In May 2026, McKinsey published a report with a provocative title: "The end of ERP as we know it?" Five years ago, this would have sounded like technological theatre. Today, it demands serious attention from everyone leading an organisation of any scale.
The argument is not that ERP disappears. The argument is that its role changes so profoundly that executives who do not think about what is coming will face a fait accompli within three years.
What McKinsey says, and what it actually means
McKinsey describes five shifts that will play out over the coming years. They range from architecture to implementation speed, from vendor dynamics to the fundamental question of whether you should still build your own solutions. But there is a common thread: AI does not only change what ERP does, it changes how your organisation relates to it.
The key concept is "headless ERP". Your system of record remains, but your employees no longer interact with it directly. AI agents take over the interaction. They process orders, monitor inventory, flag anomalies, and manage escalations. Your SAP or Microsoft Dynamics becomes less a working tool and more a data vault that agents operate. That sounds technical. The organisational implications are not.
Shift 1. ERP is operated by agents, no longer by people.
The five characteristics McKinsey describes for tomorrow's ERP are: a centralised view on business value (value mission control), an agent-based operating model, processes that empower people rather than replace them, a shared conceptual framework across the enterprise (enterprise-wide business ontology), and a stable, well-maintained core (clean core).
What this means for your organisation: the employee who today approves purchase orders in an ERP screen will tomorrow do so through a conversation with an agent that provides context, filters exceptions, and offers advice. The workflow changes, the role changes, the skills you need change with it. Those who do not factor this into their HR and training strategy now will be playing catch-up.
Shift 2. Agents on top of legacy are not a structural solution.
There is a tempting path: keep your existing ERP running and layer AI agents on top. It costs less, it is faster, and you avoid a migration. McKinsey compares this to RPA, robotic process automation, which was deployed in exactly the same way a decade ago. It delivered short-term gains but did not resolve the underlying data and process problems.
Agents on a fragile foundation amplify the fragility. If your data is wrong, your agent gives the wrong advice. If your processes are inconsistent, automation accelerates the inconsistency. ERP modernisation remains necessary, even in an agentic era. The difference is that you now have a stronger business case to fund that modernisation.
Consider a distributor or manufacturer. An AI agent can read incoming orders, retrieve product codes and customer data, propose a sales order, and enter it into the system after human confirmation. It sounds straightforward. But it only works if all customer codes are consistent, if special pricing agreements are documented somewhere rather than living in the head of the inside sales team, and if approval rules are explicitly defined. That tacit knowledge — the experienced employee who "knows" that customer X always gets a 10% discount — must be written down first. Not for the AI, but for the organisation itself. A useful rule of thumb: if you cannot explain a process to someone who starts tomorrow, it is not ready for automation.
A second warning sign: the planner who maintains an Excel layer because the ERP does not reflect reality accurately enough. That spreadsheet is not a solution, it is a symptom. An AI agent working from that data will reproduce the problem at scale.
Shift 3. Implementations become twice as fast and significantly cheaper.
This is perhaps the most concrete figure in the report: AI reduces implementation effort by at least 50%, and timelines are halved. Tasks such as configuration, testing, and training, which today consume the bulk of an ERP project, can largely be automated. Testing alone sees an 80% effort reduction, training 90%. An agentic design-and-build trajectory that previously took 6 to 9 months compresses to 2 to 3 months.
But McKinsey adds a nuance you cannot afford to miss: change management becomes the bottleneck, no longer the technology. The figures on failed transformations do not flatter technology projects: only 25 to 35% of large technology programmes achieve their intended EBITDA impact, and 65 to 80% exceed budget or timeline.
The reason is rarely the software. The reason is that the organisation is not ready. Processes are insufficiently thought through, employees have not been brought along, ownership is distributed across too many parties. Faster technology does not resolve this. It amplifies the impact of those who get it right.
Shift 4. ERP vendors are reclaiming the playing field, but competition is coming from a new direction.
SAP, Oracle, Microsoft, and the other major players are actively building end-to-end agentic deployment solutions. They are integrating promising AI startups and working to control the full stack. That is not necessarily bad news if you are already in their ecosystem, but it makes vendor lock-in a more tangible risk.
Competition for established ERP players no longer comes only from each other. It comes from cross-platform agentic providers that operate across existing ERP boundaries. That increases your options as a customer, but it also makes the market more complex. Those who make architecture decisions based on vendor roadmaps rather than their own strategic needs will sooner or later find themselves in an uncomfortable position.
Shift 5. Build or buy: the boundary is shifting.
For processes that are genuinely differentiating for your organisation, your unique way of working with customers or the specific logic of your supply chain, custom development remains defensible. But for standardised ERP processes, finance, procurement, logistics, HR, the expectation is that vendors will deliver embedded AI solutions that outperform what you could build yourself.
McKinsey expects EBIT improvements of 5% or more for early adopters who implement vendor solutions correctly and adapt them to their context. Not by building, but by buying smartly and integrating well. The question is no longer "do we build this ourselves?" but "which part of this process is truly differentiating for us, and which part is purely operational?"
What this means for your organisation now
Three insights for those leading an organisation of 50 to 500 knowledge workers.
Do not wait for the perfect moment to modernise. The figures on failed transformations work in reverse as well: those who invest early in a stable ERP core and clean data have a far stronger starting position when agentic AI becomes available for their systems. Those who wait until everyone is doing it pay a higher price for a smaller advantage.
Redefine what success looks like in a transformation project. If change management is the new bottleneck, then the measure of a successful ERP modernisation is not "did the software go live?" but "are our people working with it effectively, do they take ownership, and do they understand what the agents are doing for them?" That requires a different approach to project leadership, communication, and training. Technology adoption is a human challenge with a technological answer, not the other way around.
Use this momentum to claim ownership. The shift toward agentic ERP is an opportunity to define, as an organisation, which processes are truly strategic and which are purely operational. That exercise has value regardless of which technology you choose. It forces you to think about what distinguishes your organisation. And that is precisely the kind of question a vendor cannot answer for you.
The broader lesson
McKinsey closes with an observation that every organisational leader should carry with them: the technology is there. Implementation speed is increasing. Costs are falling. What will make the difference is not who implements fastest, but who has the organisational capacity to absorb, steer, and own the change.
That is a question of leadership, not software.
Based on McKinsey, "The end of ERP as we know it? Five ways AI is disrupting ERP", May 2026.