The Businesses AI Forgot

14 July 2026

It's been a couple months since we started talking publicly about what we're building with HappyHQ, and I've been wanting to write this for a while. Not a launch post. Not a milestone update. Just an honest account of what we've learned, where we were wrong, and how much further we still have to go.

When we started, I thought I understood this problem. I'd spent years in product, I'd watched small businesses struggle with tools that were never built for them, and I thought I could see the shape of the solution pretty clearly. I was wrong about a lot of it. I want to write about that, because I think the mistakes are actually more interesting than the parts we got right.

Where we started

I've spent the last year building HappyHQ, and if there's one thing I keep coming back to, it's this: the AI wave everyone is so excited about is being built for a completely different kind of company than the ones I care most about.

Talk to anyone in tech right now and you'll hear about AI transforming knowledge work. Coding assistants. Research copilots. Marketing automation. Most of it aimed at companies with dedicated ops teams, IT departments, and/or someone whose whole job is "figure out the new tools." That's not who I'm building for. The owners I care about and am talking with have no idea how or where to start to figure out the new tools!

I'm building for the business owner who's run the same business for fifteen years with twelve employees and a filing system that lives half in email, half in someone's head, and half in a drawer somewhere. I'm building for the owner-operator who knows every one of her three hundred clients by name and renewal date, because she has to, because there's no one else who does. I'm building for the operators who are the actual backbone of the economy and who technology keeps walking right past.

Where we were wrong

Our first instinct, if I'm being honest, was to build the automation first. We thought the value was obvious: give these businesses the workflows, the alerts, the AI-generated efficiency that bigger companies were already getting. We built early versions around that assumption. Automate the renewal reminders. Automate the follow-ups. Automate the busywork.

It didn't land, and it took me longer than I'd like to admit to understand why. We were showing up with automation for businesses that had nothing structured to automate. There was no clean data underneath any of it. There was no CRM, no consistent record, nothing a workflow could actually hook into. We were offering people a faster car with no road under it.

I sat with agency owners who were polite about it, generous with their time, and quietly unimpressed. One of them told me, in a much kinder way than this, that we were solving a problem she didn't have. She didn't need things automated. She needed someone to finally understand how her business actually ran, because right now that understanding lived only in her head, and she was terrified of what happened to the business if she ever stepped away.

That conversation changed the whole company.

What we've learned

The thing that actually changed how I think about this product was realizing you can't automate what you haven't captured. So we flipped the order. Capture first. Automate second.

HappyHQ now ingests and learns about the contracts, the emails, the calls, the scattered notes, all the ordinary exhaust of running a business, and turns it into actual institutional memory. Not a dashboard nobody updates. Not a CRM field that goes stale in a month. A living record of what's actually happening in the business, built from what people are already doing, not from a new process they have to adopt. This isn't a "one and done" situation. This isn't a "connect this" or "integrate this" problem. This work requires truly understanding the business, capturing it, and then automating it. And that's what is exciting to me.

Only once that memory exists does automation mean anything. Renewal alerts that fire because the system actually knows the contract terms. Workflows that trigger because the history is real, not because someone manually configured a rule six months ago and forgot about it.

I didn't understand how much this mattered until I watched owners try to describe how their business really runs. It's never in the org chart. It's in the specific way a certain client likes to be handled, the fact that a particular vendor always needs a reminder, the quiet judgment calls an owner makes fifty times a day without writing any of them down. That's the real operating system of these businesses, and it currently lives nowhere except in people's heads. We were building for the org chart. We should have been building for the judgment calls.

Why these businesses matter

Owner-operated service businesses with five to fifty employees are not a niche. They're everywhere. They run the talent agencies, insurance brokerages, professional services firms, and specialty shops that hold entire local economies together. They don't scale like software companies. They don't have Series B budgets for tooling. What they have is deep, hard-won institutional knowledge, most of it trapped in the founder's head, in a decade of email threads, in the muscle memory of whoever picked up the phone the most times.

That knowledge is the business. And right now, almost none of it is written down anywhere a system could actually use.

Why AI is leaving them behind

Most AI products assume you already have clean data. A tidy CRM. Structured records. Someone whose job is "make sure the systems are good." Small owner-operated businesses don't have that, and they never will, not because they're behind, but because it was never worth their time to build it. They were too busy running the business.

So the AI wave shows up promising automation, and it's mostly useless to them. You can't automate a workflow that was never written down. You can't build on a foundation that doesn't exist. Every automate-first tool asks these businesses to do the one thing they have no time or infrastructure to do: organize themselves first. We made that same mistake ourselves before we understood it.

That gap is where these businesses get left behind. Not because they don't need help. Because the help being offered assumes a starting point they don't have.

How much further we have to go

I don't want to make this sound tidier than it is. We're still early. We're still learning things every week that make me rethink parts of what we've built. Some of the businesses we talk to still don't see themselves in what we're offering yet, and that's on us to fix, not on them to understand better. Building for people who've never had software built for them means we don't get to assume anything. Every workflow, every word on the page, every piece of the product has to be tested against the real, messy way these businesses actually operate, not the tidy version we'd like to imagine.

What keeps me going is that when it does land, the reaction isn't polite interest. It's relief. It's an owner realizing that the thing she's been carrying alone in her head finally has somewhere else to live. That's worth getting wrong a few times on the way there.

Why I'm doing this

I've built products for companies of every size. What keeps pulling me back to this specific problem is that owner-operators are working incredibly hard, often harder than anyone else in business, and the tools available to them haven't caught up to what they actually need. Not more software to manage. Something that finally understands how their business actually works, and takes some of that weight off.

That's the system I'm trying to build. Not AI for the businesses that already have everything figured out. AI for the ones who've been figuring it out themselves the whole time, and deserve something that finally meets them where they are. We're not there yet. Our new idea is software plus managed services, because that's what it's going to take. It's not a simple integration or web hook. It's sitting down with real owners, real teams, and helping them to make sense of the mess. But I wanted to write down where we actually are, mistakes included, instead of waiting until it looks more finished than it is.