Why I spent a year building an AI product nobody asked me to build.

A year ago, I started building something because I couldn’t stop thinking about a chart.

The chart came from Anthropic’s research on how AI is actually used at work. It showed, across occupation after occupation, an enormous gap between what AI was capable of doing and what people were actually using it for. The capability was there. The usage wasn’t. And that gap — that quiet, expensive, unaddressed gap — wouldn’t leave me alone.

Nobody asked me to solve it. I didn’t have a customer waiting. I had a conviction, some hard-won experience, and a year ahead of me. This is a short, honest note on what that year taught me.

The idea I started with was wrong

The first version of this product was called AICareerArmor. The name tells you everything about how I was thinking: defensively. It was built around fear — protect your career before AI comes for it. Help people armour up against displacement.

It took me months to admit that the framing was wrong. Fear was the worst possible foundation. It made employees defensive, made the product feel like a threat, and — most importantly — solved the wrong half of the problem. The real opportunity wasn’t helping people hide from AI. It was helping people become genuinely more valuable with it. And the buyer who cared most wasn’t the anxious individual — it was the company that had spent a fortune on AI tools its people weren’t using.

So I killed it. The name, the framing, the go-to-market — all of it. That became Forsight: an AI workforce readiness platform, built around growth instead of fear, sold to the companies that actually own the problem.

What killing your own ideas actually feels like

Everyone says founders should “kill their darlings.” Nobody tells you how much it stings when the darling is a year of your own work. Over this year I killed a lot:

What I killed	What replaced it
The name — AICareerArmor	Forsight
The framing — fear & defence	Growth & amplification
The buyer — the anxious individual	The company that owns the problem
The business model — sell to individuals	B2B, white-label through platforms

Each time, it felt like admitting failure. Each time, the thing that replaced it was stronger. I’ve come to believe that the willingness to kill your own good-enough idea for a better one is not a sign that you’re lost — it’s the single most important muscle a founder builds. The founders who fall in love with their first idea are the ones who go down with it.

What a year of unglamorous work produces

There were no launches to announce for most of this year. No funding to celebrate. No team standups. Just a lot of quiet, unglamorous work — building the seven-agent pipeline, rewriting the positioning for the tenth time, staring at a problem until it got simpler.

Here’s what I’ve learned about that kind of work: it compounds invisibly and then shows up all at once. For months it looks like nothing is happening. Then one day you have a coherent product, a sharp thesis, a channel strategy that an experienced investor validates in the first conversation, and a story that holds up under hard questions — and you realise all of that was being built the whole time, in the parts nobody could see.

The year didn’t produce a series of milestones. It produced a foundation. And a foundation is worth more than a milestone, because everything else stands on it.

What I actually believe now

I believe the story of AI at work has been told wrong. The dominant narrative is replacement — AI comes, people go. But the companies making that trade are mostly discovering they’ve lost institutional knowledge, client relationships, and judgment that the AI can’t replicate, in exchange for tools their remaining people still can’t fully use.

The better story — the one I’ve spent a year building for — is amplification. Take the experienced people who already understand your business, and make them genuinely powerful with AI. That combination beats either humans alone or AI alone, and it’s far cheaper than replacing the people who know how the business actually works.

That’s what Forsight is for. Not to help companies replace their people with AI. To help them make the people they already have far more valuable with it.

Pradeep Shekaran is the founder of Forsight (EpicureAI Labs), an AI workforce readiness platform. If any of this resonates — as a potential partner, customer, or fellow builder — I’d love to talk.