The Builders, the Waiters, and the Last Mile
- Sales Gambit Insights
- Jun 5
- 7 min read
AI agents in sales: what SaaStr 2026 revealed
Most teams that bought an AI SDR switched it off within three months, tired rather than transformed. What the winners understood was narrower and harder: an agent scales the sales motion you already have, and most companies never had one worth scaling.
At SaaStr AI in San Mateo this May, a founder stood up during the closing session and told the room that the best salespeople win on relationships. People like them, he said. They trust them. The man taking questions, who has spent twenty years around sales, asked him when he had last met one of those buyers in person. The founder did not have an answer. Neither, it turned out, did most of the room.
We found a split. Some founders are waiting for the dust to settle and a safe playbook to appear. Others have stopped waiting and are building, failing, and learning faster for it. The distance between the two is growing by the day, and it has little to do with which tools anyone has bought.
What follows is an account of that split, and of the one decision underneath it that most companies are getting wrong.
There is a kind of founder we often think about. He runs a company somewhere past Series B, and he still steps in to close the biggest deals himself, because something happens when he is in the room that does not happen when his reps are. He has never said this aloud. Saying it aloud would mean admitting the business does not yet run without him.
This year, he tried an AI SDR. The logic was clean. Take the repetitive work off his reps, let the machine fill the pipeline, and buy back the hours. Ninety days later, he turned it off. It had sent thousands of emails to book a handful of meetings, and it had taught him something he did not want to learn. A tool that scales your sales process only scales what is already there. What was already there had always been him.
“A tool that scales your sales process only scales what is already there.”
We watched that lesson play out in front of 10,000 people. Earlier this year, we wrote about the AI Outbound Divergence, the widening gap between meetings booked and revenue earned. We thought SaaStr might complicate the argument. It sharpened it.
A floor full of the same promise
The thing nobody says from the stage is visible the moment you walk the floor. One AI SDR, then two more selling the same promise in almost the same words. AI-native CRMs in a row. Every tool for enrichment, scoring, or outbound is a slight variation on the one beside it. A question started moving through the halls, the kind that spreads because everyone is already thinking it: why not just use Claude? Founders carried it from booth to booth. The usual answer was “proprietary,” and it rarely survived a follow-up question. The people building these tools are smart, and by their own admission on stage, they are working without a map. Sam Blond, who ran sales at Brex before starting his own company, said there is no Cursor for sales. No standard has been set. The category is being built in public by people guessing well.
The guessing has a cost, and the people paying it were in the building. Between 50% and 70% of teams that deploy an AI SDR abandon it within 3 months. The ones who got burned were not the cautious holdouts. They were the believers, the early movers, the founders who bought first and asked questions later. They aimed the tool at their market and watched it scorch their sending domains and write outreach that their prospects found embarrassing. One had run 8 vendors at once, a contest so crowded that none of them got a fair hearing. He didn’t conclude AI outbound was a fraud. He was just worn out by it.
It would be easy to read that as a verdict on the tools. It is a verdict on the approach. Because in the same building, on the same days, the other half of the room was telling a different story
The line the market is dividing along
Jason Lemkin, who runs SaaStr, admitted that his own AI setup a year earlier had been bad, so he scrapped it. He pulled every vendor that wasn't built for this era, rebuilt from the ground up, and cut a team of about 17 people to 3 humans and 21 agents. The team did more, not less. He didn’t wait for a playbook. He built badly, fixed things while everyone watched, and got better in public. When a founder pressed him on the bugs still in his software, he did not flinch. The ratio of value to slop keeps moving in one direction, he said, and the only people who lose are those standing still.
“The builders are no cleverer than the waiters. They started before they felt ready”
This is the line the market is dividing along, and it has little to do with who is smarter. The builders are no cleverer than the waiters. They started before they felt ready and treated each failure as the cost of learning. The waiters are holding out for a certainty that is not coming this year, and may not come the next.
