Two projects sit on most owners' lists and never get done. One is documenting how the business runs. The other is automating some of it.
They are the same project.
That is not a productivity aphorism. It is a specific claim about where value comes from in a founder-led business, and it has a number attached: owner dependence is the most common valuation detractor we see, and the work that fixes it is the same work an AI system requires before it can do anything useful.
Why buyers discount owner dependence
Start with what a buyer is actually purchasing.
If the business cannot run for a month without you, the buyer is not acquiring a business. They are acquiring a job, plus the risk that the person who knows how to do it leaves. They price accordingly — through a lower multiple, a larger earn-out, a longer transition period, more of the price held in escrow, or all four.
This is not a subjective judgment made at the end. It surfaces concretely in diligence:
- Who approves an exception to standard pricing, and on what basis?
- What happens to the schedule if the operations manager is out for a fortnight?
- How does a new technician learn the job?
- Which customer relationships are institutional and which are personal to the founder?
- What is the process when a job goes wrong?
Answered with "I handle that," each question is a mark against the file. Answered with a document, each is a non-event.
The uncomfortable version: the more indispensable you have made yourself, the less your business is worth. Most owners have spent a decade optimizing for the opposite.
Why an AI agent needs exactly the same thing
Now the other project.
An AI system cannot do work you have not described. It has no access to the reasoning you have never articulated — which vendors get exceptions, what "urgent" means here, when to escalate rather than proceed. Hand it a vague instruction and it produces plausible output that is confidently wrong in ways you will not notice immediately.
What it needs is a written procedure: the trigger, the steps, the thresholds with real numbers, the exceptions, and what evidence should exist when it is done. We cover the mechanics of writing those in Skills: How to Tell an AI Agent How Your Business Works.
Read that list again and compare it against the diligence questions above. It is the same document.
The dividend
| The work | Automation payoff | Valuation payoff |
|---|---|---|
| Write the procedure with real thresholds | Agent can execute it | Answers the diligence question |
| Name the exceptions and escalations | Agent stops instead of guessing | Demonstrates controlled operations |
| Define what "done" looks like | Output becomes auditable | Shows management can verify quality |
| Move a task off the owner | Capacity released | Owner dependence reduced |
| Repeat across the recurring work | Compounding operational lift | The business becomes transferable |
One effort, two returns. Most owners budget for neither because each looks like overhead on its own. Together the case is much harder to argue with.
What this looks like in practice
Start with the work that already repeats. Invoice matching, appointment booking, quote follow-up, inbound call handling, month-end reconciliation. High frequency, clear right answer, currently sitting with someone whose time is worth more than the task.
Write the procedure before you evaluate any tool. This is the order almost everyone gets wrong. The procedure tells you whether the task is automatable at all — if you cannot state what a correct outcome looks like, that is the finding, and no vendor demo will change it. It also makes tool selection much easier, because you are testing against a specification rather than a feeling.
Automate the ones that qualify. File the rest. Some procedures will turn out not to be worth automating. Write them anyway. The document still does the valuation job even when no system ever runs it.
Keep the artifacts. The procedures, the exception logs, the record of what changed after the first fortnight of running. In diligence this is evidence that the business operates on systems rather than on memory.
What buyers will actually ask about your technology
Worth knowing, because the answer has changed in the last two years.
Buyers are no longer impressed by the presence of AI. Roughly 89% of US small businesses report using it; under 9% use it in the production of goods or services. Saying "we use AI" places you in the first group, and the first group is not paid for.
What gets credit in diligence:
- Does it survive your departure? A system running on documented procedures transfers. One running on the founder's undocumented prompting does not.
- Has it changed the cost structure measurably, or does it sit alongside the same headcount doing the same work?
- Who checks the output, and what happens when it is wrong? An automation with no verification step is a liability, not an asset.
- What data does it touch, and under what agreements? Client data flowing through a system nobody has reviewed is a diligence finding.
We examine what acquirers actually credit and discount in Does AI Adoption Increase Enterprise Value?.
The sequencing that works
If a transition is 12–24 months out, the order matters:
- List the recurring work — everything that happens weekly or more often and follows roughly the same shape.
- Rank by how much of it runs through you personally. That ranking is your owner-dependence map, and it is what a buyer will construct anyway.
- Document from the top of that list down, whether or not you intend to automate it.
- Automate the ones that qualify — clear right answer, high frequency, low judgment.
- Keep the evidence of what changed.
That is a genuine operational program, not a filing exercise, and it takes quarters rather than weeks. Which is exactly why it cannot be done during a live sale process, and why the owners who start early transact on better terms than the ones who start when a buyer calls.
The honest caveat
None of this is a shortcut to a higher multiple. Documentation alone does not make a business more profitable, and a buyer will not pay more for a binder.
What it does is remove a discount. The business that runs on documented process is priced as an asset; the one that runs on its founder is priced as a job with key-person risk attached. The gap between those two framings is usually larger than anything you will negotiate at the table.
And the automation you get along the way is, in the strictest sense, free — you were going to have to write it down anyway.
Related
Skills: How to Tell an AI Agent How Your Business Works covers writing the procedures themselves. The Sale Process, Start to Finish covers what diligence actually involves. For a structured review of where a business stands on owner dependence and the other factors buyers price, see the business readiness assessment.
This guide is for informational purposes only and does not constitute investment, financial, legal or tax advice.