A useful sales manager playbook creates one source of truth, one visible owner, and one dated next action for every active opportunity. It defines how deals enter and leave each stage, how to follow up without surprising customers, how to transfer a conversation with context, and which decisions require human review. The manager keeps this system reliable, removes blockers, and develops the team; they should not update the CRM for everyone or demand activity without a purpose.
Create a one-page operating contract
The playbook should answer the same questions for everyone: what qualifies as an opportunity, who can take it, how long it can remain untouched, which data is essential, how ownership transfers, and when to nurture, win, lose, or archive. A process that lives in memory or private messages cannot be inspected or reliably resumed.
Choose one official system for conversations, stage, owner, value when relevant, customer need, next steps, and win or loss reasons. A parallel spreadsheet can support a temporary analysis but should not compete with the operating record. Apply the principle of one visible owner per conversation, and keep an unassigned queue with defined coverage. This prevents duplicate answers and makes the next actor clear.
- Company or person and an authorized contact channel
- Problem or desired outcome in the customer’s own words
- Current stage and the evidence supporting it
- Primary owner and any help required
- Concrete next action, date, and responsible party
- Value and timing only when grounded in facts
- Known risk or blocker
- Source and reason when won, lost, or moved to nurture
Collect data with purpose and necessity in mind. Brazil’s LGPD, Article 6, requires legitimate, specific, disclosed purposes and limits processing to what is necessary. Even outside Brazil, data minimization is a sound operating rule: do not request identifiers or sensitive details merely to make a record look complete.
Define stages through evidence, not optimism
Salesforce Trailhead describes a pipeline as a representation of where prospects are in the sales process, and distinguishes the pipeline—which guides present work—from a forecast, which estimates future results. Stages vary by industry, product, and buying cycle, so adapt the criteria below instead of simply renaming them.
| Stage | Verifiable entry | Expected exit |
|---|---|---|
| New | Inbound contact or authorized prospect not yet triaged | Owner assigned and first approach recorded |
| Discovery | Conversation started about context, need, and impact | Problem, participants, and next step confirmed |
| Qualified | Real need, fit, and a plausible path to a decision | Solution criteria and buying process understood |
| Solution or proposal | Scope discussed; proposal is not an automatic stage change | Questions, participants, timing, and decision action recorded |
| Decision | Customer is evaluating a concrete option under known conditions | Won, lost, or nurtured with a verifiable reason |
| Closed | Outcome confirmed, never inferred from silence | Complete handoff or closure reason documented |
Do not advance an opportunity because a meeting happened, a proposal was sent, or a deal “feels hot.” Record what was observed: who confirmed the need, which commitment was made, and what event comes next. Do not keep dead deals open to protect a pipeline number. Salesforce’s guidance on a healthy pipeline emphasizes flow, qualification, and regular reviews using consistent data.
Manage exceptions every day, not every conversation
At the start of the day, inspect unassigned items, new opportunities without a first action, overdue activities, deals beyond the stage age limit, and high-priority conversations. Midday, resolve blockers such as pricing approval, specialist involvement, unexpected absence, or a waiting customer. Before closing, confirm that every active deal has a real next action and that no promise exists only in a private chat.
| Moment | Manager check | Outcome |
|---|---|---|
| Start of day | New, unassigned, overdue, and missing next actions | Priorities and owners are clear |
| Midday | Blockers, handoffs, and replies needing authority | Decisions and support are supplied |
| End of day | Next actions, dates, risks, and waiting conversations | Pipeline can be resumed tomorrow |
A follow-up should add value: confirm a commitment, answer a question, provide requested material, or check whether priorities changed. Avoid endless “just checking in” sequences. The current WhatsApp Business Messaging Policy requires a supplied number and opt-in, respect for stop requests, and approved templates outside the applicable customer-service window on the platform. Policies can change, so verify the current version when configuring automation. See the guide to WhatsApp opt-in and opt-out.
Run a weekly review that produces decisions
A pipeline meeting is not a group reading of every card. Before it starts, each seller updates stage, next action, date, evidence-based value, and risk. During the review, the manager selects exceptions: high-impact deals, stalled items, near-term forecasts, multi-stakeholder decisions, and opportunities needing help. Each discussion must result in a decision, support action, or corrected record.
- Movement: what changed, and what evidence came from the customer?
- Next event: who will do what by when, and does the customer recognize that step?
- Risk: which assumption remains untested, who is absent, or what dependency blocks progress?
- Help: should the manager coach, approve, join, remove a barrier, or let the owner execute?
- Forecast: does the category reflect current facts or hope? Record the rule used.
Use AI to prepare work, not make commitments
AI can summarize conversations, identify intent, draft a response from approved knowledge, flag a missing field, or list opportunities without a next action. These uses reduce mechanical work but still require review proportional to risk. The NIST AI RMF recommends defined human-AI roles, documented knowledge limits, and human oversight processes.
Require human approval for price, discount, contract, guarantee, promised timeline, policy exception, sensitive-data collection, disqualification, and any reply that may create an obligation. Automation should also provide a clear route to a person, consistent with WhatsApp policy. When risk warrants it, retain what was suggested, what was sent, and who approved it.
- It may summarize facts; it may not invent need or commitment
- It may suggest a next step; it may not advance a stage without evidence
- It may prepare text; it may not promise price, time, or results
- It may flag risk; a person decides and records the action
- It may use approved knowledge; unsourced uncertainty goes to review
- It must allow correction, audit, and transfer to a person
Measure flow, coverage, and record quality
Do not rely on one conversion rate to explain the operation. Define events and denominators before comparing periods, sources, or people. A commercial first-response metric, for example, needs a clear starting event and exclusions; speed alone does not prove quality.
| Metric | Useful definition | Management question |
|---|---|---|
| Next-action coverage | Active opportunities with valid action and date ÷ active opportunities | Is the pipeline executable? |
| Stage age | Time since the last evidence-based entry into a stage | Where has flow stopped? |
| Contact rate | Leads with a valid human interaction ÷ eligible leads | Can the team start dialogue? |
| Stage conversion | Qualified exits ÷ entries from the same cohort | Which transition loses demand? |
| Follow-up completion | Actions completed on time ÷ planned actions | Are commitments kept? |
| Closure-reason coverage | Wins and losses with valid reason ÷ closed deals | Is there evidence to learn from? |
Read absolute volume, source, product, and sales cycle alongside each rate. Avoid rankings built on small samples or targets that reward artificial stage movement. A metric should point to an investigation and help improve the system, not create a contest to manipulate the record.
Roll out the playbook in 30 days
- Days 1–7 — map: trace five to ten current opportunities, finding parallel systems, delays, handoffs, and ownerless decisions. Select the official source and write the one-page contract.
- Days 8–14 — configure: create evidence-based stages, minimum fields, the unassigned queue, coverage rules, and closure reasons. Migrate only necessary data and preserve relevant history.
- Days 15–21 — pilot: run it with a small team. Complete the daily routine and one weekly review. Fix ambiguous definitions, noisy alerts, and fields that do not support decisions.
- Days 22–30 — expand: train with real cases, publish AI boundaries, define metrics and maintenance owners, and record changes so comparisons remain understandable.
- Every active opportunity has one visible owner
- Every deal has a concrete next action and date
- The stage corresponds to customer evidence
- The official history contains decisions and promises
- Handoffs carry context, open issues, and expectations
- Follow-ups respect permission and the option to stop
- AI prepares; authorized people make commitments
- Reviews end with a decision, support action, or correction