<?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 The Xperience on Medium]]></title>
        <description><![CDATA[Stories by The Xperience on Medium]]></description>
        <link>https://medium.com/@lombenathan969?source=rss-110b730a08a3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*zuxu6zblotl6iLJAxqUmpg.png</url>
            <title>Stories by The Xperience on Medium</title>
            <link>https://medium.com/@lombenathan969?source=rss-110b730a08a3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 07:37:30 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@lombenathan969/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[How I use an Agentic Workflow to build Wealth Copilot]]></title>
            <link>https://medium.com/@lombenathan969/how-i-use-an-agentic-workflow-to-build-wealth-copilot-3968cc93a91c?source=rss-110b730a08a3------2</link>
            <guid isPermaLink="false">https://medium.com/p/3968cc93a91c</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[finance]]></category>
            <category><![CDATA[fintech]]></category>
            <dc:creator><![CDATA[The Xperience]]></dc:creator>
            <pubDate>Sat, 07 Nov 2026 18:22:31 GMT</pubDate>
            <atom:updated>2026-11-07T18:22:31.527Z</atom:updated>
            <content:encoded><![CDATA[<h3>How I built Wealth Copilot</h3><p><em>I’m 17, I live in Toronto, and before this project I had never written code. In one week, with Claude writing the code and me deciding what to build, I made a working web app called Wealth Copilot. It helps a financial advisor with the boring parts of the job. This article explains three things in plain words: what an AI agent is, how the system fits together, and how an AI can answer from real documents instead of guessing. That last idea is called RAG.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KnQQ6RiAnwoufA29WYYTYw.png" /><figcaption><em>Wealth Copilot front page</em></figcaption></figure><h3>Why I built it</h3><p><em>A financial advisor spends only about 20% of the day talking with clients. The rest goes to the work around the meeting: preparing for it, writing it up, answering messages, checking that nothing breaks the rules, and researching the companies clients own. My mother worked in banking, and I grew up hearing about slow systems, piles of documents and repeated checks.</em></p><p><em>So I asked one question: which of these jobs can an AI draft safely, so the advisor only has to review and approve? I focused on the advisor’s side, not the conversation with the client. The client app exists only so the advisor has someone real to work with. All the clients are fictional.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jJcdtVQCkPsT5T4N-eNmrg.png" /><figcaption><em>The whole system</em></figcaption></figure><h3><em>The whole system in one minute</em></h3><p><em>Wealth Copilot has two sides on top of one shared “brain”.</em></p><ul><li><em>The </em><strong><em>client side</em></strong><em> is where a fictional client sees a budget, goals, investments and documents, and can ask an AI assistant questions.</em></li><li><em>The </em><strong><em>advisor side</em></strong><em> is where the advisor sees a client list, notes, tasks, messages, and two AI helpers: the Advisor Copilot and the Analyst.</em></li><li><em>The </em><strong><em>brain</em></strong><em> is a separate program that decides which AI helper should answer, gives it the facts, and checks what comes back before anyone sees it.</em></li></ul><p><em>The most important rule in the whole project fits in three lines:</em></p><ol><li><strong><em>Code does the math and looks things up.</em></strong></li><li><strong><em>The AI explains and drafts.</em></strong></li><li><strong><em>A person approves.</em></strong></li></ol><p><em>Every feature in this article is a version of that rule. Nothing the AI writes is saved to a client’s record or sent to a client until a human clicks approve.</em></p><h3><em>What is an AI agent?</em></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gL7wO3Ny2EXWTIVKmYGGEw.png" /><figcaption><em>AI agent workflow</em></figcaption></figure><p><em>A chatbot answers a question. An agent does a job. The difference is that an agent can choose a step, use a tool, look at the result, and decide what to do next.</em></p><p><em>Think of a good assistant at a bank. You say, “Get ready for my 2 p.m. with Maya.” They don’t just talk. They open Maya’s file, check the open tasks, read the last messages, and hand you a one-page summary. An AI agent does the same thing with software tools.