Forrester, Salesforce, HubSpot — they've all run the same study, and they all find roughly the same number. The B2B salesperson spends 20-25% of their week actually selling. The rest goes to admin, research, comms, internal coordination, and various species of busywork.
If you're a VP of revenue, that means roughly 75% of your sales spend isn't producing pipeline. AI doesn't fix all of it. But the right rollout can return 12-18 hours a week per rep, and that math is hard to argue with.
Where the time goes
From the studies and our own audits, here's the typical breakdown for an account executive:
- 23% selling. Live conversations with prospects.
- 16% CRM and admin. Updating records, logging calls, building reports.
- 14% prospecting and research. Sourcing leads, reading company news, building lists.
- 13% writing emails. Outreach, follow-up, scheduling, internal updates.
- 11% internal meetings. Pipeline review, deal review, forecast.
- 10% proposal/SOW work. Drafting, pricing, getting approvals.
- 13% other. Training, slack, the buffer.
Three of those buckets are AI-yieldable in a big way: CRM admin, research, and email. Two more are partially yieldable: proposals and internal meetings. That's potentially 45-55% of the week.
1. CRM admin (8-12 hrs/week back)
The pattern that works: rep talks to a prospect, dictates a 90-second voice note, AI structures it into CRM updates — only the fields that matter, in the format your team uses. The rep reviews and approves in 30 seconds.
What stops working: AI that updates the CRM without the rep's review. Field drift, made-up next steps, fictional commitments. Always keep the rep in the approval loop.
2. Research and prospecting (4-6 hrs/week back)
The pattern that works: rep gets a daily brief on their top 20 accounts — news, funding events, exec changes, hiring signals — with a suggested play for each. Five minutes of reading replaces an hour of clicking around.
For outbound: the AI drafts a research brief on each prospect with three angles. The rep picks the angle and writes the note. The note is more relevant than the alternative.
3. Email (3-5 hrs/week back)
The pattern that works: the rep's drafts get auto-completed in their voice based on prior correspondence. They proofread, send. The system learns from edits.
What stops working: AI that sends without review. There's a 1% case every week where the AI says something subtly wrong. That 1% costs more than the 99% saves.
4. Proposals and SOWs (2-4 hrs/week back)
The pattern that works: structured brief + pricing inputs in, draft out. The drafting was never the slow part — pricing review and partner approval were. Compress those by making them reviewable in minutes.
The goal of AI in sales isn't to replace selling. It's to make selling the largest slice of the week instead of one of the smaller ones.
What doesn't work
- AI SDRs that auto-send. Volume goes up, reply rates collapse, brand takes a hit.
- Forecast AI. The data quality in most CRMs is too poor for forecasting AI to be better than your VP's gut. Fix the data first.
- "AI coaching" from call recordings. Sometimes useful, often noisy. Pilot before scaling.
The order of operations
- Voice-note → CRM is the gateway tool. Rep buy-in is immediate because it removes the most-hated task.
- Daily account briefs are second. Easy, visible, replicable.
- Email assist comes third. Personalize by rep.
- Proposal automation when proposal volume is high enough to justify it.
Ship in that order, you'll get the team's trust. Ship in reverse, you'll get a pilot that nobody uses.