The 10x Sales Rep
A deep dive into what's possible with AI.
I used to be a software engineer. More than 20 years later, I still remember reading an essay from Paul Graham, the founder of Y-Combinator, called “Great Hackers”. He said this:
A great programmer might be ten or a hundred times as productive as an ordinary one, but he'll consider himself lucky to get paid three times as much.
The bold part stuck with me. It also stuck in the psyche of developers, startup founders and VCs enough to become part of the Silicon Valley ethos.1 (The non-bold part matters too—more on that later.)
Over time this solidified into a simple concept: the 10x developer. A special group capable of producing 10x more than their peers.
For software engineers like me at the time, it was something to aspire to. I don’t know if I achieved it, but I did build a few things that others found difficult.
For founders, the goal was to hire some 10x developers, provide them interesting problems, ignore their quirks2 and reap the rewards.
For investors, the goal was to back founders who could attract 10x engineers—fund a Steve Jobs who brought along a Steve Wozniak.3 The economics were simple. Make these engineers pretty rich while making yourself very rich. They wouldn’t cause trouble trying to claim more of the value they created because they just wanted to work on interesting problems.
AI (probably) changes this equation. I know lots of extraordinary developers becoming even more extraordinary with a dose of Github Copilot, Cursor or Windsurf. Maybe you don’t even need to be a good engineer. Maybe every vibe coder can spend $20 a month for their very own 10x engineer.
But this isn’t about developers. It’s about whether we can apply the 10x concept to sales and how AI might help us create the 10x sales rep.
Can there be a 10x sales rep?
First, let me define what I mean by the question. Obviously some reps produce millions in revenue while others produce thousands. Why not call that 10x or even 100x?
If a rep at Oracle is closing $10M deals, it’s not fair to compare them to a mid-market rep selling a $50k SaaS product or an SMB rep banging out $10k deals. In sales, the normal way to get 10x the output is to sell 10x larger deals. Sure, skill matters but the game you’re playing matters more.
I’m talking about reps who are all playing the same game.
In particular, I’m thinking of the “classic SaaS” mid-market rep—the one selling $50k deals with a $1M quota. What does a 10x rep look like in this context? Could someone actually carry a $10M quota and succeed?
From bell curves to power laws
We all know some reps are better than others. Sure they’re rockstars, but they’re not 10x. Sales is more like sports in that way.
Take pro baseball. Going back to the 1871 season, the league-wide batting average has itself averaged .262. Bat .300 and you’re very good. Bat .400 and you’re one of the best to do it. The best single season batting average ever was .471 by Tetelo Vargas of the New York Cubans—only 1.79x the long term average.
Of course batting average is capped at 1.000 so even if you got a hit every at bat you could never be better than ~4x the average. So let’s take another metric that tries to encapsulate the total runs created by a hitter: RBat+ which should be uncapped. The math is wild but doesn’t matter that much. Just know that 100 is average. The player with the greatest RBat+ season ever is Barry Bonds in 2004 at 282. The most juiced player in his most juiced season managed to be 2.8x better than his peers. That’s a lot, but not 10x.
Unlike building software, sports operate on bell curves, not power laws—there are limits to human performance. The only way to achieve 10x performance is to turn sales into a game that operates on power laws where inputs can be magnified into exponentially more output. That means software-style productivity that removes the limits of human performance.
Interestingly, we already have a clear example of this in sales: PLG. PLG approaches the sales problem differently. Instead of trying to recreate what humans do, PLG treats it like a sewing machine and uses an entirely different method. It lets the product sell itself and removes the sales rep entirely. This means a small growth team can be responsible for many millions in revenue by operating at a different level of abstraction through software.
But (so far), pure PLG doesn’t seem capable of moving $50k B2B software. If that’s the game we’re playing, we need to find a different way to remove human performance limits. And that—it seems—means applying AI.
In the rest of this article I’ll investigate three different scenarios for creating a 10x rep, how AI might be applied to each, and how likely each one is to succeed.
Three 10x scenarios
I’ve mapped out three potential ways to get this mythical mid-market rep from a $1M quota to a $10M quota:
The Volume Play - Use AI to massively increase the number of accounts the rep can prospect into and the number of opportunities they can manage at once.
The Efficiency Play - Use AI to double win rates and opportunity creation rates.
The ACV Play - Use AI to make the rep so effective at qualifying and working deals that they can double their ACV. Note this assumes that a product with a $50k ACV has a $100k ACV use case for the same approximate target customer. I think that’s fair. I don’t think it would be fair to assume it has a $1M use case for that same customer.
Below are the 3 scenarios modeled out. This isn’t super sophisticated. For this thought experiment, I just did simple reverse pipeline math and filled in a bunch of reasonable assumptions4. I’m also not complicating this with ramp time or sales cycle length. We’ll just assume a fully ramped rep operating at steady state.
Let’s dig in.
The volume play
The Play: Use AI to massively increase the volume of accounts and deals a rep can work simultaneously. Conversion rates stay the same.