</em></p><p><em>I learned this the hard way. I started with an AI agents course and tried to build one in a tool called n8n. It wasn’t working, so I switched to building a real web app. That attempt still taught me what an agent is: a model that decides which step or tool to use next.</em></p><h3><em>Why one agent wasn’t enough</em></h3><p><em>My first assistant was a single AI that tried to do everything. It worked, but it was slow and it was hard to check. So I asked why it wasn’t a “workflow of thinking”, like the diagrams in the videos I had been watching. The brain was rebuilt around that idea.</em></p><p><em>Now a question goes through a small team, like a flowchart:</em></p><ol><li><strong><em>A safety check (code).</em></strong><em> If someone asks “Should I buy Tesla?”, the question is refused before any AI runs. Buy, sell and hold advice is not allowed.</em></li><li><strong><em>A router (AI).</em></strong><em> It reads the question and picks which specialists are needed.</em></li><li><strong><em>The specialists (AI).</em></strong><em> There is a budget coach, a goals coach, an investments guide, a money teacher and an analyst. They work at the same time, and each one only sees its own data.</em></li><li><strong><em>A combiner (AI).</em></strong><em> It merges their drafts into one answer.</em></li><li><strong><em>An answer check (code).</em></strong><em> It rejects any number that was not in the data and any advice words. If the answer fails, it is rewritten once. If it fails again, the app shows plain facts instead.</em></li></ol><p><em>Every step writes one line in a shared notebook (the programmers call it the “state”), and the app shows that list under each answer as “How I answered”. You can see exactly what happened. The tool that lets you build a program like a flowchart is called LangGraph.</em></p><p><em>[embed: node/ab31cbc5-a283]</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/586/1*pgjVh52Od-Ww1xcIlw47ug.png" /></figure><p><em>Two things made this better than the single agent. It became faster, because the router only calls the specialists that are needed and they run in parallel. And it became safer, because code checks the AI both before and after it speaks.</em></p><h3><em>RAG: look it up first, then answer</em></h3><p><em>Here is the problem RAG solves. If you ask an AI “how much can I put in a TFSA this year?”, it might answer from memory, and memory can be wrong or out of date. In finance, that is not acceptable.</em></p><p><em>RAG stands for retrieval-augmented generation, which just means “look it up, then answer”. It works like an open-book exam. The AI is not trusted to remember. Code finds the right page in a real document, hands it to the AI, and the AI must say where its answer came from.</em></p><p><em>There are four steps.</em></p><p><strong><em>Step 1: cut the document into pieces.</em></strong><em> A PDF is too long to hand over whole. So the website cuts it into pieces of about 150 words. Each piece overlaps the next a little, so a sentence at the edge is never lost. Each piece also remembers its page number.</em></p><p><strong><em>Step 2: turn every piece into numbers.</em></strong><em> This is the part that sounds like magic and isn’t. A special model reads a piece of text and gives back a list of 768 numbers that stands for its meaning. Two pieces that mean the same thing get similar numbers, even if the words are different. That is why a question about “how much can I put in my TFSA” finds a passage about the “TFSA dollar limit”.</em></p><p><strong><em>Step 3: find the closest pieces.</em></strong><em> Your question is turned into 768 numbers in the same way. The database compares them with every stored piece and returns the four closest. If nothing is close enough, the answer is simply “I don’t know”. A weak match is worse than no match.</em></p><p><strong><em>Step 4: show the source, then check it.</em></strong><em> Each piece is numbered. The AI marks what it used as [1], [2]. Code reads those marks back and shows only the sources that were really used, with the document, the page and the date. This step exists because of a mistake: an early version let the AI write “[1, 3]” when only two sources existed. The fix was not a better request to the AI. It was a few lines of code that check every number against the real list.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i97DI0EnnOOkiVWqe4GR4w.png" /></figure><p><em>The result is an answer anybody can check. A sentence with a page number and a date can be verified. A model’s memory can’t.</em></p><h3><em>The two helpers the advisor actually uses</em></h3><p><em>The client assistant is the part most people picture when they hear “AI”. But my focus was the advisor’s side, which has two helpers.