<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Neal Cabage on Medium]]></title>
        <description><![CDATA[Stories by Neal Cabage on Medium]]></description>
        <link>https://medium.com/@nealcabage?source=rss-a92eff2ab4c8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*fI0pGrr3M3BVf5f8QpaOXw.png</url>
            <title>Stories by Neal Cabage on Medium</title>
            <link>https://medium.com/@nealcabage?source=rss-a92eff2ab4c8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 05:56:43 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nealcabage/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Agentic AI — The Next S-Curve]]></title>
            <link>https://medium.com/@nealcabage/agentic-ai-the-next-s-curve-c1c4e25eda30?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1c4e25eda30</guid>
            <category><![CDATA[ai-product-management]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Mon, 15 Dec 2025 12:00:57 GMT</pubDate>
            <atom:updated>2025-12-21T03:14:08.171Z</atom:updated>
            <content:encoded><![CDATA[<h3><em>Agentic AI — the Next S-Curve?</em></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*Cjf4vuVV2i1lFBeF.png" /></figure><p>Agentic AI is the next step in the evolution of Generative AI — and it has already begun. It all starts with the LLM (large language model) — not just a model that understands a prompt and produces a smart answer, but one that can plan, orchestrate, and call tools, in order to achieve a goal. In other words, it is more than a mere langauge interview between humans and computers and can actually make reasoned decisions allowing it to act like an operator that can research, call APIs, run code, update a ticket, query a system, and return with results.</p><p>The LLM has become a practical point of execution. The model becomes a “brain” coordinating a set of tools (eg web browsing, databases, internal docs, SaaS apps, and code execution). The LLM isn’t replacing those systems (yet) — it is the layer that decides what to do next, translates intent into actions, and stitches outputs into a coherent outcome.</p><p>How is Agentic AI possible if an LLM is just probabilistic next-token prediction? It is because learning to predict language at scale forces the model to internalize patterns of human reasoning, procedures, and decision-making that are embedded in text. When prompted, the model can generate plans, choose next steps, and produce structured outputs (ex tool calls) that an external system can execute. Wrapped in a control loop (eg plan &gt; act with tools &gt; observe results &gt; iterate), the LLM becomes a practical coordinator for multi-step tasks. It isn’t inherently “wise” or guaranteed to be correct, but it’s often good enough at producing useful plans and next actions, especially when combined with tools and verification.</p><p>The near-term impact is straightforward: productivity. Agentic AI begins as a high-leverage assistant that makes humans faster (drafting, research, operational automation), then steadily becomes more autonomous — handling multi-step workflows with less supervision. Over time, that shifts the human role from “doer” to manager of a small team of agents: setting objectives, defining constraints, and approving outcomes.</p><p>In this article, I’m going to outline a few use cases that seem ripe for Agentic AI, and discuss key strategic patterns and how they are implemented, in order to get a deeper sense of how this will all work. I’m documenting all of this as part of my own learning journey but also to share my knowledge with anyone who is interested.</p><h3>Use Cases</h3><p>Agentic AI is going to work best when ambiguity is low, the process is well-defined, and success can be measured with clear signals (eg. did it resolve the ticket? did it return the right record? did it pass the test suite?). That sweet spot includes a ton of valuable work: repetitive workflows, operational “glue” tasks, and decisioning that’s bounded by policy and data. I anticipate greater ambiguity will be handled well in time, but the adoption math will start with low ambiguity and work its way into greater more ambiguous problems as capabilities improve. In the near term however, here are a few key use cases that seem highly addressable, to demonstrate how this will work:</p><p><strong>Ex1: Customer Support</strong> — Agents can classify issues, retrieve the right policy/docs, propose a response, and even take constrained actions (refund requests, account updates) with HITL approvals for sensitive steps. The big win is handling the long tail of repeatable tickets while escalating edge cases to humans with better context.</p><p><strong>Ex2: Sales Prospecting</strong> — Agents can build target lists, enrich leads, draft outreach sequences, and personalize messaging based on public data and CRM history. In practice, the highest ROI tends to be “sales ops automation” (research + list building + first-draft messaging) rather than fully autonomous selling.</p><p><strong>Ex3: Data Analytics</strong> — Agents can pull data, generate SQL, run analyses, draft dashboards, and explain results in plain language — then iterate when stakeholders ask follow-up questions. This is especially powerful when the system has reliable retrieval and governance (semantic layers, metric definitions, data permissions).</p><p><strong>Ex 4: Inventory Management</strong> — Agents can monitor inventory signals, forecast stock risk, recommend replenishment actions, and automate routine adjustments within policy bounds. The trick is integrating tightly with operational systems and making sure decisions are explainable and auditable.</p><h3>Strategy Patterns</h3><p>There are a few foundational patterns that industry leaders are converging on that provides guidance around the the sort of generalized problems that Agentic AI may be well equipped to address. Here are the four patterns identified by Andrew Ng, the AIML thought leader who also popularized the term “Agentic AI”:</p><p><strong>Pattern #1: Reflection</strong> — Instead of taking the first output, the system prompts the model to critique, improve, and validate its own work. This is a cheap way to increase quality: generate &gt; review &gt; revise, often with explicit rubrics (tone, completeness, correctness).</p><p><strong>Pattern #2: Planning</strong> — The agent decomposes a goal into steps, decides an execution order, and tracks progress. Planning is especially useful when tasks require multiple tool calls and intermediate checkpoints (eg “research &gt; synthesize &gt; draft &gt; verify &gt; format”).</p><p><strong>Pattern #3: Tool Usage</strong> — The model learns when to stop guessing and call tools: search, database queries, calculators, code execution, internal APIs. This is the core “agentic” capability: the model isn’t the whole solution; it’s the coordinator that knows when to fetch, when to compute, and when to act.</p><p><strong>Pattern #4: Multi-Agent</strong> — Instead of one generalist, you use multiple specialized agents (researcher, writer, verifier, executor) that coordinate and hand off work. This pattern can boost robustness, but it also increases orchestration complexity — so it shines most on larger workflows where specialization pays off.</p><h3>Implementation Patterns</h3><p>Under the hood, agentic behavior is usually implemented<strong> as a loop</strong> (eg observe &gt; think &gt; act &gt; observe again) with guardrails, memory, and tool integrations layered in. The patterns below are common building blocks, and you can mix them together depending on what you’re building.</p><p><strong>Impl #1: ReAct</strong> — ReAct combines reasoning with actions, so the model can gather information from tools/environments mid-task and adjust its plan. This often outperforms “reason-only” approaches because the agent can correct itself as it learns new facts. Think about how LLMs already search for information online when you ask a question. That’s taking an action but that action could be anything.</p><p><strong>Impl #2: CodeAct </strong>— This is a specialized coding version of ReAct. With CodeAct, the agent treats code execution as the primary action interface. The model writes code, runs it, inspects outputs/errors, and iterates. This is particularly effective for analytics, automation, and repeatable workflows where code is the cleanest glue.</p><p><strong>Impl #3: Agentic RAG (Vector) </strong>— Traditional RAG is often a linear pipeline: retrieve context, then generate an enriched prompt that will be passed to the LLM. Agentic RAG adds decision-making to the equation. The agent can decide <em>whether</em> retrieval is needed, <em>which</em> sources to query, and <em>when</em> to retry retrieval with a refined query — frequently using a vector database as the retrieval layer.</p><p><strong>Impl #4: Routing (MCP, A2A) </strong>— Routing is the pattern of classifying intent and sending work to the right downstream tool/agent/workflow (eg FAQ vs RAG vs human escalation). As protocols standardize tool connectivity, routing becomes less “hard-coded wiring” and more “policy + configuration.”</p><p><strong>Impl #5: Human in the Loop (HITL)</strong> <strong>— </strong>HITL is the safety valve. The agent runs autonomously for low-risk steps, but requests human approval for sensitive actions (payments, access changes, external comms) or when confidence is low. In production systems, HITL is often designed as an explicit workflow state (approve / reject / edit / escalate), not a vague expectation.