Next best action (NBA) is one recommended step for a specific deal or account, chosen from a small fixed set, grounded in evidence from the CRM, and accompanied by the reason it was chosen. A good NBA system also knows when not to act, and treats "no action" as a valid answer.
Dashboards tell sellers what happened. Ranked lists tell them where to look. Next best action closes the last gap: given this deal, right now, what is the one thing to do? This guide explains the idea, the design choices that make it trustworthy, and examples you can adopt even without software.
Next best action, defined
Next best action is a single recommended step for a specific account or deal. "Single" matters: a list of ten possible actions is just another dashboard. "Specific" matters too: the recommendation should name the contact, reference the situation, and explain why this step and not another.
The concept comes from customer-experience and marketing decisioning, where systems choose the next offer for a customer. In B2B sales the same idea applies to deals: follow up, re-engage, bring in a stakeholder, update the record, or wait.
Why a fixed set of actions works better than free text
It is tempting to let a system generate any advice it likes. In practice, a closed taxonomy of actions is easier to trust, measure and improve:
- Follow up when a buyer re-engaged or engagement spiked.
- Follow up after a meeting when a meeting happened and nothing followed.
- Re-engage a stalled deal when the deal has sat in stage too long or regressed.
- Contact an additional stakeholder when a priority deal is single-threaded.
- Update the deal when the close date has passed and the record is wrong.
- Review the deal when the data is too stale to justify outreach.
- Create a task when action is needed but no contact is associated to act on.
- No action when nothing warrants it, or a seller acted very recently.
Because the set is fixed, you can count how often each action is recommended, collect seller feedback per action type, and see where the rules are wrong.
Rules, order and contraindications
A trustworthy NBA system is a set of eligibility rules checked in a fixed order, where the first eligible rule wins. Each rule also has contraindications: conditions under which it must not fire even if its trigger is present.
- Quality guards come first. Closed deals get no action. Stale data allows an internal review, never outreach. If the evidence is too thin to score the deal, the answer is no action, with the reason stated.
- Recent seller activity is a contraindication. If the seller emailed yesterday, telling them to follow up again erodes trust in every future recommendation.
- No contact means no outreach. If no buying contact is associated, the right action is to create a task or find a stakeholder, not to "email the buyer".
- Do not invent channels. If phone numbers are not in the CRM, do not recommend a call.
Every recommendation needs evidence
The difference between a recommendation and an opinion is evidence. Each NBA should cite the records that triggered it, such as the meeting, the reply, the stage change or the close date, so the seller can verify it in seconds. A recommendation whose evidence cannot be traced to a real record should be dropped, not shown. This is the same principle as grounding in AI systems.
Examples
- Situation: demo three days ago, no follow-up logged. NBA: send a follow-up to the contact who attended, today. Why: a meeting with no follow-up past the follow-up window.
- Situation: buyer replied after a long quiet period. NBA: follow up with that buyer today. Why: re-engagement after a gap.
- Situation: proposal-stage deal well past the normal time in stage, no recent seller activity. NBA: re-engage the main contact. Why: stalled beyond the stage threshold.
- Situation: close date passed last week. NBA: update the close date. Why: the record no longer reflects reality and the forecast depends on it.
- Situation: seller emailed yesterday after a meeting. NBA: no action. Why: a seller acted recently; give it time.
Next best action and AI
Generative AI is good at drafting the message once the action is chosen. Choosing the action is a different job, and one where reproducibility and explanation matter more than creativity. Many teams therefore separate the two: deterministic rules decide what to do and why, and a person decides whether and how to do it. That keeps a human in the loop and avoids the brand risk of fully autonomous outreach.
Pipeit follows this design. It selects one next best action per deal from a fixed set using ordered rules and contraindications, attaches the evidence and the reason, and never sends anything itself.
Frequently asked questions
What does next best action mean in sales?
Next best action is a single recommended step for a specific deal or account, such as following up, re-engaging a stalled deal, contacting another stakeholder, updating the deal or taking no action, chosen from evidence in the CRM and explained.
What is the difference between next best action and next best offer?
Next best offer chooses which product or promotion to present to a customer, mostly in marketing and retail. Next best action is broader and, in B2B sales, usually means the next step a seller should take on a deal.
Should next best action be generated by AI?
Choosing the action benefits from deterministic, explainable rules so the same situation always gets the same recommendation. Generative AI can help draft the resulting message, with a person approving what is sent.
Can no action be a next best action?
Yes. If a seller acted very recently, the data is stale, or nothing warrants attention, recommending no action protects trust in every other recommendation.
For definitions of the terms used here, see the B2B sales and revenue glossary.