A conversation with Alex on building Era and Context, rethinking finance through AI agents, fiduciary design, and the future of money management.
Alex, could you walk us through your professional journey and what led you to focus on building AI-driven solutions in financial services?
Two threads from my career meet at Era.
The first is fiduciary advice. At AKQA I was a CTO-for-hire, and one of my clients was Coutts: private wealth in London and Zurich, human advisers, portfolios in the tens of millions. Done well, that kind of advice is life-changing. It’s also been out of reach for almost everyone.
The second is interfaces. At Square and then Stripe, where I ran stripe.com, I learned the interface isn’t a layer on top of a product — it is the product.
The consumer finance interface has always been built by the institution, around the institution’s products. AI flips that. The interface is now the user’s: conversational, personal, able to act on their behalf.
Era is where the threads meet. We’re delivering the kind of fiduciary attention I saw at Coutts, but at consumer scale, through an AI agent the user already uses every day. This wasn’t buildable two years ago. MCP and tool use changed that. AI could talk about your money. Now it can move it. That’s the line we’re building on.
From your perspective, where do current AI tools fall short when it comes to accessing, interpreting, and acting on real financial data?
AI assistants are excellent at general financial reasoning. They’re blind to your specific situation, not because AI is bad but because personalization requires data. Ask ChatGPT if you can afford a vacation and the answer is about a hypothetical person who roughly resembles you. The user still has to do the work.
The second problem is fragmentation. The average person has five accounts across three apps, with inconsistent categories and patchy data. Even if an AI could see all of it, it couldn’t make sense of it without a cleanup layer.
The third problem is the one most fintechs are getting wrong. People tolerate one bank because switching is painful: direct deposits, autopay, history. Switching AI assistants takes one click. The data layer needs to be persistent; the AI layer needs to be swappable. Any product that ties your financial life to a single chatbot is building on the wrong axis.
Era is the persistent layer. We connect the accounts, clean the data, and hold the rules, memory, and goals so whatever AI the user prefers can actually act on them, not just describe them. The financial context follows the user, not the assistant.
What are the core infrastructure challenges that must be solved to enable AI agents to safely and effectively operate in consumers’ financial lives?
Three challenges, in increasing order of difficulty.
The first is connection. Getting secure, comprehensive access to a user’s actual accounts — banks, brokerages, credit cards, the lot — is the unglamorous foundation. The financial system wasn’t designed to be read by software, let alone by AI. Most visible AI finance products today run on partial coverage: a few major banks, no brokerages, no audit trail.
The second is memory. An assistant without memory starts over every conversation. That’s fine for a recipe; it’s useless for money, where context (goals, patterns, history) is the entire point. And memory has to be portable. The user will switch AI tools. They shouldn’t have to re-explain their financial life every time they do.
The third is the one almost no one is solving: action with accountability. Moving money on a user’s behalf is a regulated activity. If an AI is doing it, a regulated entity has to stand behind every action, one that takes legal responsibility for the advice. Otherwise it’s unsupervised investment guidance with nobody on the hook.
What does it mean to be an “AI-native” Registered Investment Adviser, and how does this differ from traditional RIAs in both philosophy and execution?
A traditional RIA is built around a human adviser, a CRM, and a small book of clients. The product is the adviser’s judgment. The economics only works above an account-size threshold — usually $500K — which is why genuine fiduciary advice has long been a luxury good.
AI-native means taking the same regulatory structure (SEC registration, fiduciary duty, duty of care) and building it around software instead of a human adviser. Era Financial Advisers LLC is a registered RIA. The firm holds the fiduciary obligation: the same duty of care that sits behind a private banker at Coutts. Humans build the system, supervise it, and remain accountable. What’s automated is delivery, not responsibility. Because delivery isn’t priced by the human hour, the math works at every account size.
The philosophical difference is who gets fiduciary attention. The execution difference is that everything is built around a user talking to an AI agent, not reading a dashboard. That changes what you log, what’s writable, and how disclosures work.
The financial system is complex on purpose, and that complexity is most of why ordinary households leave money on the table: unclaimed 401(k) matches, idle cash, fees they never see. AI-native is how someone earning $80K finally gets the same caliber of attention as someone with $5M.
How did Era approach SEC registration, and what lessons did you learn about aligning innovation with compliance?
We pursued SEC registration early, before launching anything that would qualify as investment advice. The subsidiary structure was deliberate: Era Financial Advisers LLC is the adviser of record for our paid users, so regulatory responsibility is unambiguous from day one.
Two lessons.
The first: compliance is cheaper as architecture than as retrofit. Most companies in this space build the product, then figure out the regulatory wrapper later. That works until it doesn’t, and when it doesn’t, the cost is sometimes existential. Building the structure first felt slow at the time. It’s the only thing that’s let us move quickly since, because we’re not constantly negotiating with our own past decisions.
The second: the SEC is more open to novel models than founders assume. The current Division of Investment Management has explicitly framed itself as “The Innovation Commission” and asked firms with new approaches to come in and talk. That’s a posture shift founders should take seriously. Showing up with a clear theory of your activities (what’s regulated, who’s responsible, where the user is protected) is the difference between a productive conversation and a defensive one.