</em></p><p><strong><em>The Advisor Copilot</em></strong><em> writes drafts. It can prepare a one-page brief before a meeting, turn a pasted meeting transcript into notes and a task list, and draft a reply to a client’s message. It never saves or sends anything on its own. The advisor reads, edits and approves.</em></p><p><em>Before the advisor sees a draft, a compliance check reads it. This check is plain code, not AI, so it behaves the same way every time, and it quotes the exact words it doesn’t like. I tested it with a bad draft on purpose: “you should buy more Apple stock now, it will go up 20%”. It was refused, with the problems listed. If a draft fails twice, the app falls back to a simple template with only facts.</em></p><p><strong><em>The Analyst</em></strong><em> researches a company from its official filing in six steps: pick the company, check the data, read the key numbers, see the ratios, see the risks, and write a hand-off note. The step that matters most is the second one. Before any ratio is worked out, code checks that the numbers agree with each other, that everything is in one currency, and that no part is bigger than its whole. If a check fails, the workflow stops. It never quietly fixes the numbers. The ratios and the risk flags then come from fixed rules, so a flag is a fact to discuss, never a reason to buy or sell.</em></p><h3><em>How I kept it safe</em></h3><p><em>In finance, “the app hides it” is not security. So the rules live in places that can’t be skipped.</em></p><ul><li><strong><em>The database decides who sees what.</em></strong><em> A client can read only their own rows, and an advisor only their own clients. A test with 22 checks tries to break these rules, and I ran it after every database change.</em></li><li><strong><em>Numbers come from code, never from the AI.</em></strong><em> After the AI writes, a check compares every number in its answer with the data it was given.</em></li><li><strong><em>Keys stay on the server.</em></strong><em> The browser never talks to the AI or the market data directly.</em></li><li><strong><em>A human approves.</em></strong><em> Nothing is saved to a client record or sent to a client without a click, and every change is written to an audit log.</em></li></ul><p><em>Using the live app also found real mistakes. The most serious was a hole that let a signed-in user change their own “demo” setting, which would have let someone see real accounts. The database now blocks that, and the security test proves it.</em></p><h3><em>What I learned</em></h3><ul><li><em>An agent is a workflow, not a chatbot: small steps, some AI and some code, sharing one notebook.</em></li><li><em>The AI explains and code calculates, and a check rejects any number the AI was not given.</em></li><li><em>RAG beats memory for facts, because a passage with a page and a date can be checked.</em></li><li><em>Security belongs in the database, with a test that tries to break it.</em></li><li><em>A human stays in charge: every advisor tool ends in a click.</em></li></ul><h3><em>What’s next</em></h3><p><em>No working advisor has used the app yet, so the first step is to show it to one and ask what they would really use. I also want a scored set of test questions for the advisor tools. The client assistant already has 29 questions with known answers, but the Copilot and the Analyst don’t have any yet.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3968cc93a91c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Most Expensive Problem in Banking Isn’t Math. It’s Misunderstanding.]]></title>
            <link>https://medium.com/@lombenathan969/the-most-expensive-problem-in-banking-isnt-math-it-s-misunderstanding-ba4b11563cf4?source=rss-110b730a08a3------2</link>
            <guid isPermaLink="false">https://medium.com/p/ba4b11563cf4</guid>
            <category><![CDATA[banking]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[finance]]></category>
            <dc:creator><![CDATA[The Xperience]]></dc:creator>
            <pubDate>Tue, 29 Sep 2026 04:28:28 GMT</pubDate>
