How AI can cut the effort of an ERP change request by half
A step-by-step look at where AI saves time in an ERP change request, from analysis to client reporting, and where people must stay in control.
Every ERP support contract lives on change requests: a new report, a change in an approval flow, an adjustment after a regulatory update. They are small projects, but there are many of them, and each one goes through the same steps. That is exactly where AI pays off.
The same change request, step by step
Take a typical change request of 18 consultant-days. This is an illustrative scenario, not data from a specific client, but the order of magnitude matches what I see in Oracle Fusion and EBS support.
- Analysis (4 days → about 2). AI drafts the requirements, the impact analysis and the questions for the business from the meeting notes. The consultant reviews and decides.
- Documentation (3 days → about 1). Functional and technical documents are generated from the analysis and the configuration, then validated by the team.
- Build and test (6 days → about 4). AI proposes test cases and test data, and helps with reports and extensions. Configuration and final testing stay with the experts.
- Support follow-up (3 days → about 1.5). Incidents after go-live are analysed faster: logs, known issues and similar tickets are summarised in minutes.
- Client reporting (2 days → under 1). The weekly status is written from the real data of the project instead of being rebuilt by hand.
In total, around 9 days instead of 18. Multiplied by the change requests of a year, the saving is the size of a full project.
What does not change
AI does not decide. A consultant validates every analysis, every document and every test result, and the client signs off as before. AI removes repetitive work; it does not remove responsibility. That is also why it must run inside clear rules: which data it can see, where it runs and who reviews its output.
How to start
Do not start with a big programme. Pick one recurring flow, for example the monthly reporting or the documentation of change requests, measure the time it takes today, apply AI for a month and measure again. If the numbers are good, extend it to the next step.
That is how I work in the ERP accounts I manage: AI is not a side tool, it is part of every step of delivery.