AI Just Broke Math. What Does That Mean for Your Pipeline?

TL;DR

What does OpenAI disproving a math conjecture have to do with sales?

The point is that AI found a counterexample humans kept missing because it didn't assume the conjecture was true. The same thing happens in sales: AI can surface what actually converts instead of repeating assumptions nobody has tested in years.

How should sales teams actually use AI in their outbound process?

Use AI to pressure-test the assumptions your process is built on, not just to send more emails faster. Running controlled tests on things like email length or follow-up frequency can reveal that widely accepted 'rules' are wrong for your specific audience.

What is the difference between using AI to go faster versus using it to be more efficient in sales?

Faster means higher volume on your current process. Efficient means finding the process that actually works first, then running it. Teams winning right now are using AI to question their model, not just to fill the pipeline with more of the same outreach.

An AI model just proved a decades-old math conjecture false. Not 'helped a researcher find a flaw.' Straight up disproved it. On its own. That's not a research footnote. That's a signal.

I'm not a mathematician. I run a company that obsesses over outbound sales and pipeline efficiency. But when I read that story, my first thought wasn't about geometry. It was about all the assumptions baked into how we run sales that nobody has bothered to question in years.

We treat a lot of our go-to-market beliefs like math conjectures. Assumed true. Never really tested. Just passed down from one revenue leader to the next.

The Assumptions Running Your Pipeline Are Probably Wrong

Here's a short list of things I've heard treated as gospel in B2B sales: Send more emails to get more replies. Follow up seven times minimum. Tuesday morning is the best time to reach a prospect. Personalization means using their first name and their company name.

Some of these have data behind them. Most of them are just things that worked once, got repeated enough times, and became 'how it's done.'

I've been guilty of this. Early on, I inherited a cadence from a sales playbook someone had built in 2019. We ran it for almost a year before we actually looked at the reply rates by step. Step four had a 0.3% reply rate. We were burning sender reputation and annoying prospects just to hit a number on a sequence template.

We killed step four. Replies went up. Unsubscribes went down. The 'rule' was wrong.

What AI Actually Does to These Assumptions

The reason that OpenAI story matters isn't that AI is smarter than mathematicians. It's that AI doesn't carry the same assumptions into the problem. It doesn't 'know' the conjecture is supposed to be true. So it finds the counterexample that humans kept walking past.

That's exactly what good AI tooling does to your outbound motion when you let it. It doesn't care that your old SDR swore by a certain subject line formula. It tests. It finds what actually converts. It surfaces the counterexample.

The problem is most teams aren't using AI that way. They're using it to do more of the same, faster. More emails. More 'personalization' at scale. More of the thing that already wasn't working, just cheaper.

That's not AI improving your process. That's AI accelerating your mistakes.

The Real Unlock Is Questioning the Conjecture

Before you add another AI tool to your stack, do this first. Write down the five assumptions your current outbound motion is built on. Not your tactics. Your assumptions. Things like 'our ICP responds better to problem-led messaging' or 'we need a minimum of 200 touches per month per rep to hit quota.'

Then ask: when did we last actually test that? Not gut-check it. Test it.

If you can't answer that question, you're running a sales process built on conjectures. Some of them are probably right. Some of them are almost certainly wrong. And one of them might be the thing quietly killing your conversion rate.

The math world just got a reminder that held beliefs don't survive contact with real testing. Sales is no different.

What We Changed When We Ran This Exercise

We did this internally about six months ago. Listed our core assumptions. Ran controlled tests on the ones we couldn't verify. Three of them held up. Two didn't.

One of the ones that failed: we assumed shorter emails always outperformed longer ones in cold outbound. That's basically industry doctrine. Turns out, for our specific ICP (technical founders at companies between 50 and 300 employees), a slightly longer email with a concrete technical hook outperformed our short version by about 40% on positive reply rate.

We would never have found that if we'd just kept optimizing inside the assumption.

AI helped us run those tests faster. But the decision to question the assumption in the first place? That was human. That's still your job.

Don't Automate Your Blind Spots

There's a version of AI adoption in sales that makes you faster and a version that makes you more efficient. Those aren't the same thing.

Faster means you execute your current process at higher volume. Efficient means you find the process that actually works and run that.

Most teams are chasing faster. The ones winning right now are chasing efficient. They're using AI to pressure-test their model, not just to fill their pipeline with more noise.

A math conjecture that stood for decades just got disproved because someone (or something) was willing to actually check. Your sales assumptions deserve the same treatment.

What's the one belief your outbound motion is built on that you've never actually tested? Drop it in the comments. I'll tell you how I'd go about breaking it.