Most companies put AI into one department at a time. Finance gets a tool for the close, marketing gets one for content, sales gets one for follow-up, and each team gets faster at its own part of the job. The work that slows a company down usually sits between those teams, in handoffs that no team's tool touches and no manager's job description covers.
A group of economists argues as much in a Harvard Business Review piece, The Deployment Failure Holding Back Your AI Investments. Their case is that the gains from most AI purchases stop at the edge of the department that bought them, which goes a long way toward explaining why so much spending has shown so little in productivity. I agree, and I think the fix starts with naming who owns the handoffs.
Each department got faster and the company did not
Each department head can show that their own tool works, and most of them can show it honestly. No one is measured on whether an order moves from sales to operations to billing faster than it did last year. So no tool was bought for that trip, and the days the order sits in someone's queue go unwatched.
That is how a company can be full of AI and no faster overall. One of the surveys the HBR authors cite, a National Bureau of Economic Research working paper, now covers nearly 9,000 senior executives in seven countries. Most of their firms use AI. Asked what it has done for "annual labor productivity growth" over the last three years, the executives put the lift at 0.2%. Most companies do not fail at AI because of the technology. They fail because nobody owns it, and the work between departments is the work most likely to have no owner at all.
Give the handoffs an owner before you give them a tool
The fix starts with a map. Pick one piece of work that crosses several departments, such as a new customer order, a month-end close or a new hire, and walk it from start to finish. Write down every step, who holds it, and how long the work sits between one step and the next. Then share that map with every team on it, so each of them can see where their part slows the whole.
Look hardest at the handoffs where two teams use the same word to mean different things, such as a deal that sales calls closed and finance does not. Once the map shows where the work waits, you can decide whether AI belongs at that spot or whether the fix is a plain decision about who goes first.
Someone has to hold this job, and it is rarely any one department head, because each of them is measured on their own piece. In a mid-market company it usually lands with a COO, a CFO who already sees every team's numbers, or a fractional executive brought in to draw the map and keep it current while the owner keeps running the business.
If your departments rarely pass work to each other, start inside them
Some companies run separate business lines, each with its own team from first call to final invoice, and very little moves between them. For a company like that, putting a tool inside each team is a sensible place to start, because most of the work never crosses a line.
Walk one piece of work end to end before the next AI request
Before the next AI request reaches your desk, ask someone to walk one piece of work through every department it touches and bring back the map: each step, who holds it, and how long it waits in between. The longest wait on that map is where your next dollar should go, and it may not be a tool.
The partnership behind that map is the subject of Share the vision, then let your partner draw the map, and deciding which lever to pull first is covered in Pricing, people and AI are one decision. If you want to see who could hold the job in your company, start with what a fractional COO or CFO does. And if one handoff in your company has been slow for years, tell me which one.
John Wyatt · John Wyatt's page · All posts · RSS