Most B2B sales copilots get installed, used for a week, and quietly abandoned. The failure pattern is consistent: vendors build what they think salespeople need rather than what salespeople actually do. This post is the shortlist of features that salespeople actually use past week three — and the anti-patterns that explain why most products don't make the cut.
Features that stick
Call summarization and action items
The highest-ROI feature, by a wide margin. Sales calls produce context, commitments, and open questions; extracting these into CRM notes and follow-up reminders is tedious work reps routinely skip. An AI that listens to the call (or reads the transcript) and produces a summary, CRM-ready notes, and an action item list earns its keep in week one.
Critical nuance: the summary must be editable in-line and the rep must sign off. Fully automatic CRM updates breed distrust — the one time the AI mis-hears or mis-categorizes, the rep never trusts it again.
Account research and intelligence
Before a call with ACME Corp, a rep wants to know: recent news, org changes, product launches, relevant social signals, competitor moves. Historically this takes 20-40 minutes across LinkedIn, news sites, and internal notes. An AI that produces a two-paragraph brief plus a deeper dossier in 60 seconds saves hours per week per rep.
Next-step reminders with context
'Follow up with Priya about pricing by Thursday' — not just the reminder, but the context. Who is Priya, what was the last exchange, what specifically did I promise. Reps keep multiple deals live; ambient context prevents dropped balls.
Email drafting with voice matching
Draft a follow-up email matched to the rep's writing style and the specific prospect context. The key is editability — the rep reads, tweaks, sends. Not auto-send; the send button is the moment of human judgment and must stay human.
Proposal and deck configuration
Given a discovery call, generate a starting proposal with modules appropriate to the prospect's identified needs, pricing that matches their tier, and competitive positioning relevant to their stated alternatives. Reps tune from there. Saves 2-4 hours per proposal.
Anti-patterns that kill adoption
Surveillance vibes. AI that scores reps on call performance, flags 'coachable moments,' or generates manager-facing reports about individual rep behavior gets immediately resented and actively undermined. Coaching features work when the rep initiates them; they fail when management does.
Auto-sent emails. Any feature that sends external communications without explicit rep approval is a production incident waiting to happen. Reps are rightly paranoid about their reputations.
Cold outreach generation at scale. Separate category of product, separate ethics. Most reps at B2B companies don't use this and don't want to. Cold outreach tools aimed at SDRs are a different market.
Coaching nags. 'You missed asking about decision criteria on this call.' Reps know. The nag is condescending and ineffective; good managers deliver this feedback directly, and AI delivering it poorly is worse than no feedback at all.
Dashboard tax. AI that generates more dashboards for the rep to check defeats its own purpose. The AI should reduce the number of places reps have to look, not add new ones.
The integration reality
B2B sales copilots live or die by their integration with the existing stack — Salesforce, HubSpot, Outreach, Gong, Zoom. The copilot must read from these tools and write to them; any context switch breaks the value proposition. Plan 40-60% of engineering investment on integration and data sync, not on the AI itself.
Rollout pattern that works: one pilot team (5-10 reps), single highest-value workflow (call summaries usually), two weeks of tuning, then expand. Companies that roll out 'complete AI sales suites' to all reps simultaneously see the week-3 abandonment we opened with.