Recognising the purchase cycle: how to make repeat business plannable
How an AI agent derives each customer’s purchase cycle from their order history and prompts your sales team when renewed demand can be expected.
For years, a regular customer has ordered from you at fairly even intervals. Then the next order fails to arrive. Nobody notices: sales is busy with urgent requests, and the customer has not complained. When someone follows up months later, they have long since ordered from another supplier – even though their purchase cycle could have flagged the gap.
Gaps like this are not the result of carelessness. That cycle is right there in your order history – but nobody has the time to keep track of it for every single customer. Repeat business becomes plannable as soon as it is calculated rather than guessed.
Why repeat business depends on timing
Recurring demand is a reliable source of revenue in B2B: for reorders and consumables just as much as for services that customers commission again at intervals. It does not need a subscription. It needs the right moment.
Reach out too early and the contact feels like advertising. Reach out too late and the customer has already decided. Yet as a known supplier, you have a head start: according to McKinsey, 53% of the B2B decision-makers surveyed are “adapters”, relationship-oriented buyers who often stay with suppliers they know (McKinsey, The surprising economics of B2B growth, May 2026). You only benefit from that head start if you get in touch while the need is still open.
What makes up a customer’s purchase cycle
The purchase cycle is the typical interval between two orders from a customer. It is individual: one customer orders at short, regular intervals, another rarely but in larger volumes. A single reminder interval for all customers therefore almost never fits.
For a calculated cycle to hold up, three things matter:
- Only real orders count. A quote shows interest, but not a purchase. Only actual orders feed into the cycle.
- The reasoning is traceable. Whoever sees a customer flagged as due should be able to see why – in other words, which past orders the interval is based on.
- Gaps are handled honestly. If a customer has placed only a few orders, no individual cycle can be derived. A fallback rule for their customer class applies instead. How such classes come about is covered in the post on customer ABC analysis.
How an agent calculates the cycle and turns it into cases
An AI agent derives the purchase cycle from the order history in the connected ERP system. Before that becomes a due date, it checks the state of the data. If the data is incomplete, no due date is created. Better no prompt than a wrong one. Which data needs to come from the ERP system for this is shown in the post on ERP integration for AI agents.
The cycle becomes a daily routine:
- The agent works through the cases that are due and queues the first contact. It is only sent after the central checks of rules, channel and address.
- When the system picks up a new order from the ERP system – which happens at least daily – it affects the open cases. That way, your customer is not reminded about an order they have just placed.
- Once a case is closed, the agent plans the follow-up case from the cycle. A one-off action becomes a loop.
- The run only works in the territory you have approved. When you open up a new territory, it respects the customers’ cycles there instead of contacting everyone at once.
Where people decide
A purchase cycle is not a forecast. No agent knows when your customer will actually order again. It recognises when renewed demand is likely based on past behaviour, and turns that into a well-founded reason to get in touch.
People decide what that contact looks like. The first contact is built from editorially approved text modules and only mentions past orders that are on record. Your customer replies with one click on one of three fixed answers; if there is no response, a reminder follows, then a call task for your team. How such an email is structured is described in the post on the reactivation email.
Only this approved first contact runs without individual approval. Individual replies only go out once a person has approved them. Where that line runs, and why, is the subject of human-in-the-loop in sales.
From purchase cycle to accountable repeat business
The purchase cycle is a building block, not a finished revenue system. Which customers qualify, which territory you start with, what “due” means for your business and which texts may go out are things you set out in your mandate. We clarify these questions in AI-first coaching, before any agent starts work.
Then we implement: we build your agents, connect them to the ERP system covered by your mandate and run the system on an ongoing basis. For us, agentic sales – AI agents in sales – means agents handle the routine work reliably while your team has the conversations that matter. And because we are accountable for operations: you pay for results – not for activity. You will find more on the use case under repeat business and on the underlying capabilities under organise cases and plan next steps.
Your next step
Would you know today which of your regular customers will need something again in the coming weeks? If the answer is in your order history but never reaches a workflow, that is a good starting point for a conversation. Check your REVAGO potential – in five short questions.