The reward for moving is not marginal. Lovable, out of Stockholm, reached $100 million in annual recurring revenue 8 months after launch, faster than any software company before it, ahead of OpenAI and Cursor. Cursor went from $100 million in recurring revenue last January to $2 billion by February, the fastest any B2B software company has reached that figure. The public markets are running the same way now, with the largest IPOs forming around AI companies, and Anthropic crossing $30 billion in annualized revenue, roughly triple where it sat at the end of last year. The upside is real, it is documented, and it is going to the people who did not wait for permission.
So the first half of the instruction is plain. Move. Skip the 18-month evaluation, the steering committee, and the pilot that never ends. In a market where the products change every week, going slowly is a decision, and it is the wrong one.
What the agent actually scales
The second half is where the money is, and it is the half most companies get wrong. An agent does not bring judgment to your business. It scales the judgment already there. Give it a sales motion that works, and it will run that motion at a volume no human team could survive. Give it a motion that only ever held together because two or three good people were quietly covering for it, the thin discovery, the generic messaging, the process that was really just busy people filling the space where a process should have been, and it will reproduce that emptiness a thousand times an hour. One line went around the floor more than once. If you cannot sell it yourself, the agent cannot sell it for you.
For anyone who has built a real sales organization, this is the best news in a decade. The discipline you set, the discovery standard you held, the instinct your best people carry in their heads, all of it has been trapped at human scale, capped by how many good people you could hire and how long it took to make them good. The agent lifts that cap. When Anthropic’s own demand went vertical last year, the team couldn't hire fast enough, so instead of lowering the bar to fill seats, they took the methods their best reps already used and turned them into tools any rep could reach for. A new hire’s first day started to look like a strong rep’s ordinary Tuesday. Talent stopped being the bottleneck and became the template.
What they did not do is as telling as what they did. They did not tear out Salesforce, Gong, or the rest of the stack they had spent years tuning. They kept what worked and ran the intelligence through its seams. 54% of their new enterprise logos last year came in through a self-serve path that never touched a salesperson until late, and the salespeople they kept were placed with intent, at the moments when trust and judgment decide whether a deal closes. The shift, as their head of commercial put it, is that sales leaders are becoming systems thinkers. The job is no longer managing deals. It is designing the system that lets good people do their best work at a scale they could never reach alone.
The Last Mile
A banner plane flew over San Mateo throughout the conference. It belonged to Monaco, and the founder mentioned from the stage that it had cost him $15,000. Consider what his company does. Its agents handle the outreach, the sequencing, the database work, the volume that used to eat a sales team’s week. And given the choice of what to do with that money, the man whose business is automating the grind put his brand in the sky above a field of 10,000 people standing together, looking up. He did not buy more automated reach. He bought presence. Hand the machine everything it can do at scale, and spend your scarce human attention on the things that only happen in a room.
It was the note nearly every speaker returned to before leaving the stage. Keep the human in the last mile. Let the agent carry the volume, the drudgery, the research, the formatting, the chasing, the work that ate your best people’s hours and was never the reason anyone bought from you. Then put those people where they belong, across a table from a buyer, reading a room, building the kind of trust that does not survive a video call and cannot be faked by a machine that has never met anyone.
“Keep the human in the last mile”
The playbook everyone is hunting for does not exist yet, and nobody is hiding it. It is being written right now by people who are willing to do the work badly before they do it well. It will not arrive in the mail. By the time it is clean enough to hand around, the companies that wrote it will be years ahead.
What the best of them understood before they bought a single tool is the thing we have believed from the start. AI was never going to make selling less human. Used well, it clears away everything standing between your best people and the work only they can do, and gives them back the hours and the reach to win.
We should be honest about how long any of this will hold. Things are moving quickly enough that within a year, some of these names will have been passed, some of these numbers will look small, and some of these conclusions may not survive what comes next. That is what writing about this moment costs. We accept it.
We will not watch from the sidelines. We mean to keep pace, and where we can, to stay a step ahead, so that the people building go-to-market in this era have someone thinking the problem through alongside them rather than selling them the next tool. That is the work. That does not change.
If you handed your sales motion to an agent tomorrow, what would it scale?
Comments