            <atom:updated>2026-09-29T04:28:28.150Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*fboJIUhon8O2Uln8" /><figcaption>JPMorgan Chase building</figcaption></figure><p>When I was growing up, I spent hours watching my mom at work. She is a banker, and clients came to her desk all day with problems on their accounts. The conversations almost always took longer than they should have. Sometimes a client struggled to explain what was wrong because they did not know the banking terms. Other times, my mom had to repeat complicated rules step by step just to make sure the client understood what was happening. I understood why the policies existed, but I kept wondering why the process had to feel so exhausting for both of them.</p><p>A few years later, I saw the exact same barrier affect my uncle. He wanted to see a financial advisor about buying land and making an investment. What should have been an exciting milestone turned into weeks of back-and-forth appointments, stacks of paperwork, and repeated miscommunications. The advisor was trying to cover compliance requirements, while my uncle was simply trying to understand where his money was going. That gap in communication delayed his decisions for weeks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/735/0*GccfSy2bjhhAI6Nd.jpg" /><figcaption>image from : itech24</figcaption></figure><p>I used to think this was just a problem at our local branch. But as I started researching banking systems, I realized this breakdown happens everywhere. In many commercial banks, relationship managers spend only 25% to 30% of their working hours actually talking with clients. Meanwhile, the general public remains largely cut off from helpful financial guidance: in the United Kingdom, for instance, only 8.6% of adults received regulated financial advice over an entire year.</p><h4>The Real Problem: Access Is Solved, Understanding Is Not</h4><p>For the past decade, financial technology companies have focused almost entirely on getting people connected to financial tools. By that measure, the world has made massive progress. According to the World Bank’s Global Findex Database 2025, 79% of adults worldwide now own a financial account, up from 51% in 2011. While roughly 1.3 billion adults remain unbanked, about 900 million of them already own a mobile phone. The basic infrastructure to move money exists almost everywhere.</p><p>However, having an account does not mean someone understands how to use it safely or effectively. The actual bottleneck in finance today is not digital access; it is basic financial comprehension.</p><ul><li><strong>Low baseline literacy:</strong> Only 34% of adults across 39 surveyed countries reach the minimum financial literacy target score established by the OECD.</li><li><strong>Misunderstanding core principles:</strong> Even among people who actively hold savings products, only 46% understand how compound interest functions.</li><li><strong>A lack of comparison:</strong> Only 26% of consumers compare financial products across different institutions before signing up. Most simply accept whatever form is placed before them.</li></ul><p>Financial networks have mastered the ability to transfer funds across continents in fractions of a second. Yet the moment an everyday client sits down to discuss an investment, a mortgage, or a business loan, the conversation stalls because neither side speaks the same language.</p><h3>Speed Around the World: What Kenya and Canada Show Us</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/735/0*SSdDcY-W6ziJ70p4.jpg" /></figure><p>When my family lived in Nairobi, paying for everyday groceries or transportation did not require visiting a branch or signing paper forms. Almost everyone simply used M-PESA on their mobile phones. You sent money directly through simple text menus, the transaction cleared instantly, and both sides received a confirmation receipt on the spot. In Kenya, financial account ownership stands at 90.1%, and 52.6% of adults use mobile money every single day. The system succeeded because it met people where they already were, without unnecessary institutional hurdles.</p><p>In contrast, banking in developed economies often feels surprisingly sluggish. Canada still has no real-time payment system that connects all financial institutions. Payments Canada first promised its instant payment system, the Real-Time Rail (RTR), for 2019. After multiple vendor reviews and operational delays, the phased rollout is targeted for late 2026, with full participant adoption stretching into 2027.</p><p>This contrast shows that slow banking is rarely a software limitation. Developing nations proved years ago that instant, low-cost digital payments are entirely possible. In Western economies, legacy systems, entrenched fees, and administrative habits have simply delayed the transition toward simpler, faster systems.</p><h3>Why do we need to visit a physical bank branch</h3><p>When clients are asked why they still have to visit a physical bank branch to sign documents, they are usually told that the law requires it. But regulatory standards tell a very different story.