I used to walk into sales calls with a scrawled note on the back of an envelope. Company name, one line about what they did, and whatever I half remembered from the last email thread. Sometimes I'd forget which pitch I was even meant to be giving until the prospect said something and it jogged my memory.
That was fine when I had two calls a week. It stopped being fine when sales calls, SalesM8 demos, and consulting calls all landed on the same calendar, some weeks eight or nine a day. So winging it stops working the moment volume goes up. You either build a system or you start losing deals to your own scattered notes.
Here's the bit that should worry more business owners than it does. The average B2B rep spends 38 minutes preparing for a single discovery call, and only 23% say they walk in feeling truly prepared. That's not a skills problem. That's a system problem. Thirty-eight minutes of LinkedIn stalking and re-reading old emails, and reps still walk in feeling like they're sort of guessing.
I don't spend 38 minutes anymore. I spend about ninety seconds. Here's what actually happens before, during, and after every call I take.
Before the call: the briefing writes itself
Every call on my calendar triggers a briefing that lands as a note attached to the invite, roughly twenty minutes before I'm due to dial in. It pulls from three places: the CRM record (every past touchpoint, tags, deal stage), any transcript from a prior call with the same person, and whatever public info exists on the company (site, recent posts, size).
The output is short on purpose. Three things: what this person cares about based on what they've actually said before, what objection came up last time and whether it got answered, and the one specific thing I should ask about that I'd otherwise forget. Not a dossier. A cheat sheet.
This isn't some bespoke thing I built from scratch either. Salesforce's 2026 State of Sales report, a survey of more than 4,000 sales professionals, found AI agents are expected to cut prospect research time by 34% once fully rolled out, and top-performing sellers are 1.7 times more likely to be using AI agents than the ones missing quota. That gap isn't about who's smarter. It's about who stopped doing manual research the hard way.
During the call: the thing that actually moves the needle
And here's the bit most people building "AI sales tools" get wrong. Everyone wants a tool that tells them what to say. The data says the bigger lever is what you stop saying.
Nimitai's analysis of 350 B2B sales calls found the median talk-to-listen ratio on closed-won calls was 41%, meaning the rep talked 41% of the time and the buyer talked the rest. On closed-lost calls, the median rep talk ratio was 58%. An 8 to 14 point drop in talk ratio correlated with a 6 to 11 point lift in close rate. Median win rate across the dataset sat around 21%, which lines up with what most B2B SaaS teams report.
That's the whole game a lot of the time. Not a better script. Less talking. My setup flags it live, a quiet nudge if I've been talking for more than about ninety seconds straight, because the natural instinct on a demo is to keep explaining features nobody asked about yet. The AI isn't clever here. It's just counting, and counting turns out to matter more than cleverness.
I built the guts of this thinking into SalesM8, the AI sales coaching product I built, precisely because the talk ratio problem is universal. Every rep thinks they're a good listener. Most aren't, and most don't find out until someone shows them the number.
After the call: five minutes, not forty
The call ends, and this is where most systems fall over because everyone builds the "before" part and forgets the "after" part is where deals actually die. Follow-up drafts itself from the transcript within a couple of minutes: what was agreed, what objection needs addressing in writing, the next step with a date attached, not "let's circle back."
CRM gets updated automatically. Deal stage moves if it should. If there was a specific commitment made ("I'll send you the case study by Friday"), that becomes a task with a deadline, not a mental note that evaporates by lunchtime.
None of this is exotic technology. 81% of sales professionals now use AI tools at least occasionally, up from 54% in 2024, so the tools aren't the differentiator anymore. What separates people getting real value from people who bought a subscription and forgot about it is whether the before, during, and after pieces are actually connected. Most businesses have three separate tools that don't talk to each other. That's not a system, that's a junk drawer with a login screen.
Where I still do the work myself
I want to be straight about what doesn't get automated, because there's a version of this post that oversells it. The AI doesn't decide what I say when someone pushes back hard on price. It doesn't build rapport. It doesn't read the room when someone goes quiet because they're worried about how this looks to their business partner, not because they're not interested.
But what it does is remove the admin around the conversation so I can actually be present for the conversation. That's the whole point of the Autonomous Business framing I keep coming back to. Automate the memory and the paperwork. Keep the judgement calls human. The moment you flip that, and let AI make the calls while you do the admin, you've built it backwards.
What to actually build if you're starting from scratch
And you don't need SalesM8 or a custom setup to get most of the benefit. Start with one piece:
- Pick the before-call briefing first. It's the biggest win for the least effort. A note pulling last touchpoint, deal stage, and one open question, sent twenty minutes before the call.
- Record and transcribe every call you're not already doing this for. You cannot improve what you can't review, and you definitely cannot review something that only exists in your memory an hour later.
- Look at your own talk ratio before you buy anything. Pull three recent transcripts and roughly count who talked more. If it's you, more than the buyer, on most calls, that's your actual problem, not your pitch.
- Automate the follow-up last, once the first two are running properly. It's the easiest to automate badly, sending generic recap emails nobody reads.
Do the maths on what this is actually worth. If you run twenty sales calls a month and even a handful convert from "would have lost" to "won" because you asked the right follow-up question instead of guessing, that's not a marginal improvement. That's the difference between a good quarter and a rubbish one.
Sources:
- AI Pre-Call Research in 2026, getdarwin.ai, 5 May 2026
- Salesforce Announces State of Sales Report for 2026, Salesforce, 3 Feb 2026
- How AI Improves Sales Rep Win Rate: The Real Data (2026), Nimitai, 25 May 2026
- AI Sales Tools Stats 2026, Stealth Agents, 6 Jun 2026
Want to see how this could work in your business? Book a call and let's talk about where you're at and what's possible.
Brewed by Steven, poured by Viktor
About Steven Tann: Steven helps business owners build systems that run themselves using AI. After 10+ years helping 7,000+ businesses and building his own autonomous operations, he's the bloke who actually does it, not just talks about it. Find out more at steventann.com.