I use an AI agent to manage my contacts, my travel, and my expenses. I can describe any change I want in one sentence and it happens. And yet I still open the app to look at my upcoming trip.  It provides a great visual and links for me to call a Lyft or Uber.

There is a car wash two minutes from my house. I still wash my car myself frequently.

This explains what most predictions about AI and ERP miss. I’ll come back to it.

McKinsey published The end of ERP as we know it? Five ways AI is disrupting ERP in May 2026. This is my take on their predications.

The economics are off by an order of magnitude

McKinsey states that most large enterprises spend between $100 million and $1 billion to migrate their ERP systems, with a four to five year payback.

That is not the market I work in. A Dynamics 365 investment runs between $2 million and $40 million, and I have seen payback as fast as one year. Dynamics 365 handles $80 billion organizations just fine.

AI reduced risk before it reduced timelines

McKinsey projects AI agents cutting ERP implementation effort by at least 50 percent and halving program duration. I have not seen a project get twice as fast.

What I have seen is better, and it does not show up on a Gantt chart. Risk goes down. Change orders go down. More scope fits inside each phase, so key integrations and the value attached to them stop getting pushed past go-live just to protect a date.

Three specifics from my own projects. AI has automated 100% of order entry for several clients. AI let us build all of the integrations during the project instead of deferring them. And AI has identified risk and user resistance early, which has been a game-changer for adoption, and therefore for financial payback (ROI).

Change management is still the clock

This is the one McKinsey gets exactly right, and they say it almost in passing: change management will be the major constraint in the ERP roadmaps of the future.

It is the reason implementations take as long as they do, and the reason they will not take half the time. Some tasks in an implementation exist to show value to the people already doing the work, and to give employees time to adopt the new system and get comfortable with where the business is heading.

You cannot compress that with a faster agent. People move at the speed people move.

Data models will not become “dynamic business ontologies”

McKinsey predicts data models evolving from static schemas into dynamic business ontologies that agents reason across.

Businesses do not change that fast. A data model that rewrites itself is hard to work with, expensive to enable, and expensive to accommodate in every report, integration, and control that depends on it.

Document your processes and keep your core clean. The ability to enable your business the way it works is the value here.

The best part of the article, and the part I would invert

McKinsey writes that ERP vendors went to market with AI use cases whose “scope is too small to create measurable P&L impact.” That is accurate, and I would say it about specific agents even right now.

The Dynamics 365 Supplier Communications agent is a good example of scope too narrow to move the needle. The Customer Knowledge Management agent is the opposite, delivering so much value that its cost is immaterial. Same vendor (Microsoft), same release cadence, very different outcomes. AI should be judged by the outcomes it creates, not the number of agents deployed.

Where I would push back is the conclusion. McKinsey argues value creation shifts from build to buy because “the objective function of most processes is standardization, not differentiation.”  No!

ERP lets you run business processes at scale. Agents let you shape those processes around the thing you actually compete on. For twenty years the trade was standardization in exchange for scale. This is no longer necessary and we are going back to the AX 4.0 days where you get what you want, not what you purchased :)

Partners are not going anywhere

McKinsey suggests ERP vendors could take delivery back from the partner ecosystem.  Ha!

Microsoft needs partners because of nature of the actual work. If an ERP implementation were a technology upgrade, like installing the latest version of Windows, implementation partners would be worthless.

However, these are transformation projects that change processes, jobs, and how a business operates. That is close to psychologist work, at every level of the org: C-suite, vice presidents, middle management, supervisors, and the people doing the job every day. No software vendor is automating or accomplishing that.  Period.  It requires a hands-on approach tailored to the organization.

Back to the car wash

ERP will be available headless, and that part will be real. But a system gets used the way employees want to work, not the way an architecture diagram says they should.

I can tell my agent to move a flight. I still open the app to see my trip. The screen is not there because the agent failed. It is there because I want to look and book a rideshare.

If you are waiting for AI to cut your ERP program in half before you start, you are optimizing the wrong variable. Start now, use AI to take the risk out, and spend the savings on successful change management.

What is the first process you would un-standardize, if your ERP stopped forcing you into the template?