In AI, we don’t trust (yet)

When it comes to money, being almost right is often the same as being wrong.

Increasingly, people are turning to AI for advice. Yet almost nobody trusts it completely, especially when it comes to their finances.

This scepticism is justified. We have all learned that AI can sound confident even when it’s wrong.

Most of the time, a hallucination is merely inconvenient. An incorrect restaurant recommendation is forgettable. A buggy piece of code can be fixed.

You need to understand not only what the answer is, but why it is the answer. Months later, when markets fall and emotions take over, you need to be able to ask whether your original thesis still holds. That is much harder than it sounds.

Five years ago, when Sahil and I started Bharosa, we encountered a simple problem. Many important financial questions seemed to have no answers.

• How has my advisor actually performed for me?

• What sectors am I investing heavily in without realising it?

• How can I withdraw money from my portfolio tax efficiently?

Everyone knew the answers were directionally correct. Very few knew if they were actually correct.

In most industries, that’s acceptable. In finance, we felt it shouldn’t be.

The problem wasn’t a lack of intelligence on the part of people or systems attempting to answer these questions. It was a lack of rigour.

Producing the correct answer often requires hours of manual work, spreadsheets, assumptions, and human interpretation. Most people settled for rough estimates. Getting the exact answer was simply too expensive and time-prohibitive.

The deeper we looked, the more we realised something important. Many financial questions that people thought were simple were among the most complex reasoning problems.

Sahil understood how to answer these questions because he had spent years wrestling with them professionally.

My background in technology told me that if humans could answer these questions with enough time and effort, computers should be able to answer them too.

The problem wasn’t that the analysis was impossible.

The problem was that nobody had built the machinery to do it reliably.

If the underlying data exists, anyone should be able to understand what happened. They should know why it happened and the consequences of different decisions.

So we spent the next five years focused on a single idea.

Make financial analysis trustworthy.

We built systems that could explain exactly what happened, trace every number back to its source, and withstand scrutiny.

At the time, this may have seemed like an unusual obsession. Today, it feels more relevant than ever.

AI can generate endless answers to every question under the sun. But the decisions that shape our lives, careers and finances require something more.

They require trust. That is the Bharosa we are building.