</p><h3>Agentic System Components</h3><p>Okay, but what do you need to consider when actually building one of these systems? The LLM is core to Agentic AI but in its current state, it is not the whole story. Here are a few key components that will also be needed to build a complete Agentic system:</p><p><strong>Memory </strong>— Memory is how an agent maintains continuity across steps and sessions (what we already tried, user preferences, decisions made). It can be short-term (within a run) or long-term (persisted profiles, project state). This is often implemented using a Vector DB (ex Pinecone) that is optimized for use with an LLM.</p><p><strong>Retrieval </strong>— Retrieval is how the agent “grounds” itself in external truth: internal docs, tickets, CRM records, policies, code, etc. Done well, retrieval turns “clever guessing” into “evidence-backed execution.” This is foundationally what RAG is all about — infusing the process with contextual knowledge and seeking citations that offset hallucination risk. This is particularly critical for internal systems, to provide contextual knowledge.</p><p><strong>Prompt Engineering </strong>— Prompting is still the control surface: role instructions, rubrics, tool policies, formatting constraints, and “what good looks like.” In agentic systems, prompts become less about clever phrasing and more about <strong>operational discipline</strong> (clear steps, clear stop conditions, clear escalation paths). Much of that prompt engineering can be programmatic however and assembled during the RAG process, looking up key information and enriching the prompt, using templates, etc.</p><p><strong>Actions &amp; Orchestration </strong>— Actions are the tools the agent can call; orchestration is the logic that decides <em>when</em> to call them, <em>in what order</em>, and <em>what to do with results</em>. This is where loops (ReAct), planners, routers, and multi-agent handoffs live. Think about all of the 3rd party SaaS systems and utilities that are available online and are introducing MCP layers to enable interaction with LLM systems. Each of these is a potential action that could be taken.</p><p><strong>Governance &amp; Evals </strong>— Governance means permissions, audit logs, rate limits, and safety constraints; evals are how you measure whether the agent is improving (task success rate, correctness, refusal quality, regression tests). Agentic AI becomes trustworthy at scale only when it’s measurable and controllable. This is going to become a very important part of Agentic AI as companies seek to deploy them at scale, and run into trust and fairness issues along the way. There are real world implications that must be considered in order for these systems to be deployed successfully.</p><h3>Future of Agentic AI</h3><p>The near future of Agentic AI will be defined less by “the perfect agent” and more by early pioneers and serve as industry “proofs of concept” and an ecosystem that forms around enabling frameworks and capabilities, as well as standards to improve reliability and adoption. A growing share of “agentic capability” will get commoditized and bundled into mainstream products, so teams won’t always need to custom-build agents from scratch — we can already see the beginnings of this, with platforms like ChatGPT providing multi-modal capabilities, native integration of 3rd party tools, and connectors to import Slack, email, and Google docs for context.</p><p>At the same time, protocols and standards will create an integration fabric that makes it dramatically easier for agents to connect to tools, data sources, and even other agents across vendors. As interoperability improves, the default approach will inevitably shift from bespoke integration, to plugging into an existing ecosystem and composing larger, end-to-end flows with far less effort. This plumbing will also make simple, personal agents far more viable and accessible. And while it’s tempting to imagine agents that “rewrite themselves,” the more realistic (and more useful) path is continuous improvement through feedback loops: capturing failures, running evals, refining prompts and tool policies, improving retrieval, and iterating on workflows.</p><p>Systems will become better over time, at proposing fixes and testing them safely, but will find a harmonious balance between self-improvement and human-driven evaluation and guardrails.. As models and architectures improve, we should also see stronger planning and decisioning, which will reduce the need for Humans-in-the-Loop (HITL) on routine work for many tasks. In reality, there are some high-risk/impact decisions that likely will always require some degree of human insight (eg legal, medical decisions) — this balance will be more driven by our social contract with technology however, rather than a result of technical inability.</p><h3>Conclusion</h3><p>The Generative AI wave has only recently arrived and already has profound implications, but the next S-curve is forming quickly. Agentic AI moves beyond content generation into execution (eg planning work, calling tools, and completing tasks). The upside is a major leap in productivity, but the challenge is building the right balance of autonomy and accountability.</p><p>It’s still early, and today’s systems are imperfect, which is why most deployments will need to keep humans in the loop. But as agentic workflows mature into more autonomous agents, and standards reduce the friction of integration, the direction is clear: humans will shift from “operators” to managers of AI coworkers, setting objectives and constraints while teams of agents handle more of the day-to-day execution. The smartest path forward is to start with bounded workflows, measure outcomes, and expand autonomy as reliability proves itself.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1c4e25eda30" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Product Manager]]></title>
            <link>https://medium.com/@nealcabage/what-is-an-ai-product-manager-e24bf94539de?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/e24bf94539de</guid>
            <category><![CDATA[ai-product-manager]]></category>
            <category><![CDATA[aipm]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[ai-product-management]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Tue, 09 Dec 2025 01:43:16 GMT</pubDate>
            <atom:updated>2025-12-21T01:35:28.652Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Types of AI Product Managers" src="https://cdn-images-1.medium.com/max/904/1*OkNztjoZg6ZHsz8XBh7gUg.png" /></figure><p>Undoubtedly, 2025 is the breakout year for AI in Product Management. As tech investments shift toward AI and away from traditional Web/App products, everyone seems to be coming to the same realization that this is an “Adapt or Die” moment. You’re either going to rise with the tide of AI products or be crushed underneath it.</p><p>But what exactly is AI Product Management? After having several conversations on the topic, I get the distinct impression that we’re not all talking about the same thing. And so I wanted to use this article to propose a few definitions that might reduce confusion.</p><h3>Parallels for Comparison</h3><p>The term “AI PM” is analogous to “Mobile PM” in 2010. It was used to indicate specialized knowledge about mobile app use cases, UI patterns, and app store marketplace dynamics, but the core value proposition was still that of a Product Manager fundamentally — to ensure value creation for users and ensure product-market fit for what was being built.</p><p>The AI PM moniker is similar — it suggests specialized knowledge about a new capability and the related use cases it enables, but at the end of the day it is still Product Management at its core, and the primary focus should be ensuring customer value is created.</p><h3>What an AI PM Is <em>NOT</em></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/817/1*2kPTzvH7NQCoAkYOGTvqvw.png" /></figure><p>Perhaps the biggest confusion that I’m hearing in conversations is that someone is an “AI Product Manager” if they’re simply using an LLM to assist with research or writing their requirements. While this is an important step forward in productivity, using AI does not mean you are managing an AI Product.</p><p>Another misconception is that AI Product Management is an entirely new set of skills such as vibe-coding a prototype or running model Evals and this usurps traditional Product work such as market research and understanding customer needs. While it’s definitely helpful to tinker and understand new capabilities, there is a danger of getting swept up in shiny new things at the expense of the core value that Product brings to the table, which has always been product-market fit.</p><h3>Layers of AI Knowledge</h3><p>In the diagram above, I describe three different layers of AI knowledge and how they relate to AI Product Management. I’ve heard these used interchangeably as different definitions of an AI PM, and so I want to disambiguate the topic.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/904/1*qKnf9EpNZEwf0aoyYCX0dQ.png" /></figure><p><em>The three types are:</em></p><p><strong>i. AI Experience PM</strong> — This is someone who manages a product that leverages AI to provide a new tool or experience for users. This could be anything from traditional ML-powered features like onsite search and personalized recommendations, to AI-native UI patterns. The PM superpower here is mastering all the use cases that AI unlocks and cataloging examples of well-executed AI products through market and customer research, then bringing those applicable use cases to market and testing your way into product-market fit (PMF). Like traditional consumer products, success for AI experiences is measured using outcome metrics (eg adoption, engagement, retention).