The Agentic Revenue CompanyRevenue as a Service
Blog

Next best action: how a playbook determines the next step in sales

Next best action in sales: how an agent uses playbook, contact outcome and customer class to propose the next step – and a person decides.

Your sales rep hangs up the phone. The customer said: “Get back to me after the end of the quarter, we will know more then.” Now the real work begins. When exactly is the right moment? Is there an open case with another contact at the same company right now? Is an email enough, or does it need a call? The question of the next best action comes up again after every customer contact. In many sales teams, it is answered from memory and gut feeling.

Why the next step decides revenue

Repeat business and customer reactivation rarely happen in the first contact. They come from a chain of steps that reliably follow one another. If the chain breaks, the opportunity is left lying. Usually not because someone made the wrong decision, but because nobody made one.

Gut feeling has two weaknesses here. It cannot be traced, and it is not consistent: two people react differently to the same customer statement. One calls back too early, the other not at all. What matters to your customers, though, is that the outreach fits. According to a 2025 Gartner survey, 73% of the B2B buyers surveyed actively avoid suppliers whose outreach is irrelevant to them. The wrong next step is exactly that: outreach that does not fit the situation.

The playbook as a shared foundation

A playbook describes how your company proceeds in typical situations: which contact follows which, how much time lies between them, when a case is closed and when a call achieves more than an email. It is not a script for individual sentences in a conversation – that is what the sales call script is for. It is the rulebook for the sequence.

For an agent to work with a playbook, it needs two properties. First, it is approved: what is in the playbook was decided by your functional leads, not by a model. Second, it is versioned: when a rule changes, a new version is created, and it remains traceable which version a proposal was based on.

Which rules belong in your playbook is not a technical question. It depends on which customer relationships are worth how much to you and how you want to address your customers. That is exactly what we clarify with you before an agent makes its first proposal.

Next best action from three inputs

The agent builds its proposal from three sources:

  • The playbook defines which steps exist at all and in which order they are intended.
  • The contact outcome describes what happened in the last contact. A fixed catalogue of outcomes in the contact log ensures that “customer not reached”, “callback requested” or “no need at the moment” always trigger the same consequences.
  • The customer class determines how intensively a customer is looked after, for example how often contact is intended. How such classes are built on rules rather than gut feeling is the subject of the article on customer ABC analysis.

The result is a proposal with a reason: which step is due next and why. The reason is not decoration. It makes the proposal verifiable, and it shows your rep which rule and which outcome it is based on.

A good agent does not invent steps

A language model can phrase things very convincingly. In sales, that is both a strength and a risk. That is why we apply a clear rule: the agent only proposes steps that the playbook provides for. If the playbook provides no step for a situation, the agent does not invent a plausible alternative either.

This separation runs through the whole model. The fixed consequences of a contact outcome are controlled by the outcome catalogue, not by how eloquent a model is. The model contributes its strength where it is needed: summarising the situation from cases and contact history, and explaining the proposal clearly.

A person decides – with less legwork

In the language of our human-in-the-loop model, the next best action is an action of the type recommend: the agent proposes, a person decides. That does not slow things down, it takes load off. Your rep no longer has to reconstruct every situation on their own. They review a reasoned proposal and decide, instead of first piecing the situation together from cases, emails and notes. Where a person approves and where not is covered in more detail in human-in-the-loop in sales.

This is how agentic sales – AI agents in sales – becomes concrete in everyday work: agents organise cases and plan next steps, people decide where judgement matters. This is particularly needed in customer reactivation, where many contacts have to be followed through reliably over a longer period. And because we run the system on an ongoing basis, we keep developing the playbook with you as it becomes clear which steps work.

Your next step

Ask yourself first: does your sales team have a written, shared answer to the question of what happens after a contact? Or does every person carry their own in their head? If you want to find out whether a playbook with AI agents can secure and grow your revenue, check your REVAGO potential.


← All articles