Eight Years In, 120 Nos Later: What the AI-for-Finance Boom Looks Like to Us

Eight Years In, 120 Nos Later: What the AI-for-Finance Boom Looks Like to Us

We pitched 120 VCs on building AI for FP&A. They said it was technically impossible. Eight years later, it's a $50B category. Here's what the view looks like from where we're sitting.

I bought the domain clockwork.ai in 2018. People marked my emails as spam, they thought .ai was a country code for Antigua and I was selling bonds from a third-world prince. That's how long we've been building AI for FP&A.

Eight years later, every fund has an AI-for-finance thesis, every accounting tool has "AI" in its tagline, and the category 120 VCs told us was technically impossible is now worth more than $50 billion.

Most of the companies you've heard of in this space are eighteen months old. We've been here for eight years. That delta is the only reason this post is worth your time.

So I want to use it: tell you what the AI-for-finance boom actually looks like from inside the company that's been building it longer than the category has existed. What we saw early. What the new wave is missing. And what we're watching now.

To do any of that credibly, I have to start with the part most of the new entrants don't have. Receipts.

So let's start with 120 of them.

What 2018 actually looked like

I pitched 120 VCs over four years on AI-powered financial planning. A hundred and twenty rooms. Different cities, different partners, the same answer in 120 different costumes.

Context matters, because "AI" in 2018 didn't mean what it means today. ChatGPT didn't exist. The transformer paper was a year old. To most investors, AI meant autonomous vehicles or whatever robotics company TechCrunch had just covered.

The funds I was pitching all called themselves "AI investors" anyway. Some of them genuinely understood machine learning. Most of them were vibing.

And the vibe said no.

The three no’s

After enough pitches, rejections stop sounding random and start sounding like a pattern. Mine came in three flavors.

"This is technically impossible." The fun one. They were calling me a liar to my face. The people telling me what couldn't be built were almost never people who had built anything.

"This isn't venture-backable." One of the wealthiest partners I sat across from told me: "Make this a lifestyle company. Don't raise. Take a million a year home. Your market is only worth $500 million." That market is $50 billion today.

"Finance is an art, not a science." The deepest no. They weren't saying I couldn't build it, they were saying the category itself was philosophically wrong. The same people who told me that in 2018 are, in 2026, wiring checks to AI bookkeeping tools every other Thursday.

The part with no montage

Between 2018 and 2022, I bootstrapped Clockwork.

No check from any of those 120 rooms. I paid payroll out of the consulting business I'd built on the side. I hired with my checking account. Our first customer was a tiny company that made recycled paper straws, they let me test whether what I'd been promising for four years actually worked. It did. The model predicted their cash flow correctly. They eventually exited. We closed a seed round with a top-tier Boston VC who I'd raise from again tomorrow.

One honest caveat: the responsible advice for a founder sitting at 120 nos is stop. I advise founders for a living. If one walked in with my numbers, I'd probably tell them to stop. Conviction can become denial. Stubbornness destroys companies. I got lucky. I am the exception, not the rule. Don't mistake my outcome for proof of my method.

The gold rush, and the silence

Then ChatGPT happened.

By mid-2023, every fund had an AI thesis. Every fund had an AI partner. Funds that had passed on real AI startups for years were suddenly leading rounds in companies that were, to be blunt, ChatGPT wrappers built in a weekend.

None of the funds that called Clockwork "technically impossible" came back yet. They were too busy figuring out what an LLM actually was.

When an investor today asks you "what model are you built on?", most of them mean "which LLM are you wrapping?" They missed the thirty years of machine learning that came before ChatGPT. They're not investing in AI — they're investing in LLMs that came out three years ago, and they don't realize those aren't the same thing.

There's a name for that now: AI-washing. Jack Dorsey cut 4,000 people at Block in February for it. The investors who told me finance was "an art, not a science" in 2018 are now writing checks to AI close tools and AI reporting wrappers. The category they swore was conceptually wrong is the category they can't fund fast enough.

Nothing changed about the work. The work that was "technically impossible" in 2018 is the same work. The only thing that changed is that someone else made it look like the LLM era, and the investors got FOMO.