What inspired the launch of Context, and how does it fundamentally change the way AI agents interact with sensitive financial data?
We kept watching people ask AI assistants questions about their money: “Can I afford this?” “Did my paycheck land?” “What did I spend on dining last month?” The AI would give a thoughtful, generic answer. The user would still open four apps to do the actual work.
The bottleneck wasn’t intelligence. It was access. AI assistants are excellent at reasoning about money in the abstract. They just had no connection to your real financial life. Context is that connection.
We built it on Model Context Protocol, the open standard Anthropic introduced for letting AI agents talk to external data. That choice was deliberate: any MCP-compatible assistant — Claude, ChatGPT, Gemini, Cursor, whatever ships next month — can read and act on your finances through one connection. No lock-in to our model, our app, or our chatbot.
The fundamental change is that the AI stops being a generic adviser and becomes your adviser. It knows your real balance, your real recurring charges, your real goals. And because Context holds the memory, those facts travel across assistants. Tell Claude you’re saving for a house and ChatGPT knows next week.
We didn’t want to build another AI assistant. We wanted to make the one you already use actually useful for your money.
How does Context ensure permissioned, secure access for AI systems like ChatGPT, Claude, and Gemini while maintaining user privacy and control?
The most important thing: Era doesn’t move user data into the AI providers’ systems. The AI queries Context, Context returns the relevant data for that specific query, and the AI uses it in that conversation. We don’t bulk-upload to OpenAI or Anthropic. Your financial life isn’t sitting in anyone’s training pipeline. The data stays with us.
The user controls what each connected AI can see and do. Read and write access are separate scopes. An AI can be allowed to view your transactions without being allowed to create rules, and to create rules without being allowed to move money. Anything that touches real dollars requires explicit user authorization.
Every action is logged, attributable, and reversible. If an AI agent created a rule, you can see which AI, which conversation, and what changed. And you can undo it.
This isn’t just product hygiene. As a registered RIA, we’re subject to Regulation S-P, the privacy rule for consumer financial information. The control surface our users see is the architecture we’re required to ship anyway.
In an AI-driven financial ecosystem, how do you think about building trust, maintaining transparency, and upholding fiduciary responsibility?
The fiduciary standard exists because the gap between adviser and client is enormous, and the law decided to make the adviser legally responsible for closing it. We built Era around that standard. Era Financial Advisers LLC is the adviser of record for paid users: the same kind of relationship you’d get from a human adviser, if you could afford one.
Transparency means showing the work. When Context reorganizes your transactions or creates a rule on your behalf, you can see exactly what changed, why, and reverse any of it. The AI doesn’t operate in a black box.
Fiduciary responsibility can’t sit with the AI itself. Frontier models are extraordinary, but they’re not entities the law can hold accountable. The structure has to put a regulated entity behind the AI, one that takes the legal responsibility the AI can’t.
In financial services, trust isn’t a marketing claim. It’s a consequence of the structure you choose. Most AI fintech operates in frames where that duty doesn’t apply. We chose to be on the hook on purpose.
Looking ahead, what does “agent-led” financial management look like in practice, and how do you see this category evolving over the next 5–10 years?
The near-term version is already happening. People stop opening their banking app for routine questions and start asking their AI assistant. The dashboard becomes the fallback, not the default.
In the medium term, AI agents handle most financial decisions that don’t require human judgment — bill negotiation, savings optimization, tax-aware rebalancing — within rules the user sets, backed by a regulated entity. Human advisers don’t disappear. They shift to where their judgment was always most valuable: life events, complex planning, hard conversations.
The longer-term shift is that the question stops being “which finance app do I use” and becomes “which AI do I use, and what is it allowed to do on my behalf.” Money management becomes ambient. The interface to your financial life becomes the same interface as the rest of your life.
In ten years, asking your bank’s app a question is going to feel the way calling your bank feels today. There’s a better interface available, and your AI is using it.
What advice would you give to founders, operators, and financial institutions looking to build responsibly at the intersection of AI and finance?
For founders: build the regulatory structure before the product. In regulated industries, the structure is the product. Skip it and you’ve added existential risk; build it and worst case you’ve added cost. The asymmetry only runs one way.
For operators inside larger companies: don’t confuse AI integration with AI strategy. Bolting a chatbot onto an existing app is not a strategy. The real work is rebuilding the data, memory, and rules layers underneath, so an AI can act, not just answer. Most of the AI finance products that will matter in three years haven’t shipped yet, because most companies are still optimizing the old interface instead of building the new one.
For financial institutions: open protocols are coming whether you participate or not. Your customers are already using AI assistants for everything else in their lives. They will use them for money next. Institutions that make their data accessible to those agents, on terms that protect both sides, will stay relevant. The ones that don’t will become invisible to the customer’s primary interface. That’s a hard place to come back from.
Stay Ahead of the Financial Curve with Our Latest Fintech News Updates!
