Turning two-day processes into 20 minutes at Intent
Leo Kraus is the founder of Intent, a consultancy that removes repetitive admin work from small B2B firms by building targeted systems and dashboards. He starts with evidence of where work piles up, builds the smallest thing that removes it, and shows clients the hours actually recovered.
Use AI to surface signals and automate qualification, but keep humans writing the first outreach and owning judgment. Record your own outcomes from day one – those honest, local records matter far more than following every model announcement.
What’s the quick origin story of your business, and what makes what you do genuinely different from the other options your customers are looking at?
Intent started because I kept seeing the same thing in small B2B companies. The owner started the business for a reason. They were good at something, they wanted to build it their way, they wanted to serve their customers properly. And then five years later they spend most of their week on admin that has nothing to do with any of that. They know something is off, but they can’t tell you which part of the day is actually eating them.
The people who work for them have the same problem. Someone who should be talking to customers is instead copying numbers from one system into another, chasing paperwork, rebuilding the same report every month. That’s not why anyone took the job.
So what I do is give that time back. Not in theory, in hours you can point at.
What makes it different is that I don’t start with the tool. I start with the evidence. Before I propose anything I want to see where the work actually piles up. Which documents get typed twice. Which calls nobody answers. Which report someone rebuilds by hand every single month. Then I build the smallest thing that removes it, and I show the numbers behind it. No black box. If I can’t show you what changed, I haven’t done the job.
Where does most of your new business come from today, and where do you wish it came from?
Today it’s mostly direct outreach and referrals from people who’ve already seen me build something. That’s honest work, but it’s me pushing. I go find the firm, I build the case, I open the conversation.
Where I want it to come from is the work itself. The systems I build leave a visible trail — a dashboard an owner shows their accountant, a process that used to take two days and now takes twenty minutes. That’s the only marketing in this space that survives contact with a skeptical buyer, because the buyer has already been pitched AI by five people this quarter and is tired of it.
So the goal for the next year isn’t more volume at the top of the funnel. It’s fewer, better-documented builds that generate their own inbound, and outreach that’s sharp enough to be worth reading even when the answer is no.
What’s one thing you’ve actually handed over to AI in the last year, and one thing you tried it on and went back to doing the human way?
Handed over: qualification. I built a signal engine that reads public data on a company and works out whether they’re actually ready for this. Team composition, hiring patterns, what kind of work their site really describes, how long the owner has been in the seat. It scores them, and it throws out the ones that are a bad fit before I waste a conversation on them. My gut used to do that job. The engine does it better.
Went back to human: the first message. I tried letting a model write my outreach. What came back was fluent and completely forgettable. Technically fine, zero specificity, exactly the kind of thing everyone deletes without reading. So now the system does the research and puts the evidence in front of me, and I write the sentence myself. The model finds the facts. I make the claim. That split has held up better than anything else I’ve tried.
Have you noticed a change in how customers find you or what they already know before they reach out? Is anyone arriving through AI assistants like ChatGPT yet?
Not much direct traffic from AI assistants yet. In my segment, small B2B companies, the buyer still arrives through a person. Someone they trust mentions my name. That hasn’t changed.
What has changed is what they already believe when they show up. Two years ago I had to explain what was even possible. Now they’ve asked ChatGPT about it themselves and they arrive with a half formed picture. Either they’re wildly optimistic and want to know if it can just do all of it, or they’re burned, because somebody already sold them something that didn’t stick.
The question I get most often now is some version of "why can’t I just build this myself?" Which is fair, and I never argue with it. Usually the honest answer is that they can build the demo. What they can’t build is the system that keeps running on a Tuesday when the data comes in malformed and nobody’s watching.
So there’s less educating now and more correcting. I’ll take that trade.
If you were advising someone in your industry on all of this for the next 12 months, what would you tell them to actually do, and what would you tell them to ignore?
Do: write down your own outcomes from day one. What you sent, what happened, and why it ended. Not just won or lost, the actual reason. Most people in this business run on anecdote and a good memory, and a good memory is a liar. Six months of honest records beats any dataset you could buy, and it’s the one thing a competitor can’t copy off you.
Ignore: the model announcements. They matter far less to your business than you think, and keeping up with them is a full time job that pays nothing. Ignore anyone selling you a shortcut to owning data you haven’t earned. And ignore the pressure to have an opinion on every new tool that shows up. Being early to something bad is not an advantage.
Thank you to Leo Kraus and the team at Intent for sharing what actually worked with Leaders Perception readers.
Want to share your playbook?
Leaders Perception publishes founder and operator interviews every week. A few questions, about five minutes, no calls. It is free and everyone we accept gets published.