The meeting

In early 2025, one of those firms came back.

I never wanted the meeting. One of my current investors sent me a list of intros and asked which ones I wanted. I said yes, yes, yes, and an explicit no to this one. Don't introduce me. Wires got crossed. He made the intro anyway.

Their email: "Great to connect. It's been a while."

You know exactly how long it's been. You know exactly what you said.

Most of the time, the founder playbook says take the meeting. Smile. Pitch your traction. Keep the door open. It's a small world. Being polite is cheap insurance, and I do it all the time.

But there's a threshold. There's a moment where you've already won the argument and they're still acting like the meeting is theirs to grant. That's the moment the founder dance stops being insurance and starts being bullshit.

I chose violence.

I took the call out of respect for the investor who made the intro. I sat on my couch in Boston. They dialed in six-deep: both managing partners, a senior associate, an associate, and two interns. Just me.

"Why don't you share your screen and walk us through the deck?"

"No."

I told them I appreciated the time. I reminded them, in their own words, what they had told me four years earlier: Clockwork is technically impossible.

They squirmed. "Well, we didn't actually say impossible. We said it would take a long time."

You said impossible. I have the emails. The market you sized at $500M is $50B. Every incumbent is trying to do what we've been doing. You weren't early. You were wrong.

"We're excited to hear what you've built. Why don't you share the deck?"

I'm not going to pitch you. Having you on my cap table makes no sense to me. I respect the intro. I wish you the best of luck.

The associates and interns sat there like holy shit. The two managing partners, the ones with all the bravado, the ones who had felt comfortable telling a first-time founder his life's work was impossible, had no idea what to do. They'd never had this happen. Founders are supposed to dance.

I wouldn't change a single thing about that meeting.

Vindication is a filter, not a victory lap

Most people will read that and assume it was about ego or revenge. It wasn't.

Vindication isn't a victory lap. Vindication is a filter. It tells you who saw the thing when it was hard to see, and who needed the market to do their thinking for them.

The investor who comes back four years later isn't a better investor. They're the same investor with new information they didn't generate themselves. The judgment is unchanged. The only thing different is that the answer is now visible to anyone with a browser tab open. That's not insight. That's reading the news.

The person who couldn't see it when it was hard is not going to see what comes next when it's hard either. You're not learning about your business in those meetings. Your business already told you the answer. You're learning about them.

This isn't a founder-only story. Every reader has a version. The boss who said you weren't ready and now asks for advice. The friend who said the idea was bad and now DMs you to say I always knew. The family member who couldn't see who you were trying to become and now wants to be close again. Same person. Same instincts. They just have a result now they didn't have before.

You don't owe them the meeting. You don't owe them the catch-up. You don't owe them the coffee. You don't owe them the credit.

And to be clear: it isn't anger. If I see that investor at a conference, I'll shake his hand and mean it. He was wrong. That's not a grudge. That's information.

Or, to borrow a line from Drake: where were you when I was shooting in the gym?

What we see from year eight

I told you at the top this post was only worth your time because of the vantage point. So here's what eight years of building AI for FP&A, through one gold rush, multiple "AI is overhyped" cycles, and a 100x category expansion lets us see that most of the new entrants can't.

The boom is real. Most of the companies riding it won’t survive.

The wrappers won't survive the next model release. The AI-bookkeeping tools that are really just rebranded automation won't survive two quarters of a real CFO using them. The "AI finance" tools built by people who've never sat in an FP&A seat won't survive a real close, a real board pack, a real cash crunch. The companies that AI-washed themselves into a check in 2024 will quietly drop the AI claim by 2027 or get acquired for parts.

What survives is the work nobody could see in 2018. The actual machine learning on financial time series. The proprietary data. The vertical depth in finance. The eight-year head start on understanding why a real CFO does the work the way they do, and why a generic LLM is the wrong tool for almost every job that matters inside the close.

We've been here. The category just arrived. We're glad to have the company.

What we're watching now: what survives the wrappers, what the next eight years actually looks like, what the FP&A function becomes when the dust settles, is the next post.