For most of my career, an employee with a good product idea had two options. They could pitch it internally and watch it die in a meeting, or they could quit and build it themselves. Either way the company lost: the idea, the person, or both.
Intrapreneurship was the management-book answer to that, and it mostly lived inside large companies with innovation labs and budgets to burn. For a small business it never added up. Building software meant hiring developers, and a first version cost six figures before anyone knew whether a customer would pay for it.
AI coding tools broke that equation, and I’ve watched it happen up close.
An agency owner I know let two of their marketers spend half their working hours on an idea of theirs: software that does the ad-account management they’d been doing by hand for clients for about a decade. They built the first version in under two weeks, without a developer. Within months it had paying customers, and before the end of the year it was a profitable company with them as co-owners.
The way the owner set it up is worth copying.
Domain Knowledge Is Now the Scarce Input
When building was expensive, the developer was the bottleneck and the person who understood the problem was a stakeholder in somebody else’s project. AI has flipped that. A first version of most business software is now a few weeks of work for someone who knows exactly what it should do.
The people who know exactly what it should do are usually already on your payroll. Someone who has done a job by hand for ten years knows the edge cases and the mistakes clients keep making. They also know how to sell it, because they’ve been explaining that work to customers for years.
So the owner’s risk shrinks to a few months of part-time salary, lent to someone whose expertise you already trust.
Start With Work You Already Sell
The strongest candidate is a service your business already charges for. If clients pay you every month to do something by hand, software that does it is an easy sell: the budget line already exists, and you’re offering a cheaper version of something they buy today.
It also means you have a customer list to test it on, case studies to point to, and staff who can run demos without a script.
Ideas that sit outside what you do are harder. Without an existing budget line, you’re creating demand from nothing, and that’s a different business.
Give Them Time and One Number
The agency’s test was simple. Two people got half their hours for two months. The only thing that mattered was whether five prospects asked for a demo. They got dozens, and a few paid before the trial was over.
That one number keeps the trial short and stops it drifting into a hobby. It also settles the kill decision before anyone is emotionally invested, and when the builders are colleagues or friends, you want the number to make that call.
Pick a number that measures demand: demo requests, signed letters of intent, a handful of paying customers. Finishing the dashboard doesn’t count.
Spin It Out When It Passes
Once the idea proves itself, it should leave the parent company and become its own entity. Keeping it inside as a side project caps how much the builders can earn, and they’ll eventually notice that the upside belongs to someone else.
The shape the agency used is a good default:
- The owner keeps a majority, a little over half, and the builders share the rest.
- The new company pays the agency for the code and IP built on agency time, so ownership is clean from day one.
- The owner funds growth with a loan at a modest interest rate instead of buying more equity. The money comes back before profits are shared, and the cap table stays simple.
- The builders invest their own money as well, paid in quarterly installments over the first year, so they carry part of the risk alongside the owner.
- Their management fee starts below their old salary, on purpose: they should feel what an owner feels when revenue is thin. It steps up at revenue milestones, but only once the company has held a milestone for two consecutive months.
- An independent tax advisor checks the numbers behind each step. That settles arguments before they start, such as whether an annual subscription counts toward one month’s revenue or is spread across twelve.
- The shares vest over three years, with standard good, bad and early leaver terms.
The owner had lawyers and a tax advisor build the structure, and you should do the same. In hindsight they’d add one clause: if results fall for several months in a row, the management fee steps back down. That protects the owner’s investment if the market turns.
Buy Back Their Time
The agency also lost two of its best people to their own company, and the owner called it a big loss. The fix was an agreement that the agency can keep hiring them at a fixed hourly rate, up to a monthly cap, for the following year.
That gives the agency time to hire and train replacements, and the new company some extra income while it grows.
Where This Goes Wrong
The usual failure is choosing builders for the wrong reason. The role is mostly selling: demos, follow-ups, onboarding, support. Someone brilliant at the underlying job who dislikes talking to customers will struggle, however good the product is.
The second is running too many bets. One trial with full attention will beat four running at once, especially if you’re the one helping with the technical side.
The third is vagueness about who these people are. If you pay them during the trial, they’re contractors or employees with an option on equity. If they work for equity alone, they’re co-founders. Both are fine. Letting it stay unclear is how working relationships, and friendships, break.
It Doesn’t Have to Start With Your Own Staff
The same structure works with people outside your company. I know several capable people who are between jobs or looking for extra work, and some of them have spent years doing exactly the kind of work software could take over.
For them the trial has to be paid, because equity doesn’t cover rent. That raises your cost, but it’s still a fixed, small amount, and the single demand number protects you as well as it protected the agency.
Building the software is now the cheap part. What’s still rare is the judgment of someone who has done a job for ten years, and most people with that judgment never got the chance to own what they know. A business owner can now offer them that chance for the price of a few months of part-time work. I think more of us should.
Related
Leave a Reply