AI has supercharged outbound volume and cut cost-per-touch across enterprise sales, but win rates remain flat. Karl Pinto, former Regional Enterprise Sales Director at PagerDuty, argues most teams pointed AI at the throughput layer instead of the diagnostic layer, automating activity rather than improving qualification, deal inspection, and executive access.
Enterprise sellers have never generated more activity. Their win rates have not moved. Karl Pinto, who built one of PagerDuty’s top-ranked enterprise teams, argues most companies are aiming AI at the one part of the sales process that was never the problem.
Enterprise sales has spent two years getting faster. Reps draft outbound sequences in seconds, summarize discovery calls before they have left the room, and let models score and rank the pipeline overnight. On nearly every team that has adopted the tooling, activity is up and cost-per-touch is down. Win rates have barely moved.
That gap is the part most go-to-market leaders are not discussing, and to Karl Pinto it points to a basic error in how the industry is spending its AI budget. Pinto has spent nearly two decades in enterprise software across Dell, Salesforce, and PagerDuty, most recently as Regional Enterprise Sales Director for the Northeast, where he built and led one of the company’s top-ranked global enterprise teams. His read is blunt: AI has solved a problem enterprise sales never actually had.
“Speed was never the bottleneck in a complex deal,” he says. “You can send a hundred more emails and run a dozen more calls and still lose, because the thing that decides the deal happens somewhere those activities never reach. We bought a faster car. The traffic is on a road the car never drives.”
In Pinto’s experience, enterprise deals are not won or lost on volume. They turn on two things that resist automation: whether the seller has qualified the opportunity honestly, and whether they have earned access to the person who actually controls the budget. “Most pipelines are fiction,” he says. “It looks real in the system because someone logged a meeting and set a close date. Whether it is real depends on questions a dashboard cannot answer. Does this account have a problem worth paying to solve, and are we in front of the person who signs for it?”
He describes a pattern he has watched repeat inside hypergrowth sales organizations: teams generate enormous activity against accounts that were never going to buy, then act surprised when the forecast slips. “Activity is comfortable. It feels like progress,” he says. “Qualification is uncomfortable, because half the time the honest answer is that the deal is not real and you have to walk away from it. AI made the comfortable part frictionless and left the uncomfortable part exactly as hard as it always was.”
The problem, Pinto argues, is where teams have deployed the technology. Most have aimed it at the throughput layer: writing more messages, booking more meetings, producing more first-touch volume. Few have aimed it at what he calls the diagnostic layer, the inspection work that determines whether any of that volume converts. “Point it at the wrong layer and all you do is manufacture bad pipeline faster,” he says. “Your reps are busier, your CRM is fuller, and your win rate is identical. You have automated the noise.”
The diagnostic layer is harder to build for, which is part of why it gets skipped. It means using AI to pressure-test a deal rather than populate it: surfacing which opportunities have a validated champion, which have stalled on a single contact, which carry a close date nobody has justified, which have never once touched someone with budget authority. “That is the work that moves a number,” Pinto says. “It is just less photogenic than a tool that writes your emails for you.”
Pinto’s own approach treats qualification as an operating system rather than a reporting formality. He runs his teams on MEDDPICC, the enterprise qualification methodology, but insists the acronym is not the point. “Half the companies that say they run MEDDPICC are running it as a form somebody fills in after the deal is already decided,” he says. “That is theater. The discipline is inspecting the behavior, not the field. Did the rep actually meet the economic buyer, or did they type a name into a box?”
That distinction produced numbers that are hard to argue with. The enterprise team Pinto built closed roughly seven of every ten opportunities it qualified, a win rate well above the enterprise software norm, and he personally ran the largest deal of its kind in the company’s history, a seven-figure agreement at one of the largest banks in the United States. He is direct about why those results held: the team disqualified aggressively and refused to advance a deal until it had tested its access to real authority. “Executive access is a gate, not a nice-to-have,” he says. “If we could not get to the person who owned the budget, we did not have a deal. We had hope. AI can help me find that person and prepare for the conversation. It cannot have the conversation for me.”
He sees that same gate as the right place to point the technology. Used well, AI can tell a manager which deals in a forecast have never reached an economic buyer, the exact signal most teams discover far too late. “Imagine inspecting an entire pipeline for that one question every morning, instead of finding out at the end of the quarter,” he says. “That is a real use of the tool. It is just not the one most people bought it for.”
Pinto is not skeptical of AI in sales. He is skeptical of using it to do more of what was already not working. The teams pulling ahead, he says, are the ones putting AI underneath a qualification discipline rather than on top of an activity quota. “The winners will not be the teams that sent the most emails,” he says. “They will be the teams that knew which deals were real the earliest and spent their time only on those. That has always been the game. The tooling just raised the stakes on getting it right.”
His closing point lands as a warning more than a forecast. As AI drives the cost of activity toward zero, the teams that mistook activity for progress will produce more of it than ever, and convert none of it. “Faster is not better,” Pinto says. “It is just faster. Better is knowing what to walk away from, and that is still a human decision.”
Get the most important tech news in your inbox each week.
Opinions expressed by authors are their own.