.svg)
Some deals are transactions and some are relationships, and cost, speed, and whether a human or an AI sales agent should close them all follow from which one you are looking at.
If you have followed this series, you know I usually keep to Storylane's own first-party data. This piece is different: the pattern I found in our pipeline turned out to be a symptom of a much larger shift in how software gets bought, so for once I went looking for outside corroboration, and it lined up almost uncomfortably well.
The shift is no longer subtle. As of 2026, 67% of B2B buyers tell Gartner they would rather buy with no sales rep involved at all, up from 61% two years earlier, and the share of the buying journey they finish alone before ever contacting a vendor has climbed from 57% in 2015 to roughly 80% today. Yet the lesson is not simply "spend less on sales," because buyers who go fully self-serve end up 1.65x more likely to regret the purchase, and that regret is worst among the Gen Z and Millennial buyers who now make up 71% of the market. So the real question is which deals genuinely need a person and which ones are quietly wasting one. The cleanest way to answer it is to stop sorting deals by size and start sorting them by type: transactions or relationships. Three findings from our 1,167 closed-won deals, and one customer's live deployment of RepX, drew that line for me.
Finding 1: SMB deals account for just 12% of revenue, but cost the same to close
Here is how our deals break down by size, and how much revenue each tier carries.
The top 7% of deals carry 34% of the revenue, the same slice produced by the bottom 41% put together. A team splitting its time evenly is pouring nearly half its effort into the tier that returns barely a tenth of the money.
That would be fine if small deals were cheap to close, but they are not, because a rep still runs the full choreography for them: a discovery call, a demo, a follow-up cadence, the same motion you would run for a deal worth ten times as much. And rep time is close to a fixed cost. A fully loaded SMB account executive runs $110,000 to $150,000 a year, and that number does not shrink when the deal does. So an outsized share of your most expensive people is spent on your least valuable deals. This is exactly the stretch of the funnel an AI sales agent like RepX is built to absorb, so that human capacity can follow the revenue instead of the deal count.
Finding 2: SMB deals close 4.2x faster than enterprise, because they're transactions, not relationships
Deal size and deal length move together everywhere. Across the industry, sales cycles roughly double with every 5x jump in deal value: small deals wrap in a couple of weeks while six-figure ones run four to seven months. Our pipeline sits right on that curve.
An SMB deal closes for us in one meeting over 15 days; an enterprise deal takes three-plus meetings across 64, which is 4.2x longer. It is tempting to read that fast SMB cycle as great selling, but it is really the signature of a simple decision: no committee to align, no procurement to survive, no champion selling upward. The buyer arrives mostly decided, gets a few answers, and converts. That is a transaction. The enterprise deal is the opposite animal, a relationship built over months of meetings and stakeholders won one at a time, and it takes 4.2x longer precisely because a person has to be there for all of it.
Once you see the split, serving each side diverges sharply, because transactions are won on speed, not salesmanship. The classic MIT study found that reaching a buyer within five minutes makes them 21x more likely to qualify than waiting half an hour, yet only 23% of B2B companies respond within five minutes, the average drags on for 42 hours, and 78% of buyers just buy from whoever answers first. Now add that buyers increasingly show up after hours, 40% in the evenings and nearly a quarter on weekends. A buyer with a pricing question at 11pm will not wait for a rep's calendar, and that is the whole case for an always-on agent like RepX on the transaction side: it answers in the moment, every time, while your reps keep the read-the-room judgment for the relationships that actually need it.
Finding 3: Buyers who came through RepX closed at 18.9%, nearly 4x the industry norm
This is where it stopped being a theory about our own funnel. Hospitable, a property-management software company, had this exact problem, where a one-property host demanded the same discovery call and demo as a fifty-unit account. They put RepX on their website, named it Hospi, and let it run the transaction end of the funnel: greeting visitors, answering questions, demoing the product live in the chat, and routing the ones ready to talk.
The last number is the one that matters. Leads that came through a RepX conversation closed at 18.9%, and for context, a typical B2B SaaS lead converts to a customer somewhere in the 1% to 5% range. Buyers who had a real conversation with the agent closed at several times the rate of an ordinary inbound lead, which is a strange result for an automated top of funnel until you look at what the conversation was doing.
RepX drops interactive product demos straight into the chat, so buyers watch the product work while they decide rather than reading about it. But the detail that stuck with me is what Alan Sincich, Hospitable's Head of Product Marketing, found in their HubSpot records: many of the AI-sourced buyers were booking sales meetings before they had even started a trial, backwards from the usual product-led sequence. The conversation had quietly become the qualification step. A buyer would trade a handful of messages with Hospi, get their pricing answered, see a live demo, get routed by property count, and book straight onto a sales calendar, so that by the time a human joined it was less an intro than a closing conversation.
"The conversation kind of builds trust while you're exploring our tool for the first time. If I was an outside user, I can actually see what she's talking about live in the conversation." — Alan Sincich, Head of Product Marketing, Hospitable
That is the mechanism underneath the number: an agent that runs the transaction well does not just save rep time, it hands the human a pre-warmed, self-qualified buyer, so the meetings that do reach a person are the ones that were always going to close.
What it adds up to
The reflexive objection is that buyers want to talk to humans, and they do, for the right questions in the right deal. But hold that next to the paradox we opened with, where buyers ask for a rep-free experience and then regret it 1.65x more often when they get one. Buyers are not wrong to want self-service; they are underserved by the unguided kind that leaves them to work it out alone. An agent like RepX that answers instantly, demonstrates live, and knows when to hand off closes exactly that gap, giving the transaction buyer the fast help they wanted and passing the relationship buyer to a person who can build one.
Our 1,167 deals, the industry's decade-long trend line, and Hospitable's deployment all point the same way. The transaction, small, fast, and self-directed, is a job software now does better than any rep team, while the relationship, large, slow, and consultative, is the one thing you never want to automate away, and it is where the reps freed from the first job can finally spend their time. The question is no longer whether to put an agent in front of your buyers. It is which of your deals are transactions you are still closing by hand.