A leadership team says: "we will integrate the acquisition within a year, and we will do it smarter than last time." That last sentence conceals an assumption. Smarter these days often means: with less manual work in the overlapping functions, and with AI taking over part of the integration work instead of people doing it alongside their regular work for two years. That assumption is rarely voiced, and so it is never tested either.
The difference between this ambition with AI work and the same ambition without AI is not cosmetic. Without AI, integration mainly requires project capacity: people temporarily freed up to connect systems, compare processes and merge reports. With AI, part of that work shifts to a different question: which comparisons, connections and handovers can a system do itself, which ones require a person who approves or rejects with reason, and which remain human work because the context is too specific to automate. That distinction determines which capabilities are needed, and those are different from the capabilities a classic integration plan assumes.
In one acquisition, the matching of customer and supplier data between two systems is already largely done by AI, with an employee assessing the exceptions. In another acquisition, exactly the same matching still happens entirely by hand, in spreadsheets, by people spending weeks comparing fields. The difference does not lie in the sector or the size of the company. It lies in whether the organization has already documented its own data and processes in a way that an AI system can use, and whether someone has been designated to approve the outcomes.
The same shift plays out in contract analysis, in merging reporting structures and in the initial triage of customer communication after the acquisition. Some of that work can be taken over by a system. A larger part requires oversight: a person who assesses the output and accepts or returns it with reason. And part remains human work, because it involves negotiation, culture or political sensitivity within the new organization. An integration plan that does not distinguish these three categories plans capacity in a way that turns out not to hold up afterward.
The first capability is an up-to-date picture of which tasks in the integration can be taken over, which ones require oversight and which do not. Most leadership teams lack this picture not out of unwillingness, but because no one has ever systematically drawn it up for this specific integration process. The underlying question of which work in this company can genuinely be taken over by AI is answered per task by the work scan of FTE TO AI, and that provides a starting point that is more concrete than an assumption.
The second capability is a decision right: who may approve an AI-generated comparison, connection or report without falling back on the old, entirely manual route. Without a designated role holding that mandate, oversight remains a double-check alongside the system rather than a replacement of part of it, and the intended capacity benefit disappears.
The third capability is data hygiene at the level that makes automatic matching and comparison reliable. Acquisitions almost always bring together two datasets that do not share the same structure, definitions or quality. An AI system that has to work on that data is only as good as the data allows, and that is a capability that must be in order months before the first integration step, not during it.
The fourth capability concerns the people whose function overlaps as a result of the acquisition. If part of their work is taken over by AI, a question arises about redeployment, task packages or terms of employment. Separate legal requirements apply to that, and those are not addressed here.
An integration ambition never stands alone. It often touches on flattening the organization, because acquisitions often coincide with merging management layers. It touches on becoming more customer-focused, because customers of the acquired party are the first to notice, during the integration, whether the promise holds up. And it touches on becoming less dependent on key people, because an acquisition often places extra strain on exactly the people who carry the most undocumented knowledge in their heads. Anyone who tests these ambitions separately misses the friction that arises when they draw on the same capacity.
In a year, this ambition has succeeded if the overlapping processes are no longer running twice in parallel, if the role that approves AI outcomes actually exists and is used, and if the number of manual corrections on merged data has structurally decreased rather than being temporarily relieved. It has not succeeded if the integration is complete on paper but the freed-up hours are nowhere to be found, because no one had established beforehand where those hours would come from.
To make that ambition measurable before the process starts, it helps to know how you make a vision measurable and, if the integration is too large to complete within a year, how you phase an ambition that is too big for a year.
The free readiness check from FTE TO AI consists of eight short questions, one per dimension, and gives a picture of where the organization is furthest along and where it is not. The full ambition test, with the four layers from vision to roles and decision rights, is under construction.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.