AI Needs Its Sewing Machine Era
To be truly valuable, sales AI can't just copy humans.
I gotta be honest with you all. I’m in the midst of reviewing Q1 and kicking off Q2 just like friends of the newsletter Ryan Burke and Steve Travaglini discussed last week. Between working last minute deals, scrubbing pipeline, reviewing metrics and getting started on board prep, I’ve not had a lot of time for newslettering.
That said, I did want to take a brief moment to talk about what’s on every CRO’s mind this time of year: sewing machines.
But before I do that, let’s talk about AI SDRs.
Many of you probably saw the news about the AI SDR company 11x last week. If you didn’t, here’s the tl;dr. TechCrunch took a break from proclaiming themselves an “iconic brand” to report some news—accusing 11x of fraudulently reporting ARR (based on 3 month opt-out clauses) while sporting “70-80%” churn. Naturally this was followed by a rebuttal from 11x’s CEO along with an avalanche of takes, both pro-ish and trollish.
As for me, I’m on record with my skepticism of the whole premise of AI SDRs (in today’s form) and what their use says about how much you respect your prospects.
11x is one of many companies looking at semi-rote “digital” jobs (like many kinds of sales, support, copywriting, some kinds of software development… ok honestly a lot of jobs) currently done by humans and thinking “man, we should replace all those with AI”.
Influential folks (inside and outside the sales world) who are thinking about AI’s future see a world of autonomous agents that basically act as direct replacements for human jobs. Just this morning, I saw a LinkedIn post from Yamini Rangan, CEO of HubSpot, that pretty much summed up the perspective:
But AI? It is different. It’s giving rise to a whole new concept: “work” or “results” as a service. So, what does that mean exactly?
Simply put, if software-as-a-service helps people do work, AI will do work to help people.
Take a salesperson, for example. Their "job to be done" is to close deals. Every day, they carry out tasks to help them achieve that goal. They spend hours updating records, scoring leads, drafting emails, scheduling calls, researching companies, and following up.
Software made it easier to do all this work. But the salesperson was always the one who actually did it. (I was in sales, I know what it’s like 😉)
Now, AI is bringing intelligence to software. It can think. It can reason. It can do work – like updating contact records and crafting follow-up emails.…
When we buy software, we’ll focus less on features and functionality, and more on the specific work and results it can deliver to help our businesses grow. In other words, we will think of software as “work” or “results” as a service.
Today, we use passive tools. Soon, we'll use active agents.
Today, we buy licenses for our teams. Soon, we’ll buy outcomes for our business.
Yamini’s clearly on the same wavelength as HubSpot’s cofounder and CTO, Dharmesh Shah. Dharmesh has talked about hybrid teams composed of humans and AI agents and is even creating a “professional network” for those agents.
It’s not just something in the water at HubSpot. Marc Benioff wants Salesforce to be the leading provider of “digital labor”. Kyle Poyar has been banging this drum on the pricing side. Meanwhile, plenty of VCs are excited about how AI agents will bring about a “Service as Software” revolution.
The future they describe looks a lot like autonomous agents that are largely a 1:1 replacement for human roles.
And that brings me back to sewing machines.
It’s hard to imagine now, but people (mostly women, because of course) used to spend a large percentage of their time making and maintaining clothing—it took 14 hours just to sew a shirt by hand. Sewing was a skill, albeit one that was manual, slow and just part of everyday life. Not unlike doing outreach as an SDR, I suppose.
As the industrial revolution got going in the 1700s, automation was in the air and lots of folks tried to build machines that could sew. The first patent for a mechanical sewing device was in 1755 but it wasn’t until nearly 100 years later that a series of commercial devices actually succeeded, among those from Elias Howe and Isaac Merritt Singer. Singer, as you might have guessed, was the better sales person.1
Part of the reason it took so long to build a commercially successful sewing machine is that the way humans sew is super hard to mechanize. Because of this, sewing machines don’t sew at all like humans. There are special “eye-pointed” needles, levers, bobbins and “feed dogs” all operating in a carefully choreographed dance. Seriously, just watch this short video narrated by what I can only assume is someone’s kindly YouTube grandma.
It’s effective, elegant and oddly satisfying to watch—but it’s nothing like how a human makes a stitch.
I’ve always loved this example because it’s a reminder that the best application of technology to a job to be done isn’t to blindly do what humans do in the same way they do them. The job needs to be broken down, rethought and reassembled in a new way that achieves the same thing. And it often does it better, faster, stronger.
All of which brings me back to 11x, AI SDRs and digital labor in general.
We’re in the “recreate what humans do” era of AI. Very smart people are reaching for the simple idea: recreate the same human work with AI. It’s embedded right there in the concept of “agents” and “digital labor”—11x provides an AI SDR, Devin.ai provides a software engineer, Agentforce provides a customer service agent. Those will do all the same things their human equivalents do in the same way they do—book meetings via phone/email, submit pull requests in Github, resolve tickets in a support system.
I don’t know exactly how this plays out but I have to imagine that the first truly successful AI SDR won’t really be an SDR as we know it. It’ll break down the job to be done—create demand—in an entirely different way. That solution probably won’t be built to replicate the specific capabilities and limitations of different humans that exist in today’s GTM roles. Maybe it’ll even do the job in a way that’s more appealing to the customer and more effective than today’s outbound.
I’m eager to find out what the eye-pointed needles, bobbins and feed dogs are in the AI systems of the future. My guess is that, like sewing machines, these systems won’t fully replace humans but will make them orders of magnitude more productive.2 Unlike sewing machines, it probably won’t take decades for the solution to emerge. It’s going to be an interesting few years.
I won’t go into all the history here, but the development of the sewing machine is actually a fantastic microcosm for explaining how technology impacts our lives. The story involves inventions, cutthroat global sales teams, at least one french riot and the changing role of women in society. If you’re interested, here’s a YouTube video with some history and plenty of slick technical animations to marvel at.
In reality this may replace human workers in these roles. Not unlike how US agricultural employment has declined from ~33% of all workers in 1900 to ~2% today while producing vastly for more food. It’s also possible AI’s a different beast entirely and we’re all doomed.