Founder Lesson of the Week #2

“See, the thing about building AI for financial data is that your end user is closer to a gambler than a typical ‘report viewer,’ or ‘dashboard’” joked my friend who works in UX at Bloomberg. They realized—after countless customer interviews—that 99% accuracy just won't cut it when million-dollar positions are on the line.

I’ve been thinking a lot about how the user experience for finance applications is broken. It’s not so much a technical issue with AI as it is a problem of pitching truth and accuracy in a world where trust is crucial for the user.

For investors, users aren’t just seeking productivity; they’re often confirming a hypothesis or placing a bet. By definition, investing is always two-sided—each party has its own perspective gleaned from countless insights—so while AI can guide that process, we must remind users it’s THEIR hypothesis, not an absolute truth from the machine.

I began thinking about all this when I soft-launched Avanzai Charts this winter, a sweetener to the greater enterprise version of Avanzai. It transforms free public data for users, but I noticed nearly 70% of queries revolved around forecasting the price of Bitcoin or other assets.

I spent my Christmas venting to friends, my partner—basically anyone who’d listen—because I initially blamed the users for misunderstanding. Then I remembered what a Bloomberg colleague once said: if people aren’t happy with your product, nine times out of ten you need to shift their expectations of success, not point fingers.

That insight led me to tweak Avanzai Charts’ user experience to emphasize what we CAN realistically offer and how the product should be used. Instead of peddling a crystal ball, we position it as a tool to refine their own theses and highlight insights they might otherwise miss.

In other industry verticals—like sales, marketing, and legal—AI’s value is far clearer: it makes you more productive and saves time on a tangible deliverable. You see it in more demos, more leads, or a concise summary of a thousand-page legal brief—binary outcomes that show direct value.

In the financial markets, however, beyond simple market snapshots, the average investor still holds the most leverage when using AI. After all, many intermediaries exist because of outdated infrastructure or regulatory mandates—think how electronic trading hasn’t fully caught on in certain fixed-income markets, and how post-2008 fintech solutions tried reassuring users that ‘it won’t happen again’…until COVID did.

Some of the most successful investors I know just open a spreadsheet and press F9 every morning—no Python needed. We’ve arguably created more problems with our shiny solutions, because the ends must justify the means…right?

Tools like Perplexity, ChatGPT, and Claude do a great job of helping users research and uncover facts quickly. But they don’t necessarily guide investors through building, questioning, or uncovering hidden insights for their hypotheses.

This binary mindset means users often expect AI to deliver a single, crystal-clear winner, turning it into a quest for perfect predictions that may not exist. It’s no surprise so many people ask about forecasting—everyone wants that magic bullet for tomorrow’s market moves.

At the end of the day, finance is still gambling, so even the slickest analysis can’t guarantee an outcome, often leaving users disappointed. On the institutional side, finance runs on information asymmetry; AI can supercharge insights, but it also raises the bar for proof.

With so many free tools like TradingView or Yahoo Finance, users now expect top-tier analytics without spending a dime. So any monetized AI must offer more than just rehashing data—it should help combine and cross-reference sources, revealing insights people wouldn’t spot otherwise.

This is exactly why a typical Q&A chatbot doesn’t fully cut it in finance. There’s no single truth, so users need flexible ways to shape the data around their own narratives.

The more you promise a perfect answer, the more you risk letting down those who demand a definitive outcome. It’s smarter to equip them with tools to build and refine their own theses than to hand them a one-size-fits-all verdict.

In other industries, AI clearly boosts productivity by speeding up tasks or automating processes. But in finance, you’re really offering an edge in speculation, which is an inherently riskier proposition.

The real magic is in tools that transform data so each user finds their personal truth. That means focusing on customization, correlation, and deeper insight instead of a generic Q&A approach.

Ultimately, we need to promise less and deliver more. There is no crystal ball, but we can empower people to discover their own version of the truth in finance, and hopefully get that 99% to 100%.

That is all for this week folks! If you enjoyed please subscribe! 🤝