</p><p><strong>ii. AI Capability PM</strong> — Capabilities are more technical by their nature and require going a layer deeper in the AI knowledge stack. When a company creates their own in-house AI capability, Product is often working directly with the technical team, which necessitates a deeper understanding of AI/ML concepts. It’s worth noting that this is very similar to existing Technical PM roles. In fact, Technical PMs have worked on traditional ML capabilities for decades (eg search, recommender systems, price prediction and spam detection capabilities). Internal capabilities typically track output metrics for measuring success when they’re detached from the user experience (eg Precision, Recall, F1).</p><p><strong>iii. Full-Stack AI PM</strong> — The last type of AI PM is someone who can do all of these things — they use AI tools (AI enabled), identify AI-native use cases and UI patterns (AI experiences), and work with technical teams to build internal models (AI capabilities). While it is less common that someone has full-stack AI knowledge, it can be ideal when they do, because it allows the PM to work across the full-stack and shepherd an end-to-end strategy that directly connects user outcomes and capabilities. In time, this depth of knowledge will become more common.</p><h3>So, What Should I Learn?</h3><p><em>Every</em> Product Manager should be well-versed in AI-enabled tools by now such as using LLMs and diffusion models to assist in research, writing, and diagram creation — this is “table stakes” in 2025. Beyond that, here’s what matters for managing AI products:</p><p>For an <strong>AI Experiences PM </strong>— the most important knowledge is going to be the AI-powered use cases and AI-native UI patterns. The PM needs to know what’s possible and how to map that to customer problems and value creation. I recommend going a layer deeper though, to build a foundational knowledge of what AI is and how it works, in order to build a strong intuitive sense for what’s possible. Andrew Ng has a popular course on Coursera called <a href="https://www.coursera.org/learn/generative-ai-for-everyone/paidmedia?utm_medium=sem&amp;utm_source=gg&amp;utm_campaign=b2c_namer_generative-ai-for-everyone_deeplearning-ai_ftcof_learn_px_dr_bau_gg_sem_pr-bd_us-ca_en_m_hyb_23-11_x&amp;campaignid=20730862737&amp;adgroupid=158782890201&amp;device=c&amp;keyword=ai%20for%20everyone&amp;matchtype=b&amp;network=g&amp;devicemodel=&amp;creativeid=679234260688&amp;assetgroupid=&amp;targetid=kwd-413406756375&amp;extensionid=&amp;placement=&amp;gad_source=1&amp;gad_campaignid=20730862737&amp;gbraid=0AAAAADdKX6Yja9cbRgAJ04ae4SRZvxPHe&amp;gclid=Cj0KCQiAi9rJBhCYARIsALyPDtt2gph06EAGb_23tjZIlwFqKBmlnFTxeQBgKinJsNqimGLlOrA4l3QaAsPWEALw_wcB">AI for Everyone</a>, which is a great place to start.</p><p>For <strong>AI Capability PMs</strong> — you’ll want to build a deeper understanding of AI/ML for this one. You don’t need to be a Data Scientist, write Python or do Linear Algebra, but you will be actively working with Data Scientists, Data Engineers and ML Engineers, and you’ll need to be able to speak this language. For example, what are the model types, core concepts like vector embeddings and vector search? You’ll also want to deeply understand concepts like precision and recall, since these output metrics are often how AI capabilities are measured. The <a href="https://www.coursera.org/specializations/machine-learning-introduction/">Machine Learning Specialization</a> (also Coursera) is a great resource for learning this level of knowledge.</p><p>You’ll probably also want to familiarize yourself with available in-house data that can be used in these models, who the stewards are of that data, and applicable governance and policies (eg PII, GDPR, CCPA), since 1P data is the backbone of in-house AI Capability creation.</p><h3>Conclusion</h3><p>To close, AI is undoubtedly an important innovation on the digital product landscape and it mandates that we all learn the foundations to ride the next wave of product development, or be crushed underneath it. But it’s important to not lose sight of the North Star for Product Management, which still and always will be — product market fit (creating value for users/customers).</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e24bf94539de" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Understanding AI/ML]]></title>
            <link>https://medium.com/@nealcabage/what-is-ai-ml-669f50b4c808?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/669f50b4c808</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-product-management]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Sat, 29 Nov 2025 12:00:40 GMT</pubDate>
            <atom:updated>2025-12-21T01:37:37.643Z</atom:updated>
            <content:encoded><![CDATA[<p>Artificial Intelligence has quietly reshaped how modern software is built. For decades, most systems were engineered using explicit, human-written rules: <em>if X happens, do Y</em>. That approach works well when the world is predictable, but it breaks down when patterns become too complex, dynamic, or ambiguous for humans to fully describe.</p><p>AI represents a fundamental shift in how software works. Instead of relying on hand-coded logic, modern AI systems learn from data. By observing a large volume of examples, these systems infer patterns probabilistically, allowing them to recognize, predict, and even generate outcomes that were never explicitly programmed. This data-driven approach underpins everything from product recommendations and fraud detection to image generation and conversational systems like ChatGPT, and it forms the foundation of what we now call Machine Learning.</p><p>AI is best understood as a layered set of approaches, where newer techniques build on those that came before it. This is often visualized as an “onion” model (AI &gt; ML &gt; NN &gt; DL &gt; GenAI). While newer layers enable new and more powerful capabilities, they do not necessarily replace earlier ones. From traditional recommendation systems to modern generative models, these techniques all coexist within the broader AI ecosystem, though some exceptions do exist, which we’ll cover later in the Generative AI section of this article.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/345/0*Rr-UCBcur4HwG8zs.png" /></figure><p>Artificial Intelligence as a formal field emerged in the 1950s, with the 1956 Dartmouth Workshop often cited as its point of origin. This early era laid the theoretical groundwork for many core ideas in machine learning that are still in use today. However, progress was limited for decades, as researchers lacked the massive datasets and computational power required to train these models at scale. As a result, the period from roughly the 1960s through the late 1990s became known as the “AI Winter,” marked by stalled breakthroughs and reduced investment.</p><p>By the 2010s, we finally began to see sufficient data being collected with the industry push into “big data”. As organizations began collecting large amounts of data on user behavior and site usage, this empowered many new predictive analytics methods that were previously impossible.</p><p>In parallel, a new field called Deep Learning that leverages neural networks and expands on the original neural network concept, really took off. This new approach to AI, pioneered by researchers like Geoffrey Hinton (the “Godfather of Deep Learning”), uses architectures that allow for much deeper, hierarchical analysis of data. This new approach was serendipitously enabled by the rise of GPU processors (like those from Nvidia), providing the parallel processing power needed to train these deep networks efficiently. Soon after, the race was on to unlock key use cases. This included Natural Language processing (NLP) and image recognition, through the work of Ilya Sutskever (OpenAI fame).</p><p>By 2020, we were beginning to see broad adoption of AI/ML, though it wasn’t yet apparent to most. By this point, AI/ML was present in everything from personalized product recommendations on Amazon, music recognition via Shazam, traffic routing via Waze (now Google Maps), and language comprehension capabilities of Siri and Google Translate. Up to this point, the primary focus was on discriminative tasks-pattern recognition, detection, and prediction.</p><p>What fundamentally changed in the early 2020s was the widespread accessibility of Generative AI. Rather than simply detecting patterns or making predictions, this new class of models could be used to new, high-quality text, images, and media. This was enabled by a pivotal new architecture from Google called Transformers (“ <a href="https://arxiv.org/abs/1706.03762">Attention Is All You Need</a> “, 2017). This architectural shift allowed models to process vast sequences of data far more efficiently, directly leading to the modern Generative AI boom. And within a couple years, this is when we saw the launch of Chat GPT 3 (November 2022) and a new class of Diffusion Models for media generation such as Stable Diffusion and Dall-E.