AI helps a Salesforce sales team most with the reading and writing around selling: researching an account before a meeting, summarizing calls and email threads, suggesting the next step on a deal, flagging stale opportunities and preparing forecast reviews. It helps least when opportunity, contact and activity data is thin or wrong, because every suggestion is drawn from those records. Fix the data a use case depends on, start with one workflow reps already find tedious, and keep people in charge of what goes to customers.
Five jobs AI does well for sellers
Salesforce now presents AI across the sales cycle, from prospecting and account research to call insights, record summaries, deal insights and predictive scoring. Packaging and names shift from release to release, so judge each feature by the job it does rather than its label. Five jobs cover most of the value we see in mid-market sales teams:
| Job | What the rep or manager gets | Records it draws on | Who checks the output |
|---|---|---|---|
| Account research and meeting briefs | A one-page view of the account, open deals, recent contact and likely talking points | Account, contacts, opportunities, cases, activity history | The rep, before the meeting |
| Call and email summaries | Key points, objections and agreed next steps from a conversation or thread | Call recordings or transcripts, email activity | The rep, before saving or sharing |
| Next-best-action suggestions | A recommended step on a deal or account, with the reason behind it | Opportunity stage, activity recency, contact roles, playbooks | The rep decides whether to act |
| Forecasting support | Deals whose score, close date or amount looks at odds with their history | Opportunity history, scores, forecast categories | The sales manager in the forecast call |
| Pipeline hygiene | Lists of deals with past-due close dates, missing fields or no recent activity | Opportunity fields and activity dates | Sales operations, weekly |
Notice that none of these replaces a seller's judgment. They shorten the time between a rep sitting down and a rep knowing what to do, which is where most CRM time goes today.
Account research, briefs and summaries
Meeting preparation is the easiest win to explain to a sales team. A rep who used to click through the account, three opportunities, a case list and an email folder gets a short brief instead, and spends the saved minutes thinking about the conversation. The brief is only as current as the records behind it, so a team that logs meetings inconsistently will get briefs that miss the last two touchpoints.
Call summaries follow the same logic. Salesforce's Einstein Conversation Insights can generate summaries of recorded voice and video calls, including next steps and customer feedback, although the generative features depend on your edition and add-ons, so confirm what your contract includes. Whatever tool produces them, agree where summaries live: on the call record, the opportunity, or both. A summary that sits outside Salesforce cannot feed the next brief.
Our own work with an affiliated B2B marketing company (the two firms share ownership) shows how far this can go once the basics are in place. Claude-based call summaries and highlights were surfaced in appointment briefing emails and the account-manager queue, and call-quality scorecards with grades, supporting transcript excerpts and coaching notes were written back to Salesforce after being calibrated against human-scored calls. That calibration step is what made managers willing to coach from the scores.
Next-best-action, forecasting and pipeline hygiene
Next-best-action works when the recommended steps are ones your sales leaders already agree on. Write the plays down first: what a rep should do when a deal has no economic buyer attached, when a proposal has gone unanswered for two weeks, or when a renewal is approaching with open support cases. An AI suggestion that points to a known play gets used; one that invents a new approach every time gets ignored.
For forecasting, treat AI as a second opinion rather than the forecast itself. Pipeline Inspection in Sales Cloud can show opportunity scores as tiers, with arrows when a deal moves up or down a tier, and Einstein Deal Insights adds predictions and recommended actions about deal health. The useful question in a forecast call becomes why a deal the rep calls committed has slipped a tier, not whether the model is right.
Pipeline hygiene is the least glamorous use and often the one that pays back first. A scheduled review that lists past-due close dates, deals stuck in a stage and opportunities with no logged activity gives managers a short agenda for one-to-ones. It also improves the data every other AI feature depends on.
The data prerequisites
Sales AI reads the same fields your reports do, so the same gaps hurt it. Before switching anything on, check the records behind the use case you picked:
- Opportunity stages have written exit criteria, so a stage means the same thing for every rep.
- Close dates and amounts are updated when deals move, not only at quarter end.
- Contacts are linked to opportunities with roles, so briefs and suggestions know who the buyer is.
- Activity is logged where Salesforce can use it; check where captured email and meetings are stored, since some capture settings keep them out of standard activity records that reports and automation read.
- Closed-won and closed-lost reasons are required and reported on, because predictive scores learn from how past deals ended.
- Duplicate accounts and contacts are merged, or a brief will show half the relationship.
You do not need a perfect org, only a reliable slice. If you are rolling out call summaries, the call and opportunity data matters; account hierarchy cleanup can wait for a later phase.
Getting reps to use it
Sales teams adopt tools that save them time on work they dislike, and they drop tools that add clicks. Pick the first use case with your top reps, not for them: ask which task eats their Monday mornings and start there. Put the output where they already work, on the opportunity page or in the daily email they read, rather than in a new tab.
Managers set the tone. If the forecast call starts with the AI-flagged deals, reps learn that keeping records current keeps them out of the spotlight. If managers never mention the tool, reps conclude it is optional. Measure a small set of signals from the start, such as how many briefs were opened before meetings or how many suggested next steps were accepted, and review them with the team after the first month. Pull a feature that is not used rather than leaving it to clutter the page.
Finally, be clear about what the AI can see. Reps trust summaries more when they know the tool respects the same record access they have, and leaders sleep better when that is written into the rollout plan. Our AI security guide covers those controls in detail.
