An AI-native legacy business is the same company, with the same customers, people and functions, in which AI does the routine preparation, named people make the decisions, and the rules between the two are explicit. Each of those things can be checked, which is what makes the idea useful to an owner or a board.

Picture a third-generation building-supply distributor on a Monday morning. Today, a branch manager starts the week by pulling last week's open quotes out of email, checking stock in one system and prices in another, and deciding which customers to call. In the AI-native version of the same company, the open quotes, stock positions and price exceptions are already gathered and checked when she arrives. She still decides who gets a call and what discount is acceptable. What changes is where her first hours of the week go.

What we mean by "AI-native"

An AI-native company is the same company, with the same customers and people, in which AI does the routine preparation (gathering, checking, drafting, matching), named people make the decisions, and the rules between the two are written down. What you get from it: when a board, a lender or a buyer asks what the company is doing about AI, the answer is already on paper, and the time the team gets back goes where the owner chose.

Five changes make it checkable:

  1. AI prepares the work; people make the decisions.
  2. The company record answers management's questions.
  3. Decision rights and approvals are written down.
  4. The team gains capacity, not a second job.
  5. Success is measured as business change against a baseline.

AI prepares the work; people make the decisions.

Sales, operations, billing and reporting all still exist. What moves is the preparation inside each one: gathering information, checking it against the rules, drafting the first version. Current tools are good at that kind of work when the inputs are known and a person reviews the result. They are poor at judgment calls, relationships, and anything nobody has recorded.

So each function is designed around that split, instead of a tool being added on top of the old way of working. The test is simple: for any recurring task, can you say which part AI prepares and which part a person decides? If the answer is "the person does all of it" or "the tool does all of it", the work has yet to be redesigned.

The company record answers management's questions.

Most established businesses already hold the answers to their hardest management questions, spread across contracts, orders, invoices, emails and reports. Those records were written for people to read. In a company that has made the change, management can ask which customers are buying less than last year, and why, and get an answer built from the records themselves, with the source visible behind it.

This does not require a perfect data warehouse first. It requires deciding which records matter, fixing the few that are wrong, and keeping the source attached to every answer so a person can check it.

Decision rights and approvals are written down.

In a legacy company, authority is often unwritten. Everyone knows the controller approves larger credit limits, until the controller is on holiday. Once AI starts preparing work, unwritten rules become a risk, because a system follows what it was told, whatever everyone else assumed.

So the company writes down, for each workflow that matters: who decides, what needs approval before it happens, what counts as an exception and who handles it, and how to stop or reverse the system. Every system has an owner and a written list of what it may do without asking. Sensitive actions, such as a price change, a payment or a message to a customer, wait for a person's approval until the system has earned more room with a measured track record.

The list is short. It is also the first thing a board, a lender or a buyer will ask to see, and it is what lets a board approve the plan without approving a black box.

The team gains capacity, not a second job.

The common failure is a new tool that adds work: another inbox to check, another dashboard to feed, one person who quietly becomes its full-time minder. In an AI-native company, the time saved on copying, chasing and checking goes somewhere the owner chose in advance: more customer conversations, faster quotes, growth that does not require new hires for administration.

If nobody can name where the recovered time should go, the project is a demonstration.

Success is measured as business change.

Tools adopted and features switched on do not count. What counts is what the owner already cares about: margin, response time, cash collection, errors and capacity, against a baseline taken before the change. A change that cannot be measured does not go first.

In practice that means choosing two or three measures for each workflow, writing down today's number, and checking the same number after the change, in the same way.

Four things it is not

How far along are you?

Five questions to put on the agenda of your next leadership meeting:

If most answers are no, you have found where the plan begins.

Where to start

Start with a decision before any tool: what your company should look like with AI doing the routine preparation, and which three moves come first. Then test those moves against evidence, one workflow or one set of records at a time. Then build the first piece, measure it, and decide what comes next. That order keeps control visible and keeps the business running while it changes.