Strategy
12%
report redesigning work at scale, with a new operating model behind it. Nearly half added AI to the old map.
Deloitte AI Institute Pulse Check, 20261Headcount is not the only way to add capacity.
Companies talk about AI transformation when they mean they are tired of adding people to a process that grew too many handoffs. The useful move is leverage: the same team, a better workflow, and a result that shows up in the operating numbers.
How I see it
Operating leverage, not a transformation program
Operating leverage is the reason a COO should care about AI. If the work gets faster, cheaper, or able to absorb more volume without a matching increase in staff, the economics of the function change.
That does not require renaming the company around AI. It requires finding the workflows where coordination, exceptions, and document handling consume the calendar, then deciding how far to go: assist the people, redesign the process, or assign an outcome to an agent.
Transformation language hides the decision. It suggests a program, a steering committee, and a multi-year operating model. Those structures can wait. The first proof is one process with a baseline and a production path.
The companies that get this right treat AI as a capacity decision. They ask what the function could handle if the expensive steps moved. They do not ask how to become an AI-native organization.
How the engagement runs
What you actually buy.
A service is a sequence with an artifact at each step. It is not a transformation program you purchase as a bundle.
Name three bets
Each bet has a workflow, an owner, a baseline, and a reason the economics could matter.
A ranked list a COO can fund this quarter.
A capability map or a vendor landscape.
Write the no-list
Record which ideas are real problems that are not AI problems, too rare, or political.
Popular ideas can be stopped in writing.
A roadmap where everything is still a yes.
Sequence one workflow that can ship
The first bet must reach production without a platform program.
A contained path with an owner and a kill criterion.
A transformation sequence that needs a new operating model first.
Do not buy this for
These engagements fail for predictable reasons.
Funding a program instead of a workflow
A transformation office can produce governance and decks. It cannot create leverage until a real process changes.
Measuring activity instead of capacity
Licenses issued and prompts written are not operating results. Cycle time, cost per case, and volume per person are.
Assuming more AI means fewer people immediately
The first win is often the same team doing more, with less overtime and less rework. Headcount decisions come after the workflow is stable.
Before we start
Questions that decide whether to engage.
Is the function adding people because volume grew, or because the process got slower?
If the process got slower, leverage is a process problem first.Which steps consume calendar time without creating the customer or financial result?
Handoffs, queues, and packet assembly are usually the leverage points.If volume rose 30 percent, what would break first?
That breakage point is a better target than a generic AI program.Can you add capacity without adding a new layer of review?
If every new tool creates another checkpoint, you are adding cost.Will leadership accept a contained win instead of a company-wide program?
If the only acceptable answer is transformation, the work will stall.
How we size it
Leverage testoutput the function can absorb ÷ people and loaded cost required to absorb it
AI is useful when that ratio improves without hiding work in unpaid overtime or quality debt.
This is the wrong engagement if
Do not start this engagement yet.
- Leadership wants a transformation brand more than a capacity result.
- The bottleneck is a policy or a staffing model that AI cannot change.
- There is no volume. A rare process cannot create leverage.
- The company will not retire steps after a new path works.
A useful next step
Bring one expensive workflow. Get guided into production.
We guide the implementation, go deep on the technical path, and stay hands-on through operations — or tell you when a simpler answer is better.
Discuss an AI opportunity