</p><p>Next, I’ll go into each of the major types of AI and describe how each works and what it is used for. Below is a visualized mental map that I created, to simplify what can otherwise become overwhelming:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/944/0*Oy1aGNqP-Ox9ydn9.png" /></figure><h3>Machine Learning</h3><p>Machine Learning (ML) is broad and encompasses any algorithm which learns patterns from data to make predictions or decisions without being explicitly programmed. This is compared to traditional Symbolic systems (aka Expert Systems( like Wolfram Alpha for example (eg Mathmatica) which is primarily based on vast collections of humans-defined rules. ML includes a variety of statistical techniques that primarily fall into 3 categories: Supervised, Unsupervised, and Reinforcement Learning.</p><h3>Supervised</h3><p>Supervised learning leverages common statistical patterns to identify patterns through the use of “labeled” data. Labels are human annotated notes that accompany each data element, telling the algorithm what it is an example of. There are two main types of Supervised learning that describe how you might use this labeled data:</p><h4>i. Regression</h4><p>Regression modeling is a type of Machine Learning used to predict a continuous numerical value (like temperature or price) based on input data. The goal is to find the best-fitting function (often a line or curve) that mathematically models the relationship between the inputs and the output.</p><p>The most common algorithm is Linear Regression, but many others exist for complex relationships. When a model’s prediction differs from the actual data point, the difference is an error. The system learns by using an optimization process called gradient descent, which iteratively adjusts the model’s parameters to reduce the overall error rate until a point of convergence is achieved, making accurate prediction possible. Once achieved, we have a model that reflects that data and prediction becomes possible.</p><h4>ii. Classification</h4><p>Classification predicts a category or label for an item, based on input features associated with that item (ex its a cat since it has both whiskers and pointy ears). This is an opposite approach from Regression, which is predicting a series of data along a single dimension/feature. There are two main types: (i) Binary Classification — choosing between two options (eg <em>Spam/Not Spam </em>) and (ii) Multi-Class Classification — choosing from three or more categories (eg vs vs ).</p><p>The goal of the model is to learn a “decision boundary” that accurately separates the data points belonging to different classes. Rather than simply assigning an absolute label, most sophisticated models (eg Logistic Regression and Naive Bayes) generate a probability score that represents the model’s confidence that an item belongs to a specific class. Or, in the case of K-NN, it looks for which neighborhood an item seems to cluster into and then takes a vote as to the most common class of those items in that cluster. There are many techniques like this but the goal is the same — to convert this probability gradient into a definitive label, a classification threshold is applied. This threshold is critical because it governs the trade-off between the evaluation metrics Precision and Recall (see <a href="https://upload.wikimedia.org/wikipedia/commons/b/bc/PrecisionrecallDogExample.svg">Wikipedia diagram </a>), allowing model performance to be tuned based on the real-world cost of different types of errors.</p><h3>Unsupervised</h3><p>The next major category of Machine Learning is Unsupervised Learning, which looks for hidden patterns and relationships in data that has no human-affixed labels. This approach is especially valuable when identifying naturally occurring groupings that do not lend themselves to pre-definition, such as uncovering new market segments or finding groups of customers with shared needs that were not known ahead of time.</p><h4>i. Clustering</h4><p>The most common application of this is Clustering, which divides data into groups based on similarity. K-Means Clustering is the most popular algorithm for this task. K-Means is similar to the classification algorithm K-NN in that it relies on proximity in the feature space (vector embeddings), but its ultimate purpose is opposite. While K-NN assigns a new item to a known, pre-labeled class, K-Means works by iteratively adjusting central points (centroids) until the distance between the data points and their assigned groups is minimized, concluding with the discovery and definition of the new, unknown cluster memberships.This can be especially useful when trying to identify groups such as market segments, when you don’t have any prescribed groupings to start with. Imagine for example, taking a customer survey and identifying persona/market segments, based on similar clusters of needs.</p><p>Beyond clustering, Unsupervised also includes other key concepts like Dimensionality Reduction, which simplifies features using techniques such as Principal Component Analysis (PCA) by transforming the data to a lower-dimensional space while retaining crucial information. This can significantly simplify the data before which will make the clustering analysis far more efficient. Additionally, Association Rule Mining finds relationships between variables in large databases, often used for manual analysis in retail to identify items frequently purchased together in “market basket analysis”, though it’s unlikely to be used in production features.</p><h3>Reinforcement Learning</h3><p>Reinforcement Learning (RL) is a machine learning paradigm where an agent learns to make a sequence of optimal decisions by interacting with a dynamic environment. Unlike supervised learning (which uses fixed labels) or unsupervised learning (which finds patterns), RL involves the agent performing an action and then receiving a numerical reward signal (a positive or negative score) from the environment. The agent’s goal is to learn a policy-a strategy that maps states to optimal actions-that maximizes the total cumulative reward over time, effectively helping the system learn when ambient context is complex or incomplete.</p><h4>i. Model-Based</h4><p>Model-Based RL explicitly learns or is given a model of the environment, allowing the agent to predict future states and rewards before taking an action. The ability to simulate outcomes internally makes it very “sample efficient” (fewer real-world interactions required). A real world example is the Roomba vacuum cleaner that builds and uses an explicit map of your house to plan an efficient cleaning path is a good example of this approach.</p><h4>ii. Model-Free</h4><p>Model-Free RL skips modeling the environment and learns instead by trial-and-error, optimizing its actions or the policy itself (eg, Q-Learning). This approach is most useful in complex environments where an optimal model is difficult to create. It is like learning to ride a bike-you don’t need to know the physics behind it, you just learn which actions lead to a successful outcome. This is a common approach in training agents to play video games.</p><h4>iii. Multi-Armed Bandit</h4><p>This is one more example of Model-Free that is worth noting. With MAB, the chosen action does not affect the next state of the environment, meaning there is no long-term sequence of decisions to plan. MAB demonstrates the classic <em>“exploration versus exploitation </em>“ trade-off, deciding whether to use the best-known action (exploit) or try an unknown one (explore) and finding the optimal balance along the way. MAB is commonly used in A/B testing, particularly in contexts such as ad optimization, wherein the system is quickly trying to figure out which version is most performant and optimizes toward showing that version of the duration of the ad campaign.</p><h3>Neural Networks</h3><p>Neural Networks are a concept originally pioneered in the 1950s with the idea networks of neurons, called Perceptrons. The concept was mostly dormant however during the “AI Winter”, though the key addition of “Back Propagation”, which provides the mechanism for learning by comparing previous states, was added in the 1980s. This paved the way for future breakthroughs that would be realized finally in the mid 2000’s when data and compute power finally caught up (the Deep Learning era).</p><p>In Neural Networks, each neuron takes a set of inputs, multiplies them by their associated weights, adds a bias value, and passes the resulting value through an activation function, generating a calculated output. The network is a series of these layers of Perceptrons and the process of data moving sequentially through them is called the forward pass, which culminates in the final output layer’s activation function making the network’s prediction.</p><p>After the forward pass generates a prediction, a “loss function” quantifies the error by comparing the prediction to true answer. Backpropagation then calculates the gradient of the error with respect to every weight and every bias, working backward from the output layer to the input layer. An optimizer then uses these calculated gradients to adjust the weights and biases across the whole network. This is the continuous, iterative learning process that refines the system, making it more accurate with each cycle, similar to the iterative human learning process.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/244/0*v3FdMzaVSp_zRDew.png" /></figure><h3>Deep Learning</h3><p>By the mid 2000’s, the stage was set for the AI breakthroughs that followed. This is the point where reality caught up with theory because all of the data being collected in the “big data” era and all of the advances in compute power being made possible with cloud data centers running large quantities of GPUs and Google’s TPUs, were finally coming online. What researchers quickly realized was that the original theories were not incorrect — they just required a level of scaling to be realized, that was finally possible. Applying Neural Network at scale and thus giving the networks far greater depth than before, is the hallmark of the “Deep Learning” movement that began around this time.