The AI adoption curve: where does your company stand?

Every company I talk to is using AI to some degree. Someone in finance drafts emails with ChatGPT. A sales manager summarizes calls with whatever came free with the CRM. Leadership has a line about AI in this year's plan. So I've stopped asking whether a company uses AI. I show them where they sit on the AI adoption curve, and what it takes to reach the next stage.

Our engineers work inside private equity portfolio companies, franchises, multi-site operators, manufacturers and car washes. Day to day, those businesses share very little. A car wash counts cars and memberships. A manufacturer counts parts, shifts and scrap. Yet every one of them sits somewhere on the same curve, with the same five stages, and most of them sit lower on it than they think.

Stage one: exploring

At the bottom of the curve are the companies still exploring. Leadership wants AI in the business, often because a board member, an investor or a competitor brought it up, and there is no plan behind the interest yet. The data is scattered across a dozen systems and a shared drive. A few employees use ChatGPT on personal plans, pasting in whatever they need that day.

Every company starts here. The risk is staying. A year of scattered experiments builds a few good habits and leaves the cost of running the business where it was.

Stage two: identifying

One step up are the companies identifying. They accept that AI can help. What they can't say is where, or whether it will earn back what it costs. The waste is real and spread thin: a team that rekeys purchase orders from email into the ERP every morning, software licenses nobody has opened since the spring, a month-end close that waits on three spreadsheets from three departments. Each one costs money, and nobody has put a number on any of them.

The work at this stage is a ranking. Map how work actually moves through the company, price each tedious step in hours and dollars, and put the list in order. The top items become the business case for the first build.

Stage three: connecting

Then come the companies that need to connect their data. Their information lives in systems that never talk to each other: QuickBooks for the books, Salesforce for the pipeline, a point of sale at every location, an ERP in the warehouse. Every report starts as an export and ends as a spreadsheet. Copilot can find last quarter's board deck, but it can't tell you the correct revenue figure, because the answer depends on which system you ask.

This is where I find most companies, usually paying for AI licenses nobody uses. The AI has nothing reliable to read. What this stage needs is a data foundation: one place where the accounting, sales and operations data sits together, cleaned and joined, with access set per person. Once that exists, a question about the business gets one answer.

Stage four: implementing

Above that, the data sits in one place AI can read, and the company starts implementing. Custom AI applications get built around specific workflows, with the people who run those workflows in the room: invoice matching for accounts payable, contract review for operations, the weekly report that used to take a manager's Monday. Each tool is measured against the business case from stage two.

Costs start to come down here, and training starts to matter most. A tool the team doesn't trust goes unused, whatever it cost to build.

Stage five: automating

At the top are the companies that are automating. They have redesigned how their people and their agents share the work. Agents handle the follow-ups, approvals and data entry, using the company's own data and its own rules, and the team spends its time on customers and decisions. Output grows, expenses shrink, and the AI investment shows a positive return on the books.

Why most companies sit lower than they think

Companies tend to place themselves by the tools they have bought. A Copilot license for every employee feels like stage four. The curve measures three other things: whether the company's data is connected, whether AI is doing measured work, and whether anyone has redesigned a process around it. By that measure, a company with expensive licenses and siloed data is at stage three, and plenty of companies with an AI strategy on paper are still exploring.

The stages also come in order. Skip the identifying and the first tools are built on a guess about where the money is. Skip the data foundation and an application ends up reading numbers the business itself doesn't agree on, which tends to show up the first time two reports disagree in a meeting.

Climbing the curve

Every company can climb the curve. It takes proper planning, a data foundation, applications built for the business, and hands-on training for the people who use them.

At Straterai we start every engagement at the stage a company is actually on and build up from there. Wherever your company is today, the first step is an honest read of where it sits.

James