Your 2027 AI Plans Have a Competitor, and It Is Your ERP
Here is a scheduling conflict most manufacturers have not put on one page. SAP ends mainstream maintenance for ECC 6.0 on 31 December 2027. Migrations to S/4HANA typically run 18 to 36 months. Industry estimates suggest a majority of ECC customers had not started a serious migration as of recent surveys, and the compatibility packs that let companies defer the harder parts expired at the end of May this year.
Now put your AI roadmap next to that. Both projects need the same three things: budget, the attention of your senior IT people, and above all the handful of employees who genuinely understand how your data and processes actually work. One of those projects has a hard external deadline. The other does not. You can predict how that resolves.
The squeeze is on people, not licences
The cost that hurts is not the software. It is that S/4HANA demand is concentrating into a two-year window, with consulting rates widely expected to climb and experienced talent in short supply as the deadline nears. Treat the specific percentages from consultancies with appropriate suspicion – they are selling urgency – but the direction is not in doubt, because the deadline is fixed and the supply of people who have done this before is not.
Internally the constraint is sharper. Every manufacturer has two or three people who know why the material master is structured the way it is, which fields are trustworthy, and what that undocumented workaround in the production planning module is actually for. Those people are the critical path for the migration. They are also the critical path for any AI project touching operational data, because AI on ERP data is mostly a data-understanding problem. There is one of them, and two queues.
What this means depends on which side you are on
If you are running ECC and have not started, the honest read is that your AI ambitions for 2027 should be modest and narrow. Not zero – but agents that touch quoting, order processing, or planning are reading from a system you are about to rebuild, which means you would be automating against a moving target and rewriting the integration afterwards. Better to spend that capacity on use cases that sit outside ERP: document handling, maintenance data, quality inspection, engineering support. Those survive the migration intact.
If you are already on S/4HANA or not an SAP shop, this is a competitive window rather than a problem. A large share of your competitors are about to spend two years and a lot of money on a project that produces no new customer value, because it is a platform replacement rather than an improvement. Their attention is committed. Yours is not.
The part worth arguing for internally
There is a version of this that goes better, and it requires making an argument to whoever owns the migration budget.
An ERP migration forces you to look at every master data object, every process variant, and every customization you have accumulated over fifteen years. That is precisely the audit that every AI initiative discovers it needs and cannot justify funding on its own. Done as a pure technical conversion – lift the old mess into the new system, change nothing, finish faster – you arrive in 2028 with modern software running the same undocumented processes, and your AI problem is exactly where it was, only more expensive.
Done deliberately, the same project produces clean master data, documented process logic, and a data model something can actually read. That is the foundation the AI roadmap has been waiting for, paid for out of a budget that is being spent anyway.
The choice between those two outcomes is made early, in how the programme is scoped, and it is usually made by people who are not thinking about AI at all. If you want the second version, the time to say so is while the scope is still being written – not in 2028, when someone asks why the new system still cannot tell you what a job actually costs.
“Content generated using AI”