The Paintbrush, not the Canvas: Closing the AI Collaboration Gap

Sixty-nine percent of businesses now use some form of AI but more than 80% report no meaningful impact on productivity. That gap between adoption and results is the central puzzle in enterprise AI deployment, and a new study from CambrianEdge.ai – presented in June before the British Parliament – gives it a name: the collaboration gap.

The finding is deceptively simple. Organizations have bought the tools. What they have not built is the connective tissue that turns one person tinkering with a chatbot into a team producing better work together.

In CambrianEdge’s survey of professionals across 14 countries, 55% named their biggest AI challenge as either solo use or the absence of any structured handoff between human and machine. Sixty-two percent had no defined process for passing AI-generated work to a human for review. Twenty-seven percent reported zero shared infrastructure and not a single collaborative tool.

“The AI tools work,” notes Harjiv Singh, CambrianEdge’s founder and CEO, in an interview. “The organizational infrastructure around them does not.” This is not a skills problem, he argues, but a design one. When teams reach for different models in isolation – one person on Claude, another on ChatGPT, a third on Gemini – every handoff becomes a place where context and quality standards leak away. Give someone an iPhone and an Android phone, Singh says, and they still need an application layer on top before either becomes useful to an enterprise.

The research, which was commended by Lord Raj Loomba, a member of the House of Lords, backs the diagnosis.

Organizations with all five of CambrianEdge’s infrastructure layers reported impact at three times the rate of those with none; teams with defined handoffs saw strong outcomes 71% of the time, against 38% for those without. The same tools, in other words, produce wildly different results depending on how the work around them is organized. All of which stays abstract until you watch it happen – so I went looking for someone actually doing it, and found Hannah Feminella.

A brand rebuilt in two weeks

Feminella is the CMO of Brownstone Social, a New York members’ club, café and studio she runs with her husband Joe. The business she describes is, in one sense, the opposite of a technology story: three spaces under one roof, all built around the increasingly rare experience of people being in the same room. “The best parts of life happen when you put your phone down,” she says. Which makes what she did next all the more interesting.

Earlier this year the couple decided to rebrand, pivoting the club – formerly First Round’s On Me – into Brownstone Social. Normally a job of that size (competitive analysis, logo, color palette, the physical design of the space, positioning, channel strategy) would have consumed months. Her team ran the whole exercise collaboratively on CambrianEdge’s platform and shipped it in two weeks. The competitive analysis alone – mapping the Soho Houses and newer clubs, what members loved, hated or shrugged at – collapsed from weeks into a day. All told, she reckons the rebrand saved her roughly a month and a half.

What made it work was cohesion, rather than the speed of any single tool. Feminella’s team is deliberately lean – four core members and three interns – running marketing across four companies at once. Before, keeping a consistent voice meant re-explaining the brand’s tone to one model, then another, then a video generator, and hoping the emotional core survived. On a shared platform she could load the brand’s DNA once, connect the team’s documents and tools, and keep everyone working from the same understanding. “They can go on and know what Hannah’s thinking,” she says.

The results, by her account, were significant. Since the rebrand, Brownstone Social has seen what she describes as a 120% uptick in membership and more than a doubling of bookings for its studio space. Some of these figures are her own estimates, but the direction is unambiguous – and she attributes it partly to speed. “Attention is expensive,” she says. “If we had taken a month or two, there would have been a wane in attention.”

“Enhance, never do”

The striking part of Feminella’s story is where she draws the line. She emphasizes that collaborative AI is not a shortcut to quality. It’s a caution that maps almost exactly onto Singh’s point about handoffs. “I would caution people not to confuse collaborative AI with automatic quality,” she says. “Sharing context makes the output more consistent, but it doesn’t remove the need for taste, or accountability, or – arguably most important – human review. Nothing goes out before it’s checked by a human.”

Her rule is almost a mantra: enhance, never do. She never asks the platform to invent a brand from nothing; she brings it the emotional core – this is how I want people to feel when they walk in – and lets it build outward. Even her copywriting works this way: She writes how she’s feeling first, then asks the AI to polish it. The machine sits in the middle of a human process, not at the start or the end of it.

When I mentioned that this was, more or less, the argument of a book that Jerry Wind, professor emeritus of marketing at the Wharton School of the University of Pennsylvania; Deborah Yao (editor in chief and founder of The AI Innovator) and and I recently wrote – Creativity in the Age of AI – she laughed.  Feminella said she was going to steal the line I offered: The mistake people make is to treat AI as the canvas, when it should be the paintbrush. By the end of our conversation she had made it her own. “It should not be the destination,” she said. “It should not be the canvas. It should be the paintbrush.”

The real bottleneck

The small case and the large study rhyme neatly. CambrianEdge’s report argues, across 14 countries, that the barrier to AI value is not the technology but the organizational design around it: shared infrastructure, deliberate handoffs, human judgment held in place while the work moves fast. Feminella, running a four-person team above a café, reached the same conclusion: The value showed up not when she used AI, but when her team used it together, from a shared foundation, with a human at each end.

“AI is moving faster than our ability to absorb it into how teams actually work. If we do not address this through infrastructure, through policy, through deliberate organizational design, we risk reducing a transformative technology to a collection of individual tools,” Lord Loomba wrote in a foreword to the report.

‘The conversations we must have now, in Parliament and in boardrooms alike, must ensure that as we shape the future of AI, intelligence remains human,” the parliament member wrote. “The collaboration, the judgment, the standards. These are ours two build.”

The temptation, watching AI improve month over month, is to keep asking what the tools can do next. The more useful question, for most organizations, is the one both the data and practitioners keep considering: not whether you use AI, but how and whether you use it alone or together. Getting that right, as Lord Loomba put it, is not a technology problem. It is a human one.