At Operator Collective's Spring Gathering, Cait Keohane, Chief Customer Officer at Airtable, led a roundtable on one of the most pressing questions facing customer-facing leaders today: how do you guide your customers through an AI transformation when your own organization is navigating the same journey?
The maturity gap is real, and the range is wider than you think
Walk into ten enterprise conversations right now and you'll find companies that have already rebuilt entire workflows around AI sitting alongside companies still figuring out where to start. One leader at the table described a Fortune 50 CIO who had gone through the full arc in six months: told her team to build or die, watched them build things nobody wanted to support, went to vendors who couldn't handle enterprise requirements, and ended up back where she started.
The range of maturity is genuinely enormous, and the implication for customer-facing teams is that "meeting customers where they are" now requires a far more sophisticated read than it used to. As Cait put it, you walk into a conversation and there's a wide breadth of maturity around AI: some companies are already power users, others are at the very beginning of their journey and not quite sure where to start.
Partners can build. Transformation takes more than that.
Several leaders at the table had reached the same conclusion independently. The partner ecosystem isn't the problem. Partners can execute. But real AI transformation requires workflow redesign, change management, and a feedback loop back to the product team that most implementation partners simply aren't structured to provide. Those are different skills, and the gap between them is where customer transformations stall.
At Airtable, partners have historically functioned well as builders who augment the workforce. But guiding an enterprise through a genuine rethinking of its processes from the ground up is a different engagement entirely. As one voice at the table put it, automating a process that was already broken doesn't constitute transformation. It just makes the broken thing faster.
Another leader at the table had gone furthest in solving for this. She described building an incubation model where her team goes and does the work directly for the customer on specific workloads, and is seeing a 12x ROI compared to routing those customers to partners. Critically, it also creates a feedback loop with the product team that partners simply can't provide. She was candid that it caused real tension with the partner ecosystem: partners felt like she was eating their lunch. But the results made the case.
The emerging answer for several companies at the table: pull certain workloads in-house, at least to get customers through the early stages, and be deliberate about which partners are actually equipped for transformation work versus pure implementation. That line is sharper than it looks from the outside.
Rebuilding the customer success role
If there was a theme running through almost every company represented, it was this: the old customer success model, managing the relationship and routing the technical questions elsewhere, has run its course. What's replacing it looks different at different companies, but the direction is consistent.
At Airtable, Cait has evolved the CS team toward what she calls Technical Success Managers: deeply technical operators who are embedded in customer environments, carrying the change management and governance work that used to fall between teams. The early results are encouraging. Sales teams are pulling them into more engagements, and as Cait put it, they're not closing deals, but they're doing the heavy lifting that makes deals worth closing.
Other leaders at the table are solving this differently. One company has built hub-and-spoke teams pairing workflow designers with hands-on-keyboard implementers at roughly a two-to-one ratio, on the observation that workflow design consistently takes longer than the technical implementation itself. Another described an incubation model where her team goes and does the work directly for the customer on specific workloads, and is seeing a 12x ROI compared to routing those customers to partners.
One of the investors at the table noted that the profile of who needs to be in front of customers is changing. The job used to be about retention and relationships. Now it requires someone who can walk in, quickly understand what transformation actually means for that business, and drive toward it. That's a different person with a different skill set, and most organizations are still figuring out where to find them.
The incentives problem nobody has fully solved
The roundtable got candid on a question that most companies are still circling without landing on: how do you incentivize teams to actually use AI tools, and how do you know when usage is translating into real value? Leaderboards, spot bonuses, AI champions programs, and usage telemetry all came up, and none of them felt fully satisfying as standalone answers.
Token consumption as a metric has a real flaw: someone can consume 50,000 tokens and still underperform on every business metric that matters. Teams are also often less efficient during the learning curve before they become more efficient. Figuring out how to credit the people doing the hard early work of experimentation, whose value won't show up in output metrics for months, remains an open problem.
Operator Collective's own research, shared during the conversation, found that the majority of senior operators report having no formal KPIs for AI implementation at all. Among the organizations that do have a framework, it tends to map onto either a sense of productivity, an efficiency measure, or raw usage, all of which have limitations that the table surfaced in some detail.
Redesign first, then automate
The sharpest point of agreement across the table: AI applied to a broken process produces a faster broken process. The goal isn't to automate the existing workflow, it's to redesign it first, then figure out where AI fits into the redesign. Cait brought a framing from Airtable CEO Howie Liu that landed with the group: it would be wasteful to use AI to recreate processes that were already cumbersome. You have to step back before you build.
That's a harder ask than it sounds. It requires someone in the engagement who can hold a vision of what's possible while doing the unglamorous work of change management. That person might sit in customer success, in professional services, in a close partner relationship, or in a role that doesn't quite exist yet at most companies. But it's not a feature of the product, and it can't be delegated to whoever picks it up by default. For companies serious about helping their customers transform, building that capability has to be a deliberate investment.
This article was adapted from a roundtable discussion at Operator Collective's Spring Gathering, held under Chatham House rules. Insights are shared without individual attribution.
