Human Expertise
AI can generate the answer in seconds. It still doesn’t know which question to ask.
The most dangerous moment in a Business Central project isn’t when something breaks. It’s when something works perfectly — but solves the wrong problem. That’s a human catch, not a prompt.
We use AI tools in our work. We’re not here to argue against them. They’re genuinely useful for accelerating analysis, drafting documentation, and exploring technical options faster than was possible two years ago.
But we’ve started to notice a pattern in the conversations we have with new clients. Someone tried to use AI to solve a Business Central challenge. The AI produced something. It looked reasonable. It was implemented. And then, somewhere between three weeks and six months later, reality arrived.
Not because the AI got the answer wrong. Because nobody asked it the right question in the first place.
“A prompt is only as good as the understanding behind it. And understanding a Business Central environment — really understanding it — takes years, not tokens.”
The gap no one talks about
AI optimises for the question asked. Experience knows when to ask a different one.
Here is what AI does brilliantly: it takes a well-formed question and produces a well-formed answer. If you ask it to write an AL extension that does X, it will write an extension that does X. If you ask it to design a workflow for process Y, it will design that workflow.
Here is what it cannot do: walk into a client meeting, listen to someone describe their approval process, and realise that the real problem isn’t the approval process at all — it’s that the purchase order structure was set up incorrectly four years ago and every workaround built since then has been compensating for that original mistake.
That kind of diagnosis doesn’t come from a prompt. It comes from having seen the same pattern fifteen times across fifteen different companies, in fifteen different industries, and knowing what it means before the client has even finished the sentence.
What this looks like in practice
Two approaches to the same problem.
Consider a common scenario: a finance team is frustrated that their month-end reporting takes too long and the numbers never quite add up.
| AI-led approach | Experience-led approach |
|---|---|
| Prompt describes the symptom. AI suggests a reporting extension, a Power BI connector, or a new dimension structure. Solution is implemented. Numbers are faster — but still don’t add up, because the underlying posting inconsistencies were never identified. | Consultant recognises the symptom as a data integrity pattern. Traces it back to three years of inconsistent dimension posting. Recommends a data remediation sprint before any reporting tool is added. Numbers finally add up — for the first time. |
The AI solution wasn’t wrong. It was an accurate answer to the question it was given. The experienced consultant asked a better question.
“The value of a senior BC consultant isn’t that they know more facts than an AI. It’s that they know which facts matter — and they’ve seen what happens when you ignore them.”
The guardrail role
Experience isn’t the opposite of AI. It’s the thing that makes AI safe to use.
We’ve reframed how we think about this internally. Our senior consultants aren’t competing with AI tools — they’re the guardrail that makes those tools viable in a production Business Central environment.
An AI can generate an AL extension. Our consultant reviews it and spots that it touches a table that a critical integration depends on — something the AI had no way of knowing. An AI can propose a new approval workflow. Our consultant flags that it will break at year-end because of how the fiscal periods are configured. An AI can suggest a Copilot prompt structure. Our consultant points out that the underlying data it will query hasn’t been validated and will produce numbers the business will trust but shouldn’t.
Speed without that guardrail isn’t efficiency. It’s risk that hasn’t surfaced yet.
What this means for you
Choose your partner for their judgement, not their output speed.
If you’re evaluating BC partners in this environment, the right question isn’t “how fast can they deliver?” The right question is “what will they catch before it becomes a problem?” That’s a harder thing to demonstrate in a proposal. It shows up in track record, in the depth of questions they ask in the first meeting, and in whether they’re willing to push back on a requirement that doesn’t serve you.
At NAV SEAL, that’s the standard we hold ourselves to. We use every tool available — AI included. But the judgement that determines when and how to use those tools, and when to override them, is ours. Built over years. Not generated on demand.
Looking for a partner who brings judgement, not just speed?
NAV SEAL works with Business Central customers who want experienced architecture oversight — not just faster delivery. If you’re planning an AI or automation initiative and want a second opinion before you commit, let’s talk.
Visit navseal.com or connect with us on LinkedIn to start the conversation.
#BusinessCentral #Dynamics365 #HumanExpertise #ERPConsulting #AIStrategy #MicrosoftPartner #NAVSEAL #DigitalTransformation
