<?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 Amar Dahiphale on Medium]]></title>
        <description><![CDATA[Stories by Amar Dahiphale on Medium]]></description>
        <link>https://medium.com/@amar.dahiphale?source=rss-4a3f36ad88fa------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*qdEBxVZulOC8tJXA</url>
            <title>Stories by Amar Dahiphale on Medium</title>
            <link>https://medium.com/@amar.dahiphale?source=rss-4a3f36ad88fa------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 22:48:03 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@amar.dahiphale/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[The Quiet Shift: What We’re Missing in the Move to Agentic IT]]></title>
            <link>https://medium.com/@amar.dahiphale/the-quiet-shift-what-were-missing-in-the-move-to-agentic-it-75f076362717?source=rss-4a3f36ad88fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/75f076362717</guid>
            <dc:creator><![CDATA[Amar Dahiphale]]></dc:creator>
            <pubDate>Sun, 02 Aug 2026 10:36:45 GMT</pubDate>
            <atom:updated>2026-08-02T10:36:45.363Z</atom:updated>
            <content:encoded><![CDATA[<p>A few months back, a friend of mine who leads DevOps at a mid-sized SaaS company told me a story that I haven’t been able to stop thinking about.</p><p>It was 2:00 AM on a Tuesday, and a minor microservice latency spike triggered an automated alert. By the time he reached for his phone on the nightstand, opened his laptop, and checked the queue, the incident ticket was already closed.</p><p>This wasn’t just a basic script restarting a container. An AI agent had caught the anomaly, pinpointed a memory leak in a recent commit, spun up a temporary shadow cluster, rerouted the traffic, rolled back the bad deployment, and left a detailed post-mortem draft sitting in Slack for the team to review over coffee.</p><p>He sat there in the dark terminal glow, realizing the ground had completely shifted under his feet.</p><p>For the past couple of years, most of our industry conversations have been dominated by AI writing code snippets or polishing emails. But underneath the noise, a much quieter, deeper transformation is taking place across IT operations. We’re moving from <strong>Automated IT</strong> to <strong>Agentic IT</strong> — and honestly, a lot of us are looking right past the subtle clues.</p><h3>Why This Matters (Beyond the Hype)</h3><p>For decades, operational IT was built on simple, rigid logic: <em>If X breaks, run playbook Y.</em> We built massive rulebooks, endless chains of Zapier/MuleSoft triggers, and strict CI/CD pipelines.</p><p>That worked fine when systems were smaller. But modern software architectures — with microservices, multi-cloud setups, and edge nodes — have grown way too complex for manual monitoring. Things break in weird, novel ways that hardcoded rules simply can’t predict.</p><p>This is where <strong>Agentic IT</strong> comes in. Instead of following a strict flowchart, autonomous agents evaluate context, plan multi-step fixes, and handle unpredictable operational hiccups right alongside human engineers.</p><p>While headlines focus on flashy consumer AI features, the plumbing of enterprise software is being quietly redesigned. If we only judge AI by chatbots, we’re missing the actual revolution happening in the backend.</p><p>Here are three subtle shifts happening right now that are worth keeping an eye on.</p><h3>3 Quiet Signals Worth Watching</h3><h3>1. Goodbye Dashboards, Hello “Intent-Based” Tools</h3><p>For years, the sign of a serious engineering team was a wall of monitors displaying dozens of Datadog, Grafana, and AWS charts.</p><p><strong>What’s changing:</strong> Engineers are spending less time staring at graphs and more time evaluating <em>agent proposals</em>.</p><p>Instead of asking a human to wade through thousands of log lines to guess why a database is lagging, an agent tracks the raw telemetry in the background and presents a plain-English suggestion:</p><blockquote>“Checkout throughput dropped 14% due to a query lock. I can re-index Table B and scale the read-replica if you want to approve.”</blockquote><p>We’re moving away from <strong>staring at raw data</strong> and toward <strong>approving proposed solutions</strong>.</p><h3>2. Code is Being Written for Agents, Not Just Developers</h3><p>We used to think of “Shadow IT” as employees secretly using unapproved tools like Notion or Trello. Today, a very different architectural shift is taking place.</p><p><strong>What’s changing:</strong> Teams are actively refactoring internal APIs and documentation so that AI agents can read and navigate them.</p><p>Traditional APIs were built for rigid, predictable software integrations. Now, teams are adopting semantic API layers and protocols like Model Context Protocol (MCP). They are structuring internal codebases so that software agents can autonomously discover tools, read docs, and execute multi-step tasks when needed.</p><p>Making systems legible for AI is quickly becoming as important as making them clean for human developers.</p><h3>3. Rethinking How We Measure Success</h3><p>Historically, Site Reliability Engineering (SRE) teams lived and died by <strong>MTTR</strong> (Mean Time to Repair) — how fast can a human jump in and fix a broken production environment?</p><p><strong>What’s changing:</strong> Forward-thinking teams are starting to track <strong>MTTA</strong> (Mean Time to Autonomy) — how long a system can safely handle edge cases and self-heal before a human needs to intervene.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SlcFUVdZ-bT_en75.png" /></figure><h3>Practical Ways to Start Playing with This</h3><p>Nobody has this completely figured out yet, and we’re all learning as we go. But if you’re curious about how to prepare your own team or systems, here are a few practical, low-risk places to start:</p><h3>1. Map Out Your “Human Router” Bottlenecks</h3><p>Look for places in your workflow where a team member’s main job is just passing data around — reading a ticket, pulling context from two or three tabs, making a low-stakes call, and hitting “run.” Those repetitive handoffs are perfect testing grounds for agentic tools.</p><h3>2. Clean Up Your Documentation</h3><p>Agents need good context the exact same way new hires do. Instead of trying to write endless, rigid conditional rules, focus on creating clear, up-to-date architecture docs and internal wikis.</p><ul><li><em>Rule of thumb:</em> If a junior engineer gets confused reading your internal setup docs, an AI agent will get lost, too.</li></ul><h3>3. Keep Humans in the Loop</h3><p>Going agentic doesn’t mean giving an AI root access and hoping for the best. It’s all about setting reasonable boundaries:</p><ul><li>Set strict approval limits.</li><li>Require explicit human sign-off for high-risk moves (like dropping databases or editing firewall rules).</li><li>Let agents autonomously tackle low-stakes, repetitive tasks (like clearing cache clusters or spinning up dev environments).</li></ul><h3>Looking Ahead</h3><p>The shift toward Agentic IT isn’t about replacing human technical talent — it’s about eliminating the repetitive firefighting so engineers can focus on creative, high-impact problem solving.</p><p>The teams that thrive over the next few years won’t necessarily be the ones with the biggest budgets, but the ones that figure out how to build a smooth, trusted partnership between human judgment and agentic execution.</p><p><strong>I’m curious to hear your take:</strong> <em>Where in your daily workflow or infrastructure do you see the most realistic potential for autonomous agents — and what makes you hesitant to hand over the keys?</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=75f076362717" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>