You want AI in your company. Start with one of four jobs.
by Luis Gomes, Founder & Growth Lead
Every week we get the same inbound. The owner of a small or mid-sized company writes: "I want to train my team on AI." Or: "We need to do something with AI." Or: "What's a good AI course I can put my team through?"
Behind every one of those messages is the same problem. The owner can feel that AI matters. They cannot articulate where it fits in their company. Generic training does not fix that. You can put a team through a two-day workshop on prompts and they will go back to work on Monday and nothing will change. Training without a specific job to do is theater.
The version of this question that produces results sounds different. It sounds like: Pick one job in my company. Install AI in that job. Show me what changes in 30 days.
That is a foothold. It is not a roadmap. It is not a transformation plan. It is one job, done, in 30 days. Once it works, the next job is easier. Once you have three working installs, you have an AI-native function — not as a slogan, but as a thing the team uses every day.
Below is the menu we walk owners through when they ask. Four jobs. Each one pays back inside a month if it pays back at all. Each one is concrete enough that a team can feel the difference in the work, not in a slide.
The four jobs
1. Time back. The repetitive output that takes the senior person on the team hours every week — automated. Monday reports. Email follow-ups. Meeting notes that someone has to clean up. Status updates the founder writes for the team. Spreadsheet rolls. The work that is not the operator's job but somehow always lands on their plate.
The math is straightforward. A marketing operator who spends six hours a week assembling the Monday report goes to one hour. The five hours go back to thinking about next week's campaign, not formatting last week's. Across a small team, that is two days a month per person, returned. You can measure it. You can defend it.
2. Faster response. The first response a customer or a lead or a teammate gets — accelerated from days to minutes. First-line support that triages and answers the easy 60% of tickets without waiting for a human. Inbound lead qualification that asks the right four questions and routes only the live ones to sales. Internal "ask the team" questions that get an answer from the company's own documents instead of a thread that takes three days to resolve.
The point is not that AI replaces the human. The human still owns the hard cases, the relationship calls, the judgment. The point is that the queue does not pile up while the human is in another meeting. Customers do not wait. Leads do not cool. Teammates stop scheduling a 30-minute call to ask a 30-second question.
3. Institutional memory. What lives in people's heads becomes a base the rest of the team can query. The brand voice rules the head of marketing has been holding for five years. The pricing logic the founder explains to every new salesperson. The reason the product works the way it does. The customer-segment patterns the senior account manager has internalized over hundreds of accounts.
Once those are in a queryable system, three things happen. New hires ramp in weeks instead of months. When someone leaves, the knowledge stays. The team stops asking the same person the same question — the system answers. This is the closest thing to a moat a small services company can build, because it compounds with use.
4. Quality floor. AI as a second pair of eyes on every piece of work the team produces. Junior writers get drafts that already match the brand voice because the system was trained on what the senior writer approved. Junior analysts get reports that flag the obvious mistakes before they reach the operator. Account managers catch errors before they go to the client because the system is reading their drafts in real time.
The team does not get smaller. The bottom of the team's output rises to meet the top. Less rework. Less single-person dependency. Less of that specific feeling — the one every operator knows — where the senior person is the bottleneck because they are the only one who catches what needs catching.
How to pick one
One question: which one hurts most today?
Don't pick the most exciting. Pick the one where the most senior person on your team is doing the most repetitive work. That is the place where the math is biggest, and the place where the team will feel the change first. The senior operator with five hours a week back is the senior operator who can finally do the work you actually hired them to do.
The owners who get this wrong almost always pick by what sounds the most like AI. They pick the chatbot, or the auto-generator, or the dashboard. The owners who get it right pick the place that has been quietly draining their best people for two years.
What 30 days looks like
Week 1 is scoping. What is the job, who does it today, what does the input look like, what does the output need to look like, what tools and data does AI need access to. Most projects fail in week 1 because nobody scoped the job — they scoped the technology.
Weeks 2 and 3 are install and test. The system gets built around the actual workflow. The team uses it. It breaks. We fix the edge cases. The team uses it again. By the end of week 3 the team should be using it more than they fight it.
Week 4 is hand-off and measurement. The team owns it. We measure the thing we said we would measure in week 1 — hours back, response time, ramp time, error rate. The number is the number. If it didn't pay back, we say so. If it did, we pick the next job.
There is no two-day workshop. The team learns by using the thing. That is how AI training actually works, because the alternative — sitting in a room learning prompt patterns in the abstract — does not survive contact with a Monday morning.
Foothold over big bang
Most owners who ask about AI are pitched the big-bang version: re-org the team, train everyone, pick a vendor for the whole stack, transform the company. That pitch is what scares owners and stalls projects. It is also where most "AI transformations" actually fail — too much surface area, too little signal that the thing works.
The foothold is the opposite. One job. One install. One measurement. The team feels the change. The company gets a specific, defensible, measurable result. Then you pick the next job, and the team has earned the credibility to make a bigger move.
The architectural insight underneath this is not new. Diana Hu of Y Combinator has named three pillars every AI-native company gets right: the model is at the center of the workflow, data accumulates with use, and the work runs as agentic loops rather than pipelines. Each of the four jobs above is a small instance of those pillars. Block calls the macro version "From Hierarchy to Intelligence" — the architectural shift from coordinating people to coordinating intelligence. Harvard Business School Online has the academic framing of what "AI-native" means in practice, separating it from companies that have bolted AI on top of an older operating model. The four jobs are the smallest concrete way to start moving toward that architecture without committing to it on day one.
What we install
This is the work Leanboat does. We pick a job with the owner. We install AI in that job in 30 days. We hand it off. We measure what changed. The team finds out what AI actually does for them by using it on real work, not by sitting in a workshop.
The first install is the cheapest, fastest way for a company to find out what AI changes for them. Once it works, the second one is faster. The third one starts looking less like a project and more like the way the company operates.
That is what AI-native looks like from the inside. Not a transformation. A series of small installations that accumulate into a different kind of company.