A mid-market SaaS company came to me last year. HR tech, roughly 180 people total, a 40-person GTM org. AI licensing sat at 87% of the team. Reported active use was 7%.
Nine months later, usage across the GTM team was above 60%. ABM pipeline was up 34%. Cycle time on enterprise deals came down by nine days.
This is what we actually did. The framework matters more than the headline number, so I’ll be specific.
What we didn’t do
We didn’t run a company-wide training. We didn’t buy new tools. We didn’t reorganize anyone. We didn’t bring in a full consulting bench.
All four were on the table in the first kickoff meeting. We talked the CRO out of each one. That conversation took about two hours, and it was the most valuable two hours of the engagement.
What we did do
We picked one motion. Outbound into a named account list for their enterprise segment.
We picked that motion for three reasons. It was measurable, with clear baseline numbers. It was isolated to one team, so political overhead was low. The baseline was bad enough that nobody would resist changing it. Those three criteria mattered more than any strategic rationale.
The working group was four SDRs, two AE counterparts, one RevOps partner, and me. Leadership agreed to get out of the way after the kickoff. That was the single most important decision of the engagement. Executive oversight in weekly check-ins would have slowed everything down.
Rebuilding the motion in six weeks
The old motion was eight sequenced touches, generic messaging, no account context, no signal-based prioritization. SDRs were hitting 80 dials a day and getting a 1.4% reply rate.
The redesigned motion started with signal data. We pulled intent signals and fit scoring into a Clay workflow. An account only moved into the sequence when its score passed a threshold. An AI step drafted the first touch using account research pulled from LinkedIn, recent press, and their last earnings call. The SDR reviewed, edited in 90 seconds, and sent.
We cut the sequence from eight touches to four. Each touch had a specific job. Trigger reference. Problem framing. Proof point. Meeting request. The SDR’s role shifted from “write and send” to “review, improve, and decide.”
From baseline to week six, reply rate moved from 1.4% to 3.8%. Meeting conversion moved from 11% to 18%. Daily dial count dropped from 80 to 35, which the CRO hated at first. He stopped hating it around week four when the meetings started stacking.
How the team actually worked differently
The SDRs started spending the time they used to spend typing on two-call research. Actual account prep. Reading the product page. Finding the internal champion. Mapping the buying committee. None of that was happening in the old workflow. Not because they didn’t want to. Because the old workflow didn’t leave room.
The AEs started getting better briefs. The RevOps partner started getting cleaner data because the sequence only fired on qualified accounts. The sales engineer started delivering better demos because the first sales call was better prepared.
None of this was about AI, and all of it was about AI. The model made deep account research tractable at scale. Without it, the SDRs couldn’t have personalized 35 touches a day. They’d be back to mass sequences in a week.
The adoption number came for free
We never ran an AI adoption program. We rebuilt the motion around an AI component. The SDRs used the tools every day because the tools were the job. Usage went up because the work changed.
That’s the only way I’ve seen usage numbers move durably. Adoption programs run in isolation produce short-term spikes and long-term drift. Workflows rebuilt around the model produce usage that doesn’t need to be pushed.
Scaling the pattern
Twelve weeks in, we applied the same shape to content production for the enterprise segment. One motion, one owner, one metric.
A month after that, we applied it to event follow-up. Same pattern. Pick the motion, baseline the number, rebuild the workflow assuming the model carries a piece of the work, measure at ninety days.
By month nine, four motions had been rebuilt, and aggregate AI usage across the GTM team crossed 60%. The usage number wasn’t the goal. It was the byproduct.
What broke along the way
Two things, both preventable.
Middle managers didn’t know how to coach the new workflow. Their playbook was “pound the phones.” Their reps were running a workflow the manager couldn’t coach. We lost a month retraining the frontline managers, which should have started in week one. If I could do the engagement again, the manager retraining would be parallel to the rep retraining, not sequential.
Marketing didn’t move at the same pace. The SDRs were operating in a new world. Marketing was still producing campaigns on the old calendar. The content wasn’t matching the motion. We had to run a parallel content redesign to catch up, which added six weeks to the timeline.
If I did it again, I’d start outbound and content at the same time, with the same metric layer across both.
What the CRO said at the end
He pulled me aside at the final review and told me he’d have spent three times the fee and still been ahead. I don’t usually quote client sentiment in case studies because it reads as self-congratulatory. I’m quoting it here because his follow-up comment was more useful.
He said the thing that surprised him was how boring the work ended up being. He’d expected big strategy, ambitious transformation, executive alignment sessions. What he got was a forensic, workflow-level redesign, week by week, with one team. No rollouts. No all-hands. One motion, then another, then another.
That’s the real pattern. Boring on the outside. Expensive math on the inside.
The takeaway
The company didn’t go from 7% to 60% AI usage by running an AI program. They got there by redesigning four motions, one at a time, over nine months. The usage number followed the work.
If your team’s AI usage is stuck in single digits, stop staring at the usage dashboard. Look at whether a single motion has been genuinely redesigned. Until something has been structurally rebuilt, nothing will move.