</p><p><em>Three notable deep learning came to prominence in the DL era (2016–2016):</em></p><h3>i. Convolutional Neural Network (CNN)</h3><p>CNNs are used primarily for analyzing images. They use a “convolutional” layer, where a small matrix of learnable weights, known as a “kernel”, slides (or “convolves”) across the entire input image. At each step, this filter calculates (dot product) with the small patch of pixels it covers (typically a 3×3 pixel patch) to check for a matching pattern. This process is repeated across the entire image, generating a feature map that indicates where that specific pattern (like a vertical line or edge) was detected. By stacking multiple convolutional layers, the network builds a hierarchy of features, and learns simple elements in early layers and then combines them into complex concepts (like eyes or entire objects) in the deeper layers. Through this process, a CNN is able to “detect” images that are present in an image.</p><p>RNNs are built for processing sequential data, such as text, audio, where the order and context of elements are crucial (not just their spatial location, as in the case of CNN). Their unique feature is the “directed cycles” which loop to find information from a previous step in the sequence to influence the processing of the current step. Similar to the human brain, this approach essentially gives the network a short-term working “memory” of past inputs, making them ideal for tasks where context is necessary. RNNs are used for things like text autocomplete, as they can predict the next word in a sentence based on all the words that came before it. Later, Long Short-Term Memory (LSTM) was added which addressed the struggle of remembering information over long sequences (direct cycles helped but were not sufficient) and this dramatically improved its ability to handle longer-form text.</p><h3>iii. Generative Adversarial Network (GAN)</h3><p>Lastly, GANs represent a really innovative new new framework and was arguably the first glimmer of the future Generative AI wave. In this model, there are two different neural networks that compete: a Generator and a Discriminator. The Generator network attempts to create realistic synthetic data (like images or music) from random input, and the Discriminator network acts like a critic, trying to correctly distinguish between the real data from the training set and the fake data created by the Generator. This process drives both models to iteratively improve: the Generator gets better at producing data that fools the Discriminator, and the Discriminator gets better at detection.</p><p>While GANs gained initial fame for early image generation, their usefulness can be quite broad, including things such as data augmentation (eg creating synthetic training examples), super-resolution (enhancing low-quality images), and image-to-image translation (e.g., turning a sketch into a photorealistic scene).</p><h3>Generative (GenAI)</h3><p>The Generative AI era is defined by the ability of AI models to create coherent, and original content of different “modes” including text, images, video, and code. Technically, the ability to generate assets began earlier with models like Generative Adversarial Networks (GANs) in 2014, but the revolution was truly unlocked by the Transformer architecture in 2017, following the publication of Google’s seminal paper: <a href="https://arxiv.org/abs/1706.03762">Attention is All You Need </a>.</p><p>The Attention Mechanism is a new capability, allowing for parallel processing of data and enabling the training of models on enormous data sets, leading to the development of highly sophisticated “Foundation Models” that can generalize across numerous tasks. Stemming from this major breakthrough is an entirely new class of models that take advantage of Transformers to leapfrog past their Deep Learning predecessors, from just one decade prior. The two primary model types in the Generative AI era are:</p><h3>i. Large Language Models</h3><p>LLMs are the successor to Recurrent Neural Networks (RNNs) for sequential data processing, primarily text. Built entirely on the Transformer architecture, they leverage the attention mechanism to capture long-range contextual relationships within massive datasets. The architecture is split into two dominant forms: Encoder-only models, like BERT, which focus on reading the entire text bidirectionally to deeply understand and encode context for analysis; and Decoder-only models, like the GPT genre (eg ChatGPT, Claud, Gemini), which excel at generating text by predicting the next word in a sequence. These models function as advanced statistical prediction machines, allowing for fluid conversation, sophisticated writing, and code generation at scale.</p><h3>ii.Diffusion Models</h3><p>For media generation, Diffusion Models are the new standard for synthesizing high-quality visual and temporal media such as video, replacing GANs for image and video generation. Well known examples include MidJourney, Stable Diffusion Dall-E, and Google’s Nano Banana. These models work through s process of systematically destroying an image and then reversing it, to learn how to recreate it. To achieve their high fidelity, Diffusion Models often incorporate the Vision Transformer (ViT) architecture as their core backbone, particularly for video, which requires temporal attention tracking (adding a 3rd dimension to attention, for time). This allows the generative process to leverage the Transformer’s attention mechanism for modeling global relationships across the pixels of an image or the frames of a video, ensuring the generated content is highly coherent and complex.</p><p>The power of Large Language Models (LLMs) extends far beyond mere text generation — they are now capable of taking complex, high-level instructions, understanding user intent, and acting as an orchestrator across multiple tasks and systems.</p><p>The initial step was enabling Multimodal Generative AI, where the LLM routes a creation request to a specialized model, such as passing a prompt to a Stable Diffusion model-to execute a task it cannot do itself. This concept, however, has much deeper implications, as the LLM is not merely a translator; it is an Agent orchestrator capable of calling any tool necessary to achieve a goal. This can include web searches, running code, querying databases, or interacting with any 3rd-party service via MCP/API. The ability to understand, plan, act, reflect autonomously on behalf of the user. likely defines the next major wave of innovation beyond foundational Generative AI. And almost as if its right on queue — AI research has recently turned to reasoning and world models, which appear to represent the next level of cognition (See: LeCunn’s and Sapient’s ).</p><p><em>Note — the world of Agentic AI is very interesting and active — I’ll probably write a separate article on this topic alone (stay tuned!).</em></p><p>Finally, I want to close this article by sharing a compilation of use cases that I put together, to really drive home my understanding the various algorithms and use cases of those algorithms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/942/0*E5rbvpObqHliJtXf.png" /></figure><p>Wrapping up, AI is a data-driven predictive approach to building systems. By gathering data and modeling it properly, it is possible to create systems that excel far beyond the need for humans to write declarative rules. This foundational shift from prediction to creation has opened up a world of exciting possibilities. Although Machine Learning had been in production use on popular sites and apps since the mid-2010s, it only recently captured the imagination of culture at large, when we entered the Generative AI era, starting in 2017, and particularly as punctuated by the release of ChatGPT in November 2022.</p><p>Research continues and the rate of development of this capability is expected to only increase. Meanwhile, there are many exciting possibilities and use cases that have yet to be fully realized, even if you were to take a snapshot of what’s possible today alone. And it you put all of this together, it is a wondrous future to consider, and a varietal playground for people who work in the space of innovation and digital products.</p><p><em>Originally published at </em><a href="https://nealcabage.com/wp-content/html/ai-ml.html"><em>https://nealcabage.com</em></a><em> on November 29, 2025.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=669f50b4c808" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[3 Pillars of Product]]></title>
            <link>https://medium.com/@nealcabage/3-pillars-of-product-861680d9441d?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/861680d9441d</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-skill]]></category>
            <category><![CDATA[product-development]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Sun, 20 Dec 2020 12:00:42 GMT</pubDate>
