We have now watched a genuinely large number of funds use AI for the first time, and the remarkable thing is how identical the journey is. Different fund sizes, different strategies, same seven stages, almost in the same order, often with the same quotes. So we wrote them down. This is part comedy, and a part diagnostic.
You will recognise your fund and you may even recognise yourself.
Stage 1: Denial
"AI is hype. Our business is relationships and judgement. A model cannot do what we do."
Every fund begins here. AI is dismissed in roughly the same way.
This is technically true and completely beside the point. Nobody is automating the judgement. They are automating the four hours of document-wrangling that happens before the judgement. But at Stage 1, the fund hears "AI" and pictures a robot partner, rejects the robot partner, and feels wise. Meanwhile the fund down the road has quietly stopped reading CIMs by hand.
That feeling lasts until a competitor closes a deal faster, or an LP asks about your "AI approach" and the room goes quiet.
Stage 2: The secret intern phase
Denial breaks not from the top but from the bottom. An analyst, unsupervised and slightly bored, discovers that a chatbot summarises a CIM in thirty seconds. They tell no one. Productivity quietly doubles for one person.
The fund is now "using AI" without a single partner knowing it, on personal accounts, with confidential deals. Nobody has decided anything. It is happening anyway.
"Oh, you can just... paste the whole thing in?"
Stage 3: "We'll just build it ourselves"
A partner finally notices, gets excited, and arrives at the conclusion every smart, well-resourced team arrives at: how hard can it be? We will build our own.
"How difficult can this really be?"
Someone has watched enough product demos to conclude they're mostly wrappers around foundation models.
There are diagrams. There is a Notion page.
Someone creates a Slack channel called #ai-project.
A technical associate volunteers to connect a few APIs over the weekend.
For about three weeks, morale is extremely high.
The prototype summarises three CIMs beautifully.
Everyone congratulates themselves on saving six figures in software costs.
Nobody has uploaded a scanned PDF yet.
This stage is intoxicating and almost always a trap, for reasons that become clear in Stage 4.
Stage 4: The building-is-hard hangover
This is where software starts behaving like software.
The prototype works in the demo and falls apart on the third real deal. It hallucinates a number. It cannot read the scanned PDF. The model provider ships a new version and everything needs rewiring. Someone realises that "build it ourselves" was not a weekend project but a permanent engineering commitment to security, accuracy, evaluations, and maintenance, forever, by a team whose actual job is investing.
The prototype slowly is abandoned. Everyone goes quiet about it. We have seen this graveyard many times. It is a peaceful place.
Stage 5: The existential dread
This is the least funny stage. Not because AI suddenly became better.
Because everyone else quietly did. The fund looks up and sees the market moving.
A competing fund completes diligence in half the usual time.
An LP asks what the firm's AI strategy looks like. A portfolio company casually mentions they've automated work your team still does manually.
Individually, none of these moments matter very much.
Together, they create an uncomfortable question.
"Have we already fallen behind?"
Ironically, this is usually where the conversation improves.
The question changes from Should we use AI?
To Which parts of our investment process actually benefit from AI?
That's a much better discussion.
Stage 6: The cautious evaluation
Maturity arrives. The fund stops trying to build it and starts trying to choose it. Demos get booked. By this point, nobody cares whether a chatbot can summarise a CIM.
That's table stakes. Instead, the questions become much more useful.
What happens when two documents disagree?
Can it explain where every conclusion came from?
How does it handle confidential information?
Will every analyst evaluate opportunities the same way?
Can we trust it on a real deal rather than a polished demo?
This is also the stage where funds stop shopping for AI and start shopping for workflows.
Those are very different purchases.
This is the smart stage. It is also where good decisions get made, because the fund now knows what it actually needs: not magic, but a secure, consistent, source-grounded workflow that fits how they decide.
Stage 7: The workflow that fits
Eventually, AI disappears. Not because it failed. Because it became infrastructure.
Nobody discusses email strategy. Nobody schedules meetings about Excel.
They simply become part of how work gets done.
The same thing happens here.
Investment memos become more consistent. Knowledge stays with the firm.
Every deal follows the same workflow.
The analyst still exercises judgement.
They just spend more time judging and less time looking for page 18 of the data room.
Oddly enough, this is where most funds wanted to end up all along.
They just took a scenic route getting there.
This is the destination. The funds that reach it are not the ones who talked about AI at the earliest. They are the ones who got through Stages 3 to 5 fastest and with the least money set on fire.
The cheat code
The funny thing about these stages is that almost every fund believes its journey is unique.
After enough conversations, we're no longer convinced.
The firms moving fastest aren't necessarily the ones that discovered AI first. They're the ones that recognised they weren't trying to build an AI company. They were trying to build a better investment process.
That's the philosophy behind askRIA.
Instead of scattered prompts, personal accounts and undocumented workflows, every deal runs through a secure workspace where opportunities are evaluated against your investment thesis, evidence is cited and knowledge accumulates across the team instead of disappearing into someone's browser history.
If you're somewhere between Stages 3 and 5, you're in good company.
Just don't unpack.
The view isn't worth staying for.
Keep reading
*Skip to Stage 7. Run your first deal free in askRIA and see the workflow that fits.*