The Hardest Part of an AI Program Isn't the AI
Most tax leaders I talk to are not short on technology. They have provision systems, cloud ERP, automation platforms, more AI tools than they can evaluate, and more vendors than they can return calls to. What they are short on is adoption.
That gap is the whole game. An AI program in tax doesn't fail because the model is weak or the integration breaks. It fails because the organization keeps running the way it always has while a new tool sits on top, used by a few, trusted by fewer, and quietly abandoned the first time a deadline gets tight.
The technology is the easy part. The people are the hard part. And the people are where the value is.
You can't bolt AI onto an old operating model
Here is the pattern I see most often. A function buys a capable platform, runs a clean implementation, and expects behavior to change because the software now allows it to. It doesn't. People revert to the spreadsheet they trust, the manual reconciliation they understand, the review step they've always owned.
Layering modern technology on a model designed for a different era doesn't produce transformation. It produces digital complexity — more systems, same workarounds, and a new line item on the budget. AI makes this worse, not better. Point agentic tools at fragmented data, unclear ownership, and disconnected processes, and you don't get productivity. You get faster dysfunction, at scale, with less visibility into how the answer was reached.
Enterprise-grade AI demands a redesigned operating model underneath it. The tool and the model are one purchase, not two. Treating them separately is the single most expensive mistake I watch tax functions make.
Transformation is not a night job
The other pattern is just as common. Leadership approves the program, then expects an already-stretched team to absorb it on top of compliance deadlines, audit demands, and constant business change. Transformation becomes the thing you do after the real work — the night job.
It never holds. Momentum stalls, steering meetings turn into status updates, and competing priorities win every time because they always have a deadline and transformation never does. The functions making real progress treat change as a standing capability with dedicated capacity, not a temporary project with a finish line. Technology changes in months. Organizations change over years, and only when someone owns the change full-time.
Communication is the work, not the wrapper
People support what they help build. That single idea reorders how a serious program should run.
It means involving the people who will use the system while it's being designed, not training them the week of rollout when resistance and anxiety are already baked in. It means saying out loud, repeatedly, why the change is happening, how roles will shift, and what good looks like. Silence doesn't read as neutral. It reads as threat. Uncertainty becomes resistance, and resistance slows adoption long before leadership sees a number move.
Communication isn't the kickoff email. It's the continuous, sometimes tedious work of keeping a few hundred professionals aligned on why their day is changing — and what they get out of it.
Sponsorship that survives the kickoff
Most tax functions know executive sponsorship matters. Fewer treat it as more than an approval and a slide at launch. Real sponsorship is sustained and visible: leadership consistently signaling that the effort matters, that priorities will stay aligned, that modernization isn't optional, and that learning will be supported when — not if — the team makes mistakes.
The strongest cultures send a clear message: we are going to modernize, it will require change, we will make mistakes, and we are committed to the outcome and to the people getting us there. That balance of commitment and permission is what lets a team experiment instead of hide. Functions that lead with fear suppress exactly the transparency and learning a program needs to work.
And here's the quiet killer: organizations launch with enthusiasm and a detailed technology plan, then keep rewarding the same behaviors, the same risk avoidance, the same priorities as before. People notice the contradiction immediately. If the incentives didn't change, neither will the behavior.
What the functions that get this right actually do
The tax functions making durable progress share a short list of habits. They treat transformation as a permanent capability, not an episodic project. They fund real capacity for it instead of asking people to do it after hours. They invest as heavily in communication and training as in the platform itself. They build governance that survives leadership turnover. And they measure adoption as seriously as they measure implementation — because a tool nobody uses is a cost, not a capability.
None of this is soft. It is the rigor that global tax operations already apply to controls and audit, pointed at the human side of change. The discipline is the same. Only the object is different.
The real conclusion
The tax function at large multinationals is entering a period of enormous change, and AI will be at the center of it. But the organizations that win won't be the ones with the best tools. They'll be the ones that help their people move through change — deliberately, repeatedly, and with real investment behind it.
That conviction is why I built ai4tax as a program, not a product: deep tax expertise, hands-on technology implementation, and change management, delivered together. Because in tax, that's the only combination that actually holds.
If you're a CFO or Head of Tax thinking about enterprise AI built for the rigor your operations demand — let's chat.
← All insights