What AI is actually for.

I’ve spent two years putting AI systems inside real businesses, alongside running Gain Partners with Marcus. What it’s taught me isn’t the stuff that makes for great LinkedIn content. But it’s a lot more useful.

Where founders’ time actually goes

The first thing it taught me: founders almost always spend their time on work that isn’t their highest-value work.

It’s not that they’re bad at managing themselves. It’s that businesses generate friction constantly. Reporting, follow-ups, problems, summarising meetings, chasing invoices, keeping the CRM up to date. That friction lands on whoever cares most, which is usually the founder. So they work through it, call it running the business, and meanwhile the work they’re actually best at, the sales calls, the client relationships, steering the product, gets squeezed into whatever time is left.

What AI is genuinely good at

The work AI handles best is high-volume, low-judgment work. Data pulls and reporting. Content first drafts. Internal operations: process automation, financial ops, scheduling, communications, compliance. Client follow-ups with known information and predictable structure.

These are real time sinks. They’re not where your edge lives.

Take the AI sourcing pipeline we worked on for a recruitment business. It pulls candidates from across the usual sources, scores them against the profile the founder is hiring for, and surfaces a shortlist. He still picks up the phone. He just isn’t the one doing the scrolling.

It’s good at the infrastructure of a relationship

The other area where I’ve seen it make a genuine difference in founder-led businesses is customer experience. Not by replacing the relationship, but by making the consistent parts of it actually consistent.

Most founders I work with care deeply about how clients are treated. But they can’t be in five places at once. So the onboarding email goes out late, the weekly update gets skipped because Thursday was a nightmare, or the client who hasn’t heard from you in two weeks starts wondering if they made the right call.

AI handles the infrastructure of a relationship without being the relationship.

Onboarding sequences that fire on time. Status updates that go out on schedule. Follow-ups after deliverables that tell the client they’re being thought about, even when the founder is in back-to-back sales calls.

The key phrase is “done well.” Automated comms that feel robotic, generic, or obviously AI-generated do more damage than silence. If you’re going to use AI in the client-facing layer, it needs to sound like you. That takes more setup than people expect.

Take the client portal we built for a performance coaching practice. The founder still coaches, still writes the personal notes, still picks up the phone when it matters. The portal handles the bits in between: payment, scheduling, the bookings the client used to email about. The relationship stays his. The infrastructure stops being his.

Where AI isn’t good

Where AI doesn’t work, and I’ve watched founders learn this the hard way, is anywhere the relationship is the product.

AI can’t build trust. It can’t make the judgment call that costs you a deal if you get it wrong. It can’t do the creative leap that comes from knowing your client’s situation in a way you can’t put into a prompt. These are the things that make founder-led businesses work. They shouldn’t be candidates for automation.

Marcus handles the strategy and advisory side of Gain Partners. He’s in the relationship layer all day. Whenever we talk through where AI fits in, his instinct is always the same: don’t automate what makes you good. He’s right. The work that feels irreplaceable usually is. The work that feels like overhead almost always isn’t.

AI doesn’t arrive fully formed

There’s one thing I’ve watched trip up almost every founder who approaches this seriously, and it’s worth naming.

AI doesn’t arrive fully formed. The businesses I’ve seen get the most out of it didn’t start with an impressive system. They started with a foundation and built on it slowly. The ones who tried to skip that step mostly ended up with expensive tools producing unreliable outputs and scepticism about whether any of this actually works.

The foundation is the unglamorous part. Your data: clean, connected, consistently collected. Your processes: defined well enough that a system can follow them. It might mean changing how your team logs information, or finally merging the four spreadsheets that should have been one database two years ago. Without it, you’re not building AI solutions. You’re building brittle automations that work sometimes, fail quietly other times, and end up costing more time than they save.

Get the data right first. Then build one thing on top of it. Watch it run. When it’s stable and you trust it, build the next thing. Feature by feature, over months, the system compounds.

That’s the version that eventually runs your business like clockwork. Not the version you buy in week one.

Pick one task

Pick one recurring task that takes you two or more hours a week. Not a category of work, a specific task. The report you run every Friday. The follow-up emails after discovery calls. The meeting notes you transcribe and format before sending to clients.

Pick that one thing, build or configure something to handle it, then watch it for two weeks. Does the output hold up? Does anything break? Does it save you the time you thought it would?

If yes, pick the next one.

There’s just a Friday afternoon where you don’t have to build the report, and you use that time to call a client you’ve been meaning to call for three weeks. The ROI isn’t in the automation. It’s in what you do with the hours you get back.

Most founders aren’t failing because they lack AI tools. They’re failing to protect the time they need to do the work only they can do.

The systems I’ve built that actually stick are boring. They run in the background. They quietly do the job, week after week, and the founder gets a little more of their time back. Which is the whole point.

Read next: Working with athlete-founders.