Building AI Middleware Beats Buying Off The Shelf For Brand And GTM Teams

A communications audit used to take a junior strategist a full week: collect materials across a client and its competitors, read everything, synthesize positioning statements, map messaging, and evaluate visual identity. That same work now runs in an hour, thanks to a tailored AI assistant. And the team that used to spend the week on process work is now free to spend it on creative strategy. That compression is happening across every phase of brand and go-to-market execution, and it changes not just how fast teams move, but what they can afford to attempt.

David Mayer is a Senior Partner and Director of Marketing and Customer Strategy at Lippincott, a creative consultancy that helps global brands define and activate their market positions. A member of the firm's management team, Mayer has spent more than 20 years helping companies build brand propositions that connect business strategy to customer experience. Lippincott is a division of Marsh McLennan.

"We're using AI on two levels. First to take process work like communications audits that used to take a week and get it done in an hour," says Mayer. "And second, to enable things we simply couldn't do before, like synthetic interviews to deepen our research." That second level is where the capability expansion gets interesting. Mayer's team now runs synthetic interviewees alongside every traditional qualitative study. The AI-generated respondents are queried in parallel with real participants, and the results are compared for thematic consistency. After 12 to 18 months of this dual-track approach, the findings are clear.

• Where it works: "In terms of thematic topic areas, synthetic interviewees are pretty spot-on with what you see in traditional research," Mayer says. The team uses them heavily in exploratory phases to identify gaps and surface angles that might otherwise be missed.

• Where it doesn't: Synthetic respondents show a consistent confirmation bias. "If you ask a synthetic to rate a brand on a one-to-ten scale, they'll tend to agree more often than what's actually in the market," he explains. "That's a nature of the LLMs." For now, synthetic research supplements rather than replaces traditional qual.

The efficiency gains compound when AI moves deeper into the creative workflow. Mayer points to GM's Metropolis platform as a model: a custom middleware layer that plugs into different LLMs on the back end, connects them to the company's creative process, and allows teams to swap models based on cost and performance. The result is programmatic advertising that automatically generates localized variations, showing a Silverado in a rural, suburban, coastal, or urban context depending on the viewer's location.

• Material at scale: "It makes a couple of percent difference on behavioral response rates, which doesn't sound like much," Mayer says. "But at the scale that you're doing it, it's pretty material."

• 100x output: In a traditional creative workflow, designers have bandwidth to create a small number of options for a given use case. With AI middleware and policy constraints in place, that output expands by a factor of 100. Designers then shift from generation to curation, becoming guardians of quality who select, refine, and contextualize the best options.

Mayer's recommendation for most clients is clear: build, don't buy. Taking a single tool off the shelf creates two risks. The first is ecosystem lock-in that becomes difficult to exit as workflows harden around one vendor. The second is subtler but more damaging over time.

• Fluency risk: "If you just buy off the shelf, you risk not becoming AI-savvy," Mayer warns. "You're just using it as a basic tool. And I think that will hurt you in the long run, as we're all going to need to understand how to use it the same way everyone uses Excel and PowerPoint." His counsel pushes back against IT teams that benchmark a single LLM for cost efficiency. "None of them are good at everything. You need to allow your teams to mix and match."

• The gap: The cultural gap between strategy and execution remains the biggest friction point. Mayer describes sitting at industry events where CMOs present ambitious AI implementations on stage, then hearing marketers at lunch ask what "Gen AI" actually means. "There's a level of cultural transformation that needs to happen," he says. "What's possible versus what's actually happening is going to take quite a few years."

Still, Mayer sees incumbents holding their advantage in the long run, even as startups move faster in the short term. CIOs at large organizations are rightly cautious about uncontrolled AI use, and that caution creates a temporary speed disadvantage. But the institutional knowledge, customer relationships, and brand equity that incumbents carry are difficult for any startup to replicate, regardless of how fast they ship.

For revenue teams worried about what AI means for their roles, Mayer offers a reframe. The mundane work, the first drafts, the process tasks, those are the pieces AI absorbs. What remains is the work that actually matters. "AIs are very smart, but they're not wise," he says. "This is going to dial up the importance of relationship and context, which is a very human endeavor. I actually feel very positive about what this is going to enable creatives to deliver."