</p><p>The Financial Action Task Force (FATF), the global standard-setter for anti-money-laundering measures, has clarified that digital identity verification “may present a standard level of risk, and may even be lower-risk” compared to in-person onboarding. In Canada, FINTRAC regulations explicitly allow financial institutions to verify an individual’s identity without requiring their physical presence, using authenticated government photo ID, credit file matching, or dual-process verification methods. The expectation that clients must sit in an office to sign forms with a pen is largely an institutional habit, not a legal requirement.</p><p>Worse, when that in-person meeting does take place, both people in the room are often frustrated:</p><ul><li><strong>Advisors are buried in data entry:</strong> Relationship managers often describe themselves as “data-entry clerks.” Similarly, a study by Kitces Research found that financial advisors spend only about 20% of their time directly meeting with clients. When asked about their single biggest operational frustration, advisors overwhelmingly cited writing and formatting meeting notes.</li><li><strong>Clients feel intimidated:</strong> The client is frequently overwhelmed by dense legal disclosures and technical terminology. Instead of asking clarifying questions, many nod along, sign the papers, and leave without understanding what risks they agreed to take on.</li></ul><p>Because the advisor is busy typing notes to satisfy internal auditing, and the client is hesitant to admit confusion, the meeting fails to achieve its primary purpose: clear, honest human communication.</p><h3>AI as a Translator, Not an Underwriter</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*9-ywicmX3I9VrGex.jpg" /></figure><p>Artificial intelligence is frequently promoted as a replacement for human bankers. But delegating financial decisions to algorithms creates serious legal, ethical, and practical hazards.</p><p>A clear example is the Apple Card, investigated between November 2019 and March 2021. When prominent cardholders reported that spouses sharing assets were granted drastically different credit limits, the public suspected algorithmic discrimination. The New York State Department of Financial Services investigated roughly 400,000 applicants and concluded there was no evidence of unlawful lending discrimination under fair lending laws. Even so, cardholders said they could not get a clear explanation of why their limits differed, because the scoring model functioned as an opaque black box.</p><p>Consumers understand this danger. According to a 2026 Gallup and Edward Jones survey, only 27% of adults trust AI models to provide financial guidance. However, research by the CFP Board shows that 52% of investors are comfortable acting on AI-assisted financial recommendations if a certified human advisor verifies the information first.</p><p>Global privacy regulations also strictly restrict purely automated decision-making. The European Union’s AI Act classifies AI systems used to evaluate creditworthiness as high-risk, requiring strict human oversight, detailed audit logs, and a clear right to explanation under Article 86. In Canada, Quebec’s Law 25 requires organizations to disclose whenever an exclusively automated decision is made, granting individuals the right to review the parameters used and submit observations to a human representative.</p><p><strong>Appropriate vs. High-Risk Applications of AI in Advisory Workflows</strong></p><p>Practical Administrative Uses (Low Risk)Unsafe Autonomous Uses (High Risk / Restricted)Transcribing meeting discussions into structured notesAutomated credit limit approvals or loan denialsTranslating financial jargon into plain language for clientsUnsupervised investment or product recommendationsPre-filling compliance disclosures from verified dialogueSolely automated risk profiling without human review</p><p>These guardrails outline the only sustainable role for AI in finance: acting as an administrative assistant and real-time translator that supports human conversation, rather than an autonomous decision-maker.</p><h3>What I Am Building, and Three Risks That Could Stop It</h3><p>My goal is to develop an in-person advisory co-pilot: an AI system that runs securely during client meetings to eliminate administrative friction. Instead of forcing the advisor to stare at a screen, the system listens to the conversation, takes accurate compliance notes, flags terms that the client might find confusing, and auto-fills standard documentation. The advisor can give their client undivided attention, while the client receives plain-language explanations of the agreements they are considering.