Move to agentic resource planning: what this means
The most common misunderstanding about agentic resource planning, often called agentic ERP, is that it needs a large transformation programme. It does not. The agents run on top of the ERP you already have, so the move to agentic resource planning starts without touching the ERP and widens as trust builds.
1. Start with a read-only connection
In the first step the agents connect to the ERP with read-only access. Logo, Netsis, Mikro, Canias, Nebim, SAP, Microsoft Dynamics 365 or Oracle: whichever system you run is read through a ready connector or directly from the database. Nothing is written at this stage.
The point of this step is to get to know the data: which fields are filled, which codes are inconsistent, which tables are current. An agent’s recommendation is only as good as the data it reads.
2. Pick one decision
Rather than handing all of resource planning to agents at once, pick a single decision. A good first decision is:
– Frequent (daily or weekly).
– Reversible (a wrong transfer can be corrected; a wrong dismissal cannot).
– Measurable (did stock-outs fall, did excess stock clear).
Inter-store stock transfers, safety-stock updates and collection priorities fit this description.
3. Run it in the shadow first
For a period the agent produces recommendations that are not applied; they are set beside the team’s own decisions. Two questions are asked: does the agent see the problems the team would see? Where it does, is its recommendation sensible?
Because every recommendation shows its reasoning and the data behind it, the team can discuss concretely why it would decide differently.
4. Move to approved write-back
Once trust is there, recommendations start arriving in an approval box. The responsible person sees each one with its reasoning and approves or rejects it. An approved recommendation is written to the ERP: a transfer order, a purchase request, an updated parameter. The reason for a rejection is recorded and taken into account next time.
The limits stay clear at this stage too: the agent can write only the operations it is allowed to, every write depends on an approval, and every write is logged.
5. Build workflows that start from signals
Once the first decision is settled, the loop runs on its own: the agent reads the data regularly, raises a signal when a threshold is crossed, the signal starts a workflow and the right team is notified in Teams, Slack, Telegram or WhatsApp. Nobody waits for a report; the problem and its likely fix arrive together.
6. Widen the scope
Each new decision area goes through the same path: read-only, shadow, approved write-back. From inventory to capacity, from capacity to cash, from cash to people.
Common mistakes
– Choosing the hardest decision first. The first decision has to succeed for the team to trust the next.
– Skipping the shadow period. Opening the approval box before measuring what the agent gets right and wrong loses trust at the first mistake.
– Leaving data problems to the agent. An agent does not fix inconsistent units or duplicate product codes; it makes them visible. Fixing them is still the team’s job.
Summary
Moving to agentic resource planning is not an ERP replacement; it is a decision layer built on top of the ERP one step at a time. It starts read-only, proves itself on one decision, moves to approved write-back and widens as trust grows.