            <atom:updated>2025-12-21T00:35:25.674Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/836/0*AfBSllB9GrFwJK-p.png" /></figure><p>The core functions of <a href="https://nealcabage.com/product-management/">Product Management</a> are straight forward, clustering around a primary objective of Product-Market fit. If you start with a Northstar of generating business value by creating solutions that solve real problems and general real demand from customers, then what you need to do is obvious. Yet the adoption of Product Management into software organizations has conflated this truth with the tactical role of a Product Owner, which arguably is closer to Program Management than traditional Product Management. When you combine that with a lack of clear definition and it seems the role of Product is a little different in every organization.</p><p>To this end, I thought it might be helpful create a simple model that defines this role, and the areas we should really be paying attention to. With that, I introduce the 3 Pillars of product Management: Here’s a quick discussion of each of the three areas:</p><p><strong>I. Opportunity Discovery</strong> — This is the heart &amp; soul of Product Management, since this where we really understand the big picture about our users and their needs. We perform market research to understand the competitive set and we work with UX/UXR to conduct customer surveys and interviews to understand their needs, and bring opportunities back prototype and size with Finance. This ideally is an ongoing stream of work that informs regular planning, but it can also kick off major new initiatives when new opportunities are identified.</p><p><strong>II. Product Planning</strong> — Planning starts with an articulation of the vision, and creation of goals to determine prioritization. Once we have those in place, we can prioritize development initiatives according to what is going to be most impactful toward our vision and our goals. This is our opportunity to be strategic and make sure the entire team is aligned around the most important work.</p><p><strong>III. Feature Development</strong> — Lastly, development is where it all comes together into a tangible product. By this point, we’ve identified the most important initiatives and when they need to happen. Now, each initiative needs to be defined with product requirements, and the user interface needs to be designed and validated. And depending on process and the size/type of initiative, this is also where PM’s will get more involved with the development teams, answering questions, performing UAT, setting up and calling A/B testings, etc.</p><h3>End-to-End Process Flow</h3><p>The following diagram shows a typical step-by-step Product process and how the 3-pillars model overlays with, to put it all together. As you can see, market analysis leads to opportunities in the discovery phase, which leads to prioritization and the creation of a roadmap), which ultimately feeds the product development backlog (user-story level priorities) and a release plan for when increments will be delivered.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-sBIMl-ng5tpD6pJ.png" /></figure><p><em>Originally published at </em><a href="https://nealcabage.com/wp-content/html/3-pillars.html"><em>https://nealcabage.com</em></a><em> on December 20, 2020.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=861680d9441d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What is Product Management?]]></title>
            <link>https://medium.com/@nealcabage/what-is-product-management-b268293ce95c?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b268293ce95c</guid>
            <category><![CDATA[product-management-career]]></category>
            <category><![CDATA[product-management]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Sun, 20 Oct 2019 12:00:53 GMT</pubDate>
            <atom:updated>2025-12-21T23:47:28.616Z</atom:updated>
            <content:encoded><![CDATA[<h3>Clarifying the Role of Product Management</h3><p><em>Have you ever noticed that Product Management is different at every company? To the point that it can look like a completely different role with non-transferrable skills. At Company A it is about figuring out what to build and crafting long strategy documents. At Company B, there are no documents at all but a strong emphasis on delivery. I feel like this lack of a clear anchor is a weakness for the role, and so I’d like to take a pass at defining what I think the role should be and why.</em></p><h3>What is product Management?</h3><p>When thinking about Product Management as a role, it is helpful to begin with a definition of the object which we are managing — the product. I <a href="https://nealcabage.com/what-is-a-product/">define a Product</a> as <em>a solution offered to a market, to enable or make efficient a specific customer need or desire. </em>Implicit in this definition, Product Managers identify an under-served market segment, understand their needs and how to solve them, figure out how to differentiate the solution in the marketplace, and then take that solution to market at scale, such that we’re creating a valuable solution at a relatively low cost that will resonate with a broad audience.</p><p>This definition is intentionally rooted in a Marketing perspective, which is the origin of Product Management. In fact, if you go back to the 1930s, Product Management began in the Brand Marketing department of Procter &amp; Gamble, as a way to vertically scale the Brand Marketing department, allowing individual “brand men” as they were called, to go deep and truly own every aspect of a product, from understanding their market, to production, and taking that product to market. Inherently then, Product Management is looking at the Product from the perspective of the Marketing Mix 4P’s.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*67ElbPSTcXfI8XXF.png" /></figure><p>The mission of the Product Manager in this context, was to understand everything about their market and then ensure the correct product was built. Product would speak with customers, go deep on customer and competitive research, and spend time in the field to understand why customers behaved as they do. And thus, the value proposition of the Product Manager to the organization was to ensure the success of the product, principally through ensuring product-market fit, and reducing the risk of investing significant resources into building something that no one would buy. The Product Manager would then shepherd the product through development and taking the product to market, ensuring the end-to-end loop was successful, and adapting or correcting course where needed.</p><h3>History of Product Management</h3><p>The above description was the role of the Product Manager for the majority of the 1900s and you’d primarily only find Product Managers at CPG companies such as P&amp;G until consumer software products began rising to prominence in the late early 2000s in the Silicon Valley — and that’s when everything really changed.</p><p>In 2001 the <a href="https://agilemanifesto.org/">Agile Manifesto</a> pulled together a number of similar concepts and methods into a common mission, in juxtaposition to traditional Waterfall methodologies, to move toward small-batch incremental development. Software is inherently different from physical products insofar as it is much easier and less expensive to make small incremental changes and deliver those to the market, which means software products were able to shift to a model of providing a constant stream of incremental improvements rather than having to wait for large batches.</p><p>Scrum is the popular Agile Project Management framework that emerged around this time as well, and with it, a definition of two important roles within a product development team — the Product Owner and the ScrumMaster. Because this was a project management framework for incremental and continuous development, that means Product Management would no longer develop large Product Requirement Documents (PRDs) and throw them ‘over the wall’ to Engineering to implement in a different department. Instead, the Product Owner began co-locating as part of the team and developing requirements iteratively as well.</p><p>This new approach has its pros and cons but one notable change is that this shifted the person playing the product role from being a strategic Product Manager primarily looking at product-market fit, to a more tactical role that defines, supports, and clarifies requirements on a real-time basis. And because this is in the context of software development, there is never a shortage of clarification needed for nuanced details for various use cases.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*X8Iw-0tbnxvGQ3cL.png" /></figure><p>Finally, in the 2010s, an awakening began in the software industry, when two seminal books (Four Steps to the Epiphany and Lean Startup) introduced Lean Methodology, as a way to incrementally address product-market fit (PMF) in a way that meshed with Engineering’s new Agile development principles. Interestingly, many of these techniques for iterative product discovery were pioneered much earlier by Toyota in the 1960s, but with re-discovery, new life was given to the role of Product Management, as the value and compatibility of the role in the new Agile paradigm, became clear to the rapidly growing software industry. The number of people holding the title rapidly increased and Product Management, often referred to as a mini CEO of the product, was soon the most <a href="https://www.wsj.com/articles/coveted-job-title-for-m-b-a-s-product-manager-1456933303">popular career path</a> for recent MBA graduates.