AI Required: Data Enrichment, AI SDR, AI Meeting Avatar
First let me just say that the math changes a lot if you can source large numbers of inbound leads. In fact, there probably needs to be some pipeline creation support for even the baseline case to make sense. That may come from inbound or an SDR. In this thought experiment, though, we’re trying to imagine a single rep being responsible for this entire number end-to-end, so let’s run with that.
The first place to apply AI here is to automate the truly massive number of touches necessary to build pipeline in the first place. Even if we assume that 10% of engaged accounts result in an opportunity (“opportunity creation rate” aka OCR), we’re still left with having to work 10,000 accounts. We assume all those accounts need a decent level of multi-threading and fairly lengthy sequences. Seems reasonable for $50k ACV.
That works out to 1,800 touches per business day. To support this we’d need an AI SDR doing massive outreach in parallel across multiple channels—email, phone, LinkedIn, etc.
We need that AI SDR to produce 4 opportunities per workday5 and we need a high overall OCR of 10%. To get to these levels, we’ll need three things:
A steady supply of high ICP fit accounts that could be customers
Good data about those accounts (both fit and timing) so the AI can craft relevant and timely messaging
The ability to reach contacts at those accounts with our messaging. This means good contact data and the ability to use all high-converting channels a human could.
I won’t say the overall data problem is solved but it’s the closest to solved of any of these. There are tons of data providers—both traditional and AI-enabled. It’s at least possible to get this pretty good.
A bigger challenge is how an AI SDR might orchestrate all of the actions that go into “working” these accounts. This isn’t a solved problem, but it probably will be. There are lots of folks trying to solve it and no major technical issues.
The final piece is a little tricky. Legal and technical barriers remain to completely automating outreach across channels. Robocalls are “almost always illegal”. LinkedIn will most likely continue TOS and technical mitigations to make it hard to automate things on LinkedIn. They plenty of motivation because otherwise they risk making LinkedIn unusable.
Because of the limits on other kinds of outreach, AI SDRs usually use mass email which is both legally suspect and has horrible conversion rates. Even assuming you somehow managed a 1% opportunity creation rate via that kind of email, you’d need your AI to work 100,000 accounts to get to the number of opportunities needed to produce pipeline to support a $10M quota. Your whole TAM may not be that large.
These legal and technical restrictions kill this volume play before it even gets started. However, let’s ignore that for a second and look at managing the active pipeline required to get to $10M.
Assuming each opportunity requires an average of 3 meetings (wins probably require 5+, losses require 1-2), the rep would need to hold 12 meetings each and every day. Assuming each meeting is only 30 minutes, that’s 6 hours each day. That’s technically doable but nearly impossible to sustain.
AI could actually solve for this if you take huge leap of faith—namely that humans would be ok talking to an AI Avatar on zoom instead of a person. If, for example, you could have an AI take the first discovery meeting and a demo you might be able to cut the number of meetings by as much half to 2/3rds. The rep could then hold only critical meetings for later stage opps. That might cut them down to 4 meetings per day.
One thing that takes very little imagination and is largely true today is that AI could handle most meeting prep work and follow up comms. That would limit the number of activities the rep has to do outside of attending the actual meetings and any higher-level engagement.
Verdict: Unlikely. This kind of outbound requires significant legal changes and this kind of pipeline management requires significant cultural changes.
The efficiency play
The Play: Use AI to get smarter and improve key conversion rates. This doubles both win rate and opportunity creation rate.
AI Required: Account Research, Rep Copilot
Let’s start at the top of the funnel. The first thing to note is that we “only” need to work 2,000 accounts to hit our pipeline creation target. This is due to doubling OCR (and more than doubling win rate—more on that in a moment). Because we’re focused on being significantly more efficient, we create 10x the revenue from 2x the accounts.
This play starts with making sure that we’re absolutely focused on the right accounts. If we’re not there’s no way we could turn 20% of the accounts we touch into opportunities. The first use of AI here would be researching the accounts to target. We need to know everything we can about the accounts and their timing situation—each one must be in top right “Good Place” quadrant of the FT chart.
Just like with the volume play, the data part of this play is mostly solved.
The next piece is how you orchestrate the outreach to these accounts. You still have to do lots of touches (2x more per day than the baseline) but this seems more solvable than the volume play.
For one, you could have a copilot making decisions about sequencing and combining automated email outreach with account prioritization. With a human-in-the-loop model like this, the highest priority accounts would get human engagement. The rep could handle ~50 of the daily outreach tasks but only with the accounts where it’s most valuable. The copilot would equip the rep with highly relevant messaging custom-built for each touch.
The combination of high quality accounts, an “omniscient” level of research, super-intelligent sequencing and a human touch seems at least promising for doubling opportunity creation rate. As a bonus, the human involvement removes much of the technical and legal barriers of a fully autonomous AI SDR.
