Insights
Every ERP vendor and every large integrator is now selling AI, and the volume of it makes calm judgement hard. As an independent firm with no platform to sell, we can say the quiet part plainly: some of what AI does for ERP programmes is genuinely useful, a good deal of it is marketing attached to features you will never switch on, and the gap between the two is where money gets wasted. Here is how we would separate them.
By Christopher Fagan
This is the most real of the benefits. Data is the single biggest risk on most ERP programmes, and machine assistance at profiling, matching duplicates, spotting anomalies and suggesting corrections can take real effort out of the worst workstream on the plan. It does not remove the need for business ownership of the data, but it makes the grind faster.
Generating test cases and scripts from requirements and configuration, and identifying gaps in coverage, is a legitimate accelerator. Testing is where programmes get compressed, so anything that produces more coverage for less effort is worth having.
Drafting process documentation, configuration records and training material, and answering “how does our system do X” from the actual configuration, saves time on tasks that otherwise get skipped and cause problems later.
The suggestion that AI will design your operating model, resolve your process decisions and implement the system with minimal human involvement is not real. The hard parts of an ERP programme are decisions, ownership and change, and those remain human. AI accelerates tasks; it does not make the judgements that decide whether a programme succeeds.
Autonomous agents acting inside your finance and operational processes sound impressive and raise governance and control questions most organisations are nowhere near ready to answer. Automating a process you do not control, on data you do not trust, produces faster mistakes. The order still matters: control and data first.
Much AI functionality is included to win the deal and then never configured, because using it requires clean data, clear process and effort nobody budgeted for. You pay for it in the licence and get nothing back.
AI is changing parts of how ERP gets implemented, genuinely and usefully, at the level of tasks. It is not changing the reasons programmes fail, which are about decisions, data ownership and whether people adopt the new way of working. Keep the accelerators, ignore the theatre, and spend your attention where outcomes are actually decided.
We give independent, vendor neutral advice on ERP selection and delivery. See ERP implementation consulting and vendor selection, or get in touch.
If any of this is live for you right now, the senior people who wrote it are the people you would work with.