</p><p>While the vision is straightforward, turning it into a viable enterprise solution involves three serious risks that must be tested honestly:</p><ol><li><strong>Market demand versus budget:</strong> Advisors consistently report that documentation is their primary operational headache. However, recognizing a frustrating task does not necessarily mean institutional leaders will allocate software budget to fix it. Enterprise financial institutions have long procurement cycles, and proving that an administrative co-pilot generates measurable return on investment remains an ongoing challenge.</li><li><strong>Data security and privacy compliance:</strong> Client-advisor conversations involve highly confidential financial details, including identification numbers, asset values, and tax records. Under federal PIPEDA guidelines and provincial laws like Quebec’s Law 25, recording conversations requires explicit, informed consent and robust encryption standards. Storing audio recordings or meeting transcripts on insecure external cloud servers would create unacceptable compliance liabilities.</li><li><strong>The regulatory boundary:</strong> There is a strict legal line between administrative documentation and regulated financial advice. If the software simply summarizes a client’s stated goals, it remains a helpful productivity tool. But if the system begins suggesting specific portfolios or credit strategies, it crosses into regulated advisory territory, triggering complex compliance reviews under frameworks like the EU AI Act and GDPR Article 22. Keeping the model strictly confined to note-taking and translation is essential for feasibility.</li></ol><p>Of these three challenges, navigating institutional data privacy and security remains the most demanding hurdle. Enterprise banks will not permit software in the room unless they have absolute certainty that client data remains protected, local, and completely auditable. I drafted this vision and its challenges with help from people who took the time to question my thinking: <a href="https://www.linkedin.com/in/kenshoandoheng/">Kensho Ando Heng</a>, <a href="https://www.linkedin.com/in/zhitong-celine-zhou/">Celine Zhitong Zhou</a>, <a href="https://www.linkedin.com/in/rajikhayo/">Raji Khayo</a>,<a href="https://www.linkedin.com/in/langilleeric/">Eric Langille</a> and many more that i have meet during the Nrth Festival. They helped me sharpen the ideas above, and any mistakes that remain are mine.</p><h4>Key Takeaways from my research and professionals opinions</h4><ul><li><strong>Access is largely solved, but understanding lags behind:</strong> 79% of adults worldwide hold accounts, yet fewer than 35% meet basic international standards for financial literacy.</li><li><strong>Emerging markets showed the power of leapfrogging: </strong>Systems like Kenya’s M-PESA showed that simple, accessible digital tools can transform everyday commerce when artificial institutional barriers are removed.</li><li><strong>Physical paperwork is driven by habit, not legal mandates:</strong> International bodies like the FATF and national regulators like FINTRAC explicitly support secure digital identification methods.</li><li><strong>Advisors and clients face opposite sides of the same communication barrier: </strong>Professionals lose hours to documentation and data entry, while everyday clients struggle to decipher financial jargon.</li><li><strong>AI must serve as an assistant, never the ultimate arbiter: </strong>Technology should handle note-taking and clear translations, leaving financial judgment and relationship-building entirely to human professionals.</li></ul><h4>A Question for Financial Professionals</h4><p>If you work as a financial planner, mortgage broker, or branch banker: roughly what percentage of your client meetings is spent looking down at a screen or typing notes instead of talking directly with the person in front of you?<br>I would welcome the opportunity to hear your perspective. You can share your thoughts directly in the comments or connect with me on <a href="https://www.linkedin.com/in/nathanlombe/">LinkedIn</a> as I continue researching and refining this architecture.</p><p><strong>About the Author</strong></p><p>Hi, my name is Nathan! I am a high school student passionate about fintech and building solutions to real-world inequality. Learning how technology can make banking simpler is one of my favourite things to research! You can reach me through my LinkedIn ( <a href="https://www.linkedin.com/in/nathanlombe/"><em>Nathan Lombe</em></a>).