Will AI replace 80% of sales reps?

I've been speaking to John for about fifteen minutes already, and he's back on after three years — do you remember, three years ago I reached out and said jump on the podcast? For those who don't know him: I've followed this guy for twelve years, he's the person who introduced me to sales. So John, quick run-through — who are you and what do you do?

John Barrows, CEO of JB Sales. I've been running my own sales training and consulting firm for the past fifteen years — training companies like Salesforce, LinkedIn, Box, Dropbox, all the fun tech and SaaS companies. I grew it to about six million and twenty employees, and then 2023 hit and it all came crashing down, because all my clients were SaaS and had money in Silicon Valley Bank. It took me from six million to zero. Now it's back to me and my COO.

I'm more of a sales professional who happens to train. I'm out there selling every day, and if I didn't have the skills to sell I wouldn't be here — that crash forced me to use every one of them. My point is we've way over-engineered the sales process. We've taught tech and process instead of fundamentals, so we've got a whole generation that doesn't know how to fundamentally sell. I want to get people back to fundamentals and then augment with AI as a superpower.

Today we're talking about the future. Before we hit record I was telling John about Open Claw — an AI agentic framework that can go into your computer and pull any information it can reach. I spent the whole weekend setting it up, and I'm not a coder. I told it 'I want five leads a day' — companies that raised Series B, C or D, hiring a new person, doing ten to twenty-five million ARR — and boom, beautiful leads. Then I hesitated, because that's a lot of control to give up.

I'm higher on the risk scale than most, so don't follow my track — this is personal. But the IP thing is interesting, because I think IP is over. AI has zero guardrails here in the States, and it's already consumed the entire internet. I've been putting content online for twenty years — you could go into ChatGPT right now and say 'coach me like John Barrows,' and it will.

Last year I had Guy Kawasaki on my podcast. He'd built a Kawasaki GPT from every book, interview and podcast he'd ever done, and he said if you ask the GPT a question you'll get a better answer than asking him. So the night before, I interviewed his GPT with the exact same questions. The next day I asked him the same questions in the same order, then told him — and I had to be honest, his answers the night before were better.

He started talking about legacy, and I said no, that's for you. For someone like me trying to learn something from you, what's the point of this conversation if I can get better answers from your GPT? And he didn't have a good answer. When Guy Kawasaki, with his breadth of knowledge, couldn't answer that — as someone who makes a living transferring knowledge to other people, I knew I was screwed.

I've supported the cadence tools, the Outreaches and SalesLofts, but I've never personally used them, because I've always preached authenticity. Getting an unpersonalised email from me isn't sales — and honestly the whole SDR and BDR world is moving into marketing. It's impression points, not sales. If you look at it that way, fine: automate it, personalise it, make my inbox like Instagram so I only see what's relevant. That part I actually think is good.

What I don't trust is the control and the security. You could tell me upside down that it's secure and I won't believe you, because I believe the technology will lie to get around its own security — not the humans, the tech. I've proven it over and over: it blatantly lies, then when you catch it, it says 'oh, you're right.' Now I ask it to rate its own answer one to ten, it comes back a seven, and I tell it to go source-check until it's a ten.

We've gone from a search engine to an answer engine, and that's dangerous for most people. With search there was at least a perception of choice — you read an article or two and formed your own conclusion. Now it just gives you the answer. For the curious it's a superpower: give me another perspective, source-check that, role-play this. But most people are conditioned to be lazy — they'll ask the question and copy-paste the answer, and in that case I don't understand the value of the human.

So I've built individual custom GPTs for each part of the sales process. My meeting-prep one asks who I'm prepping for, then starts with the industry — here are the relevant trends, here's why they matter, what else do you want to know? Then the company, reviewing its 10-K reports; then the persona; then it helps me build two or three hypotheses and impact questions. It forces me to learn. It takes ten minutes instead of the hour it used to — but it's not zero minutes, because that's where the value is.

What I won't skip is the business acumen. If it tells me a company just got a new CRO, I don't want it jumping straight to 'write me an email.' I want it to tell me what that CRO is accountable for, their KPIs, what their thirty-sixty-ninety looks like and the challenges they face — then we write the email off that. Without it, we're ripping the business acumen away from these kids and just producing an outcome.

My fear isn't the tools automating sales — it's the client waking up and realising they don't need us. Any rep who thinks they're safe: go into ChatGPT and pretend you're the buyer. 'I'm a VP of sales, fifty-person team, twenty SDRs, thirty AEs, looking at JB Sales training — pros and cons, how does it compare to the top three vendors on G2, what should I ask John?' You get better info in five minutes than from a twenty-two-year-old SDR who asks bad questions and gives you no value.

I spoke at Pavilion — my keynote was 'Anything But Predictable,' about how the predictable revenue model ruined sales. Segmenting the roles turned it into an assembly line instead of a skill. The SDR role as a business unit is a nightmare: the average SDR stays at the company 1.4 years, fewer than half ever become AEs, and average quota attainment is about sixty-two percent. So we need to get back to full-cycle sales — I'd make everyone a GTM generalist for their first two years.

Here's a framing the AI actually handed me: today we're talking about the future. John says B2B sales is dead, it's all B2C now, and eighty percent of reps are going to get smoked by AI because they're average — and he's proving it by changing his own business model. If you want to be in the twenty percent who survive, listen up.

Every person should be using AI to figure out how to replace themselves. We're in a transition — you're not fully replaced, but it's the eighty-twenty rule. The human still needs to be the last mile: the QC before the email goes out, the check on the proposal before you hit send, the eyes on what it aggregated to make sure it's right.

A friend of mine, an accountant, had been out of work over a year. I asked where he was on AI and he said he'd used it to update his resume. I said, that's the most basic level there is. Go figure out how to automate eighty percent of what you do, then in the interview say: you're not hiring me for the eighty percent I've automated — you're hiring me for the thirty years of experience to know the last twenty percent matters, and to spot when the output is wrong.

Just get in the game and get curious. I'm insanely optimistic for the twenty percent — honestly probably more like ten — who are curious and willing to do the work. I'm insanely pessimistic for the other eighty percent who are lazy and want the bare minimum, whose jobs in a lot of cases are already irrelevant. White-collar workers are in trouble, and sales is near the top of the danger list. You'd better be a heavy hitter.

Thanks for watching the whole way through. If you liked it, hit subscribe — it takes two seconds and helps me reach great guests. And if you ever want to know what we do as a business, go to gorevintel.com, where we help founder-CEOs make decisions backed by intelligence. That's it for now — see you on the next episode.