</p><h3>Product Management Activities</h3><p>With the history of Product as a backdrop, I think of Product Management as having three main functions: (i) Opportunity Discovery in which we understand customer needs and market dynamics, (ii) Product Planning in which we assess the opportunities, define a rational prioritization framework, and create a roadmap, and (iii) Product Development, in which we work with our technical and design partners to realize the product we will eventually take to market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HS7fb_SxhSsOpAow.png" /></figure><h3>Definition of Success</h3><p>Because Product Management is a strategic role, meaning that success should be measured in terms of outcome, not output. A product is successful if it achieves an adoption or engagement goal, not because x number of features were delivered.</p><p>This is a paradigm shift from other roles such as <em>Program</em> or <em>Project</em> Management and so it is important to call out. In fact, even the Agile Product Owner is sometimes positioned to be more outcome/delivery driven, which is a scrum team role often played by IC Product team members, which further engrains the confusion. So, keep this important difference in mind — it is foundational to driving impact.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*nKCiZ4TJn6Oz52Pe.png" /></figure><h3>Where Does Product Strategy Fit?</h3><p>If we agree that Product Management is about driving outcomes which makes it a fundamentally strategic role, then where does the product strategy fit within an organization that already has a business strategy?</p><p>The simple answer is that strategy cascades within an organization from Corporate strategy to Business and down to the Functions of the business (eg Marketing, Finance, etc). The Product strategy is one of those functional strategies. It inherits key inputs from the cascaded Business strategy such as goals and responds with a plan to achieve those goals, by addressing customer needs. This is then passed further down to the operational/action plan, which describes what will be delivered and when.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/899/0*ObW8n1Lv7gL-usal.png" /></figure><p>Having put these layers into perspective, Product Management is not working in a vacuum, nor are we defining the business strategy. Rather, Product is working to deliver on the Product Strategy that supports the Business Strategy — with a mandate from Business to build a product in some general direction, it is Product’s role to figure out what that right product is to build, in order to achieve the goals of the Business strategy.</p><p>This structure is a good point of reference for how Product is ideally situated within a larger organization but naturally, this isn’t always the case. For example, in a startup , there is not nearly so much structure as this and Business strategy vs product strategy are often blurred together (the Business Model Canvas, for example, is a popular tool of startup product Managers). And still at other companies, for reasons I explain here, Product might not even be its own standalone, often rolling up to either technology or another functional group.</p><h3>Variability of the Role</h3><p>The last point I want to touch on while defining the role of Product Management, is why the role varies so much from company to company. Because of my role in founding and leading <a href="https://nealcabage.com/pmala/">PMA.LA</a> for several years, I had the opportunity to interact with hundreds of Product Managers and I’ve observed a few patterns that affect how Product Management is practiced:</p><ul><li><strong>Type of Product</strong> — The nature of what you’re building varies dramatically from B2C to B2B to internal. B2C, for example, will place much greater weight on a sleek user experience, whereas an internal tool or platform may place little to no value in doing so. Discovery is also very different — with B2C you must discover what they want. With B2B or internal, your key users are more than happy to tell you what they want, and delicately saying no becomes a critical skill.</li><li><strong>Seniority Level</strong> — Many junior Product Managers have great hopes of driving strategy and being CEO of their product but in truth that responsibility often lies with someone more senior in the Product organization. The role of a junior Product Manager is often more ‘heads down’, focused on feature definition and development.</li><li><strong>Product-First Organizations</strong> — Organizations who are known for their products rather than their service or arbitrage (buy/sell) activities, naturally place Product Management at the tip of the proverbial strategic spear. As such Product Management has much greater political weight to drive the strategy for the company. At service and trade organizations, Product is often supporting a strategy that is defined by another Business function, and this positions the role to be more tactical than strategic.</li><li><strong>Orthogonal Roles</strong> — Because Product Management is charged with the success of their team, that often means Product will act as the ‘glue’ and fill in various orthogonal roles where help may be needed but isn’t a core function of Product. Project Management, UX, Analytics are, and Product Ops are common areas where Product Management may fill in. The risk is that one of these orthogonal roles can take over a disproportionate amount of time and take away from core Product responsibilities. This is especially the case when a dedicated ScrumMaster or Program Manager are not available to support the team.</li><li><strong>Product Lifecycle</strong> — The role of Product at a startup is almost entirely focused on product market fit of the ‘MVP’ product. More mature products trade the need to find market fit, with the need to instead satisfy the many stakeholders that have grown up around the successful product and each has their own feature requests to support their efforts. This can shift the role from being very outward and strategically focused, to being more focused on internal diplomacy.</li><li><strong>Reporting Hierarchy</strong> — If Product is positioned to report directly to Business rather than being positioned under another business group, that positions Product to lead a functional strategy, rather than supporting one. Similarly, placing Product under a business function vs under Tech/Engineering means the role (while supporting in either of these cases) could be much more Business or Technically oriented.</li></ul><h3>Conclusion</h3><p>In summary, Product Management is fundamentally about understanding the needs and desires of the customer segment you’re serving, figuring out the right product to build to optimally address those needs, and working with a cross-functional team to continually build and refine that product, to ensure it is maximally successful in the market. The role of Product Management can vary significantly, depending on how the Product function is positioned within the organization, the maturity of the product, and the type of product being developed. Fundamentally, however, Product Management is best positioned to create value, when the focus is placed upon driving an outcome, rather than merely delivering an output.</p><p><em>Originally published at </em><a href="http://nealcabage.com/wp-content/html/prod-mgmt.html"><em>https://nealcabage.com</em></a><em> on October 20, 2019.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b268293ce95c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Product Team Competencies]]></title>
            <link>https://medium.com/@nealcabage/product-team-competencies-de260f1af501?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/de260f1af501</guid>
            <category><![CDATA[product-team]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-skill]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Wed, 12 Dec 2018 12:00:05 GMT</pubDate>
            <atom:updated>2025-12-21T00:33:38.817Z</atom:updated>
            <content:encoded><![CDATA[<p>Product Management isn’t the same at every company. Different orgs, product types, and team structures lead to confusion about what skills a Product Manager should actually have. When I was learning the craft, I struggled with this myself — and later, as a leader, I wanted a way to clearly articulate the full spectrum of product capabilities.</p><p>To address this, I developed a Product Competencies Framework — a map of the core skills that effective Product Managers need. This framework doesn’t assume every PM must excel at everything. Rather, it provides a way to visualize the range of skills, Evaluate team strengths/gaps, and support coaching and team development.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/908/0*6RSRs8GntwdtAa1D.png" /></figure><p>The Product Competencies framework describes all of the skills commonly associated with Product Management. It does not assume every PM should be able to do all of these things, but it provides a way to think about what an individual PM should be good at, if they want to focus in one direction or another. It also provides a map for considering how to balance out a team, making it easy to see if you’re heavy on certain skills but have gaps in other areas on your team.</p><p>Junior Product Managers tend to focus more at the bottom of this chart whereas senior PMs who are thinking strategically, are operating in the upper part of the chart (especially team leads, Directors). Technical PMs who work on internal tools and platforms, are generally focused on the right side, whereas consumer facing products on the left. From this, you could derive the quadrant where each PM should be focused (eg upper-left for a senior B2C PM, lower right for a junior internal tools PM).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1eDOWF-X2mNZF_jL.png" /></figure><h3>Team Assessment</h3><p>This model can be useful for evaluating, communicating, and improving the overall capabilities of your team. It can also be used to think through skills gaps on the team or what type of PM is needed for a given type of product role that you’re hiring for. For example, if your team is lacking in external market strategy, the model might be drawn as follows, and this makes it easy to illustrate where the deficiencies are that a team can rally around in the coming year.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/908/0*pA5cnV4LevcfWpPR.png" /></figure><p><em>Originally published at </em><a href="https://nealcabage.com/wp-content/html/competencies.html?2"><em>https://nealcabage.com</em></a><em> on December 12, 2018.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=de260f1af501" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Business Model Archetypes]]></title>