</p><p><strong>Works Cited</strong></p><p>Central Bank of Kenya. <em>2024 FinAccess Household Survey: Main Report</em>. Central Bank of Kenya, Dec. 2024, <a href="http://www.centralbank.go.ke/wp-content/uploads/2024/12/2024-FINACCESS-HOUSEHOLD-SURVEY-MAIN-REPORT.pdf">www.centralbank.go.ke/wp-content/uploads/2024/12/2024-FINACCESS-HOUSEHOLD-SURVEY-MAIN-REPORT.pdf</a>.</p><p>CFP Board. “CFP Board Survey: Investors Trust AI More Than Social Media, but Advice Still [confirm full title].” CFP Board, Aug. 2023, <a href="http://www.cfp.net/news/2023/08/cfp-board-survey-investors-trust-ai-more-than-social-media--but-advice-still">www.cfp.net/news/2023/08/cfp-board-survey-investors-trust-ai-more-than-social-media--but-advice-still</a>.</p><p>European Union. <em>Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act)</em>. Official Journal of the European Union, 2024, eur-lex.europa.eu/eli/reg/2024/1689/oj.</p><p>Financial Action Task Force. <em>Guidance on Digital Identity</em>. FATF/OECD, Mar. 2020, <a href="http://www.fatf-gafi.org/content/dam/fatf-gafi/guidance/Guidance-on-Digital-Identity.pdf">www.fatf-gafi.org/content/dam/fatf-gafi/guidance/Guidance-on-Digital-Identity.pdf</a>.</p><p>Financial Conduct Authority. <em>Financial Lives 2024 Survey: Key Findings</em>. FCA, 2025, <a href="http://www.fca.org.uk/publication/financial-lives/financial-lives-survey-2024-key-findings.pdf">www.fca.org.uk/publication/financial-lives/financial-lives-survey-2024-key-findings.pdf</a>.</p><p>Financial Transactions and Reports Analysis Centre of Canada (FINTRAC). “Methods to Verify the Identity of Persons and Entities.” Government of Canada, fintrac-canafe.canada.ca/guidance-directives/client-clientele/Guide11/11-eng.</p><p>“General Data Protection Regulation, Article 22: Automated Individual Decision-Making, Including Profiling.” <em>GDPR.eu</em>, gdpr-info.eu/art-22-gdpr/.</p><p>Gallup and Edward Jones. Survey on trust in AI for financial guidance, Aug. 2026. Reported in [confirm title]. <em>The Hill</em>, thehill.com/business/6014460-americans-distrust-ai-financial-advice/.</p><p>Kitces Research. “[Confirm exact title: how financial advisors spend their time].” Kitces.com, <a href="http://www.kitces.com/blog/how-do-financial-advisors-spend-time-research-study-productivity-capacity-efficiency/">www.kitces.com/blog/how-do-financial-advisors-spend-time-research-study-productivity-capacity-efficiency/</a>.</p><p>McKinsey &amp; Company. “Agentic AI Is Here. Is Your Bank’s Frontline Team Ready?” McKinsey &amp; Company, 2 Dec. 2025, mckinsey.com/industries/financial-services/our-insights/agentic-ai-is-here-is-your-banks-frontline-team-ready.</p><p>New York State Department of Financial Services. “DFS Issues Findings on the Apple Card and Its Underwriter Goldman Sachs Bank.” NYDFS, 23 Mar. 2021, <a href="http://www.dfs.ny.gov/reports_and_publications/press_releases/pr202103231">www.dfs.ny.gov/reports_and_publications/press_releases/pr202103231</a>.</p><p>Office of the Privacy Commissioner of Canada. “The Personal Information Protection and Electronic Documents Act (PIPEDA).” Government of Canada, <a href="http://www.priv.gc.ca/">www.priv.gc.ca</a>.</p><p>Organisation for Economic Co-operation and Development. <em>OECD/INFE 2023 International Survey of Adult Financial Literacy</em>. OECD Publishing, Dec. 2023, <a href="http://www.oecd.org/content/dam/oecd/en/publications/reports/2023/12/oecd-infe-2023-international-survey-of-adult-financial-literacy_8ce94e2c/56003a32-en.pdf">www.oecd.org/content/dam/oecd/en/publications/reports/2023/12/oecd-infe-2023-international-survey-of-adult-financial-literacy_8ce94e2c/56003a32-en.pdf</a>.</p><p>Payments Canada Real-Time Rail: “Canada’s Real-Time Rail System Will Launch This Year After Achieving Critical Milestone.” <em>BetaKit</em>, July 2026, <a href="https://betakit.com/canadas-real-time-rail-system-will-launch-this-year-after-achieving-critical-milestone/">betakit.com/canadas-real-time-rail-system-will-launch-this-year-after-achieving-critical-milestone/</a>.</p><p>Quebec. <em>Act to Modernize Legislative Provisions as Regards the Protection of Personal Information</em> (Law 25). Government of Quebec, 2021, <a href="http://www.cai.gouv.qc.ca/">www.cai.gouv.qc.ca</a>.</p><p>World Bank. <em>The Global Findex Database 2025</em>. World Bank, 2025, thedocs.worldbank.org/en/doc/be6615202d1f08a25855c8ac2d615122–0050012025/related/Little-Data-Book-2025-Web.pdf.</p><p>World Bank. “Digital Technology Is Unlocking Financial Inclusion.” <em>Development Talk</em>, World Bank Blogs, 2025, blogs.worldbank.org/en/developmenttalk/digital-technology-is-unlocking-financial-inclusion.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ba4b11563cf4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>