That moves us to pipeline management. The biggest change here is that we dramatically improve the win rate. Note that we assume meetings per opportunity goes up in this scenario. Ironically, that’s the price of success. If more sales cycles are going well, the rep will likely have more meetings. Even with an additional meeting per opp, the number of meetings per day is doable because the rep doesn’t have to manage as many open opps due to the higher win rate.
At ~6-7 meetings per day, the rep could probably run all those themselves without having to resort to some kind of AI surrogate and the culture shift required to make that possible. This becomes even more likely with a copilot that reduces prep time by making sure the rep is well-equipped for each meeting.
This whole scenario hinges on a massively higher win rate. Why might that be credible? Two ways:
Better Qualification - win rate would naturally increase if AI can actually deliver higher quality accounts in terms of both ICP fit and purchase timing because the rep will be spending almost no time on accounts that aren’t able to buy
Better Deal Management - reps should win more if an AI copilot could provide instant access to all relevant deal information, encyclopedic product and market knowledge, and detailed account research
Verdict: Possible. Unlike the volume play, this doesn’t require any massive change to laws or cultural norms. It requires sophisticated data integration, online research and an AI copilot with a deep understanding of context. All of those things exist today—they just need to improve.
The ACV lift play
The Play: Use AI to make the rep so effective at qualifying and working deals that they can double their ACV.
AI Required: Data Enrichment, AI SDR, Account Research, Rep Copilot, AI Meeting Avatar
This scenario is closest to the classic way to 10x the production of an AE—sell larger deals. The problem is there are some major prerequisites for selling larger deals that are out of the rep’s control and no amount of AI can fix.
Namely, the product needs to deliver enough value for the target customer that they can spend 6 figures. That puts a natural ceiling on how much the ACV can increase without completely changing the target segment, the positioning and the product—in short, playing an entirely different game.
Let’s assume those factors cap the ACV for our mid-market rep to about $100k. That makes this scenario tough.
First, the pipeline creation and management load looks too much like the volume scenario. Even though the number of opportunities required is 2x less than the volume scenario, the higher ACV means we’re probably going to need to have 2x the meetings. That puts us back to the same overall pipeline management load. Solving for both of these with AI means we’re back to legal and cultural change.
Second, we also need to introduce more account research and sophisticated deal management into the mix. The rep has to be extremely good about managing deals if the ACV is going to go up. That means we’ll need a great copilot for the meetings the rep actually attends themselves.
In the end you have to get AI to do all the things from the two prior scenarios at once and stick the landing. That’s tough.
Verdict: Worst case. Factors outside the rep’s control limit potential ACV increases. This has all the drawbacks and challenges of volume without the upside from the efficiency case.
So, is the 10x rep possible?
Honestly, when I sat down to write this I thought the answer would be a qualified “no”. Now I think it’s a qualified “yes”. It just won’t take the form that most people seem to think. It may be possible to use AI to create a 10x rep with a strategy built to augment human efficiency, not replace humans with agents.
Increasing the volume of pipeline creation management activities by directly replacing humans with AI quickly runs into legal, technical and cultural barriers. Those may go away over time but that won’t be a simple process.
Increasing pipeline creation efficiency through better account selection and increasing win rates through rep “omniscience” has challenges. But overcoming those challenges only requires incremental improvements to existing technology—better data, better integration, better orchestration—not quantum leaps.
What does it mean for sales teams?
As we solve these incremental problems, expect to see quotas increase and headcount decrease. What this means for the economics of SaaS and sales remains to be seen. 10x reps won’t be like 10x engineers—content to solve problems and make good money. Instead they’ll look to get theirs. Quota:OTE ratios will probably stay stuck at their current max of ~5:1. 10x reps won’t initially make sales teams any less historically expensive.
Over the long-term 10x reps may become the norm. If they do, that will erode rep pricing power (like what may be happening for 10x engineers). At that point, we can expect quota:OTE ratios to creep up. But until then… 10x reps mean you won’t have to manage as many people, but don’t expect your CAC to get better anytime soon.
The end result, though, is exciting. The 10x rep is possible and teams of 10x reps make new GTM models possible. The pieces exist today even though they’re imperfect—better start putting them together.
This wasn’t a new idea. There was at least one study in the 60s showing huge variations in performance among individual programmers.
Most of Paul Graham’s essay is an attempt to describe 10x engineers. The overall picture is basically a stereotype—lone wolves living in their own heads with abrasive personalities. Over time, the “10x developer” idea often got twisted into an excuse to reward assholes who didn’t play well with a team.
It was pretty rare that the great founder was also the 10x developer. Bill Gates might be an exception. He managed to surround himself with enough business weirdos that he got the job done.
For example, 20% win rates are the average. 5 meetings for a $50k deal and 10 for a $100k deal has a little data to support it and it just seems reasonable based on personal experience that every meeting is surrounded by a flurry of emails—sometimes just a single follow-up but usually more.
Obviously an AI SDR can run 24/7/365 but we’ll keep it simple. We should generally expect the end-prospects to be working during business hours so we’re still somewhat limited on how we can blast them.