"AI will run your finance department." "AI will replace your ERP." "AI makes implementation consultants obsolete." None of these are true, and believing them costs money. Seven common AI claims in ERP, what is actually true behind each, and how to tell the difference when a vendor is selling.
We wrote recently that Skynet isn't coming for your purchase ledger. It still isn't. But the claims about AI in ERP have got louder, and some of them are costing businesses money. Here are seven we hear often, and what's actually true behind them.
The reality: AI agents handle well-defined, repetitive tasks (keying invoices, drafting orders, matching bank lines) with a person reviewing the result. Judgement, relationships, controls and accountability stay with people. AI changes finance jobs; it doesn't remove the department (see AI and jobs).
The reality: AI needs a system of record to work on. Agents read and write Business Central's data, under its permissions, with its audit trail. The ERP becomes more important, not less, because it's the thing the agents trust.
The reality: Copilot's assistive features are included. The task agents are charged on consumption, per transaction (what's included).
The reality: it works on clean data and written processes. Duplicate suppliers, wrong units of measure and undocumented exceptions make agents fail, or worse, succeed confidently at the wrong thing (see getting ready for agents).
The reality: AI makes the typing faster. Someone still has to design the processes, decide what should be customised, clean the data, set the permissions and own the outcome (see AI-assisted development).
The reality: nothing is. Good agents on clean data are very accurate on routine items and hand the rest back. That's why a person reviews, and why you measure it monthly (the five numbers).
The reality: the businesses that benefit most are the ones that prepared: current versions, clean data, written processes, good permissions. Rushing an agent onto a messy system isn't being early; it's automating the mess. Preparing now and adopting carefully is not falling behind.
AI in Business Central is useful, real and getting better quickly. It's also software, with costs, limits and risks, and it rewards the businesses that treat it that way.
