Your Finance Team Doesn't Trust AI. Here's Why That's Reasonable.
User resistance to AI in ERP is not an adoption problem. It is a design problem. The finance teams that are sceptical of AI in Business Central are responding rationally to the evidence available to them. Here is what changes that.
In almost every AI deployment conversation that involves a finance team, there is a moment where the mood in the room shifts. The technology has been demonstrated. The use case makes sense. The efficiency gains are credible. And then someone asks: but how do we know it is right?
The question is not rhetorical. It is not obstructive. It is the most important question in the room, and it is one that most AI implementations in ERP are not fully equipped to answer.
The finance team's scepticism about AI is not irrational. It is the appropriate response of a group of professionals who are accountable for the accuracy of the financial record, who have seen what happens when that record contains errors, and who have been given a tool that produces outputs without reliably explaining how it reached them.
That is not an adoption problem. That is a design problem.
What the finance team is actually asking
When a finance team member says they do not trust the AI, they are usually asking one or more of four specific questions, even if they do not articulate them in these terms.
Can I see how it reached that conclusion? The output of an AI system is only as useful as the finance team's ability to evaluate it. A pricing recommendation that arrives without a reason code is not a recommendation. It is an instruction. Finance professionals do not act on instructions they cannot interrogate.
What happens if it is wrong? Every finance team has experienced a system error that produced downstream consequences. The question is not whether the AI will ever be wrong. It is what happens when it is. If the answer is unclear, or if the error correction process involves manually unwinding a series of transactions that the AI processed automatically, the risk calculation changes significantly.
Who is responsible when it goes wrong? Accountability is not a bureaucratic concern for a finance team. It is a professional one. Finance professionals are personally accountable for the accuracy of the records they are responsible for. An AI system that makes decisions without a clear accountability chain creates an ambiguity that experienced finance professionals are right to be uncomfortable with.
Will it keep doing this, or is it a coincidence? A single accurate output is not evidence of a reliable system. The finance team needs to see a pattern of accurate outputs, across the range of scenarios the AI will encounter, before it adjusts its default posture from scepticism to trust.
What changes the calculation
The finance team's trust in AI is not built through reassurance. It is built through evidence. And the evidence needs to be specific, structured, and available at the point where the trust decision is made.
Confidence scores change the calculation. When a finance team member can see, for every AI output, a quantified statement of how certain the system was when it made the decision, they have a basis for evaluating the output rather than simply accepting or rejecting it. A high-confidence output on a well-understood transaction type earns a different level of scrutiny than a low-confidence output on an unusual case. That differentiation is the beginning of a rational trust relationship.
Reason codes change the calculation. When a finance team member can see, for every AI action, a structured explanation of which rule applied and what data the AI was reading, the output becomes contestable. If the reason code references a pricing rule that the finance team knows has an exception for this customer, they can intervene. If the reason code references data that the finance team knows is outdated, they can flag the data quality issue. The AI is no longer a black box. It is a system whose reasoning can be examined.
Approval thresholds change the calculation. When the governance framework gives the finance team control over which decisions the AI may make autonomously and which require their review, the trust relationship has a defined boundary. The team is not being asked to trust the AI with everything. They are being asked to trust it with the categories of decision for which the evidence of its reliability has been established. That is a much more manageable ask.
The role of the BC partner in building trust
For BC implementation partners, the trust problem is a deployment problem that needs to be addressed before go-live, not after.
The most effective approach is to involve the finance team in the governance design process. Not as stakeholders who are informed of decisions made elsewhere, but as participants who define the approval thresholds, the exception categories, and the review processes that will govern the AI's behaviour in their environment.
When the finance team has defined the rules the AI operates within, their relationship to its outputs changes. They are no longer evaluating a system someone else configured. They are reviewing outputs produced within a framework they designed. The accountability question has a clear answer. The trust question follows from the answer to the accountability question.
The partners who build this process into their implementation methodology will find that the finance team's trust is established before go-live rather than earned painfully after it. That is a better implementation experience for the client, a cleaner go-live for the partner, and a more durable foundation for the ongoing managed services relationship that follows.