            <link>https://medium.com/@nealcabage/business-model-archetypes-d552a2e2eb37?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/d552a2e2eb37</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[business-models]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Fri, 12 Jan 2018 12:00:06 GMT</pubDate>
            <atom:updated>2025-12-21T00:32:17.819Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*DCv3frrI9VigB2hW.png" /></figure><p>The Business Model Archetypes are seven fundamental business “personalities” upon which any business model can be developed. By providing the context of all available models, it becomes easier to see how businesses relate and directions in which businesses can pivot. In this article, I describe the model and the archetypes, that I first introduced in <a href="https://nealcabage.com/book/">The Smarter Startup</a>, as well as describe use patterns where the models might benefit entrepreneurs and product strategists who use the model.</p><h3>Background &amp; Context</h3><p>I’ve noticed over time that business models often become confusing because fundamentally different ways of creating value are blended together. Teams will talk about products, services, platforms, and marketplaces interchangeably, even though these represent very different underlying business behaviors. While many companies exhibit characteristics of multiple models, one mode of value creation typically dominates.</p><p><em>Below are a few of the common business models…</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*p9eXCh1HJtvZoErp.png" /></figure><p>The Business Model Archetypes framework was created to make these underlying patterns explicit. It is a conceptual model designed to describe how businesses create, deliver, and capture value at a structural level. The goal is not to force companies into rigid categories, but to provide a small set of recognizable templates that make strategic conversations clearer and more precise.</p><p>Like other archetype-based frameworks (eg Carl Jung), this model abstracts complex real-world behavior into simplified forms. Most businesses will combine elements of multiple archetypes, but understanding which archetype is primary helps teams reason about strategy, identify sources of advantage, and evaluate potential pivots or expansions. The value of the model lies in clarity rather than precision, enabling better alignment and decision-making across product, strategy, and leadership discussions.</p><p>By understanding the fundamental patterns, we can create a unified and radically simplified model that contextualizes the possibilities. There are 3 primitive archetypes, and 4 hybrids that come from combining the primitives.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*voQ0qkzwyncTDx1S.png" /></figure><h3>Primary</h3><ul><li><strong>Product</strong>: one-time purchase of an artifact</li><li><strong>Service</strong>: manually doing something and charging a fee</li><li><strong>Trade</strong>: connecting buyers and sellers for commerce</li></ul><h3>Secondary</h3><ul><li><strong>Brokerage</strong>: providing trade as a service</li><li><strong>Subscription</strong>: productizing and semi-automating a service</li><li><strong>Marketplace</strong>: productizing trade with a self-service platform</li><li><strong>Ecosystem</strong>: platform that combines all three (Mature)</li></ul><h3>About the Archetypes</h3><p><em>Here’s a brief description of each of the archetypes to demonstrate further:</em><br> <strong><em>Indicative Attributes</em></strong></p><p>The primary archetype of Trade focuses on connecting buyers and sellers. Money is earned by buying a product for less than it is sold for. The trader’s primary occupation is sourcing something of value “in the field”, packaging it, and making it readily available to those who desire it. Prototypical examples include the eCommerce retailer and lead generation. In both of these examples, the Trader sources something, qualifies and prepares it, then sells it to a customer. In the case of lead generation the behavior is the same as retail, except that the product is information about a customer prospect.</p><p><em>Originally published at </em><a href="https://nealcabage.com/wp-content/html/BMA.html?2"><em>https://nealcabage.com</em></a><em> on January 12, 2018.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d552a2e2eb37" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Are Some People Lucky?]]></title>
            <link>https://medium.com/@nealcabage/why-are-some-people-lucky-413a638ec85c?source=rss-a92eff2ab4c8------2</link>
            <guid isPermaLink="false">https://medium.com/p/413a638ec85c</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[product-market-fit]]></category>
            <category><![CDATA[psychology]]></category>
            <dc:creator><![CDATA[Neal Cabage]]></dc:creator>
            <pubDate>Tue, 18 Oct 2016 12:00:43 GMT</pubDate>
            <atom:updated>2025-12-20T18:38:36.792Z</atom:updated>
            <content:encoded><![CDATA[<p>You are not imagining it — some people truly do have better, more predictable outcomes than others. Some serial entrepreneurs may have several successful hits in a row, and others may struggle through several businesses and never get traction.</p><p>The popular thought is that if you work hard, you will succeed. This may be true for many smaller service businesses-an accounting firm, say, or a dental practice-since there isn’t a lot of mystery about market timing or product-market fit. For these businesses, the model is well defined and there are numerous predecessors that serve as models. For innovation-driven businesses, however, the outcome is much more uncertain. You can work hard, but you may not succeed. Even more vexing are stories about reluctant entrepreneurs who succeed despite themselves.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xxKuS5L44YkLp4mS.png" /></figure><p>Clearly, there is more to the story than simply working hard. The key appears to be your ability to encounter and recognize opportunities. In a 1999 psychology experiment at Harvard University known as “Gorillas In Our Midst,” Dr. Daniel Simons asked a number of study participants to watch a group dribbling basketballs and to count the number of dribbles. Halfway through the video, a six-foot tall man in a gorilla suit walked slowly through the spectacle, stopped in the middle, and pounded his chest a few times, before walking away. Afterwards, the study participants were asked who saw the gorilla. Surprisingly, more than half of them did not.</p><p>A later study by European Sociologist Dr. Richard Wiseman expanded on this research in a book called <em>The Luck Factor</em>. In Wiseman’s seminal study on the subject, he asked a group of participants to count the number of images inside a newspaper. They were given two minutes to complete the exercise, and most people used the full amount of time given to count and doublecheck their answers. A couple of participants however confidently completed the activity within mere seconds, both of them having the right answer. It turns out that, on page two of the newspaper, Dr Wiseman had taken out a one-half page advertisement that read, “Stop counting. The answer is 43. Tell the experimenter you have seen this and win $250.”</p><p>These studies illustrate a concept called inattention blindness. When we’re too focused on solving problems, we often do not observe the opportunities right in front of us. Ironically, the answer to improving your luck, then, is not working harder-it actually might be the opposite. If you find that you are always doing “heads down” work and frustrated that you’re not reaching your goals, it may be time to turn off the computer and go to an industry networking event or mingle with your customers. Have a drink and engage in conversation. Not only might you have the opportunity to reframe the problem in a way you hadn’t thought about before, you might even make a few connections who can help you reach your destination a little faster.</p><p>When you are trying to innovate and create new opportunity, it is important to be able to step back to see the big picture. Sometimes, it is only when you’ve stepped back and stopped thinking about the problems you’re trying to solve that you can see things from a new perspective. So don’t forget to take a little time to step back from productive work. You might feel guilty, but it could be the only way you’ll see opportunity right under your nose.</p><p><em>Originally published at </em><a href="https://nealcabage.com/wp-content/html/lucky.html?2"><em>https://nealcabage.com</em></a><em> on October 18, 2016.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=413a638ec85c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>