AI is a friend, not a foe, to in-house agencies.
By Adam Kleinberg
Earlier this week, Traction VP, Lauren Evans, and I stood in front of a room full of 300+ in-house agency leaders at the ANA In-House Agency Conference in Huntington Beach (where the sunsets are so pretty they make it into blog posts) and made a promise right in the title of our talk: AI is a friend, not a foe to in-house agencies.
You could feel the room wanting it to be true.
That undercurrent — hope mixed with something closer to dread — was the whole reason we were there. Because the fear is real. If AI can produce a brief, generate copy, design an asset, and ship a campaign, what exactly is the in-house agency for?
Here's what I think most people in that room already knew but hadn't fully let themselves believe: they're asking the wrong question.
Re-orchestration, not reorganization.
Rishad Tobaccowala wrote last week in a piece called The Coming Organizational Meltdown that the biggest challenge of an AI age will be organizational design. Lauren took it a step further: "Reorganization won't cut it. What's needed is a re-orchestration."
That's the real frame. You're not moving boxes on an org chart. You're rethinking how the whole thing plays — who does what, what gets delegated to AI, and how it all connects. Or as Loredana Crisan, chief design officer at Figma, put it in a recent Futureproof Project session after describing her eight months leading Meta's consumer AI experiences:
"Instead of building pixels for the 'Happy Path,' we are now curating experiences with data sets and evals."
Same shift, different lens. The work isn't gone. It's been re-orchestrated.
The moat in-house agencies already have.
The ANA’s latest report on the State of In-House Agencies summed up the situation: the latest threat to in-house agencies is “creative stagnation.” Companies need big ideas to stand out in an AI world as day-to-day production becomes more and more automated. If in-house teams can’t develop them, brands will look outward (pendulums do swing, kids). The best opportunity for in-house agencies to show more value is showing business outcomes — at least it is according to 37% of respondents.
These teams are just scratching the surface with AI. 49% of the in-house agency people polled said they see moderately positive impacts of AI on creative development, while 21% are still experimenting and say it is too early to tell. The ANA came to a conclusion that was almost the same as the title of our presentation:
"Generative AI is a friend, not foe, to in-house agencies.”
They were just a little bit off though: the real friend isn’t just GenAI . It’s the orchestration of agents and workflows with AI.
Outside agencies sell access to talent and tools. AI is eating that model fast ... because the tools are now available to everyone.
Any outside agency can spin up the same models, the same workflows, the same capabilities.
But they can't spin up what in-house agencies have.
The brand data. The institutional knowledge of why the last campaign failed and what the CMO actually meant when she said "make it feel more premium." The relationships that get a legal review turned in two days instead of three weeks.
Every successful company has their secret sauce. Apple and Microsoft are both very successful companies, but they are very different, in ways that live in the heads and hard drives of the people inside those walls.
That accumulated context is the moat. And outside agencies can't replicate it, because they don't have access. So while they're racing to out-tool each other, in-house agencies have the opportunity to build something those outside competitors fundamentally cannot copy.
But only if you design it purposefully.
Four steps of Marketing System Design.
At Traction, we have an orchestration solution we call Marketing System Design.
- Start with the Futureproof Diagnostic.
Before you build anything, you need to know where AI actually belongs in your operation. The Futureproof Diagnostic uses the 4P framework — People, Platforms, Partners, Process — to map that out. It goes deeper than a survey or interviews: we use actual behavioral data from your team to surface shadow AI usage and identify where agent-powered workflows would make the biggest measurable impact on how your organization actually performs.
You probably have people already using AI in ways you can't see. That's not a compliance problem. As one leader put it in the room: "Shadow AI isn't about breaking rules. It's about learning what my team needs." The diagnostic makes that visible, so you can design around it instead of fighting it. - Turn your context into skills, and skills into workflows.
Think back to the early days of your career. There was a box on your resume that said “Skills.” It said things like: proficient in Microsoft Word. Organized. Detail-oriented.
Your skills look very different now. You bring strategic judgment, brand intuition, institutional relationships, creative taste. The stuff that took years to build.
Those early-career skills are the ones you outsource to AI now. You keep the strengths. You delegate the skills. Then you turn that delegation into repeatable workflows with your brand's context — the special sauce — baked right into them. - Build it together.
Lauren made a point that landed hard in the room. When a new tool gets handed down from above — here, use this, starting Monday — people don't see an upgrade. They see a threat.
But when people build the solution themselves, they protect it. They improve it. They evangelize it. The same capability, owned differently, produces completely different adoption.
This is why we run hackathons with in-house teams. Not training sessions. Not demos. Working sessions where the team builds their own workflows around real problems they actually have. Co-creation isn't a cultural nicety. It's the adoption strategy. - Deploy a custom OS built into the tools they're already using.
We built TractionOS for ourselves — our own dose of our own medicine, an AI operating system that runs how we work at Traction. What we build for in-house agencies is different: a custom OS that runs how they work, incorporating their institutional knowledge, their brand rules, their workflows, in the platforms their teams are already in.
The goal isn't to add another tool to the stack. It's to make AI invisible inside the tools they already use — so the team isn't learning a new system, they're just getting more done in the one they already know.
Here's what that looks like in practice. One of our clients, wood pellet patio heater brand called Patiofyre, has a small team with limited resources. With their custom OS (we call it Fyrestarter), they can knock out a full e-commerce campaign — copy, creative direction, channel strategy — and push it directly to Klaviyo; or generate an AEO-optimized blog post and send it straight to Shopify; or create a collateral package like in the image above. All by answering a few multiple choice questions right inside of Claude.
The institutional knowledge of how Patiofyre markets, what their voice sounds like, what their customer responds to — that's all baked in. The AI doesn't replace the team. It multiplies them.
Experimentation Velocity: the KPI that matters most.
Lauren's measure for all of this is experimentation velocity. Not efficiency. Not headcount reduction. How fast can your team try something, learn from it, and try again?
The in-house agencies pulling ahead right now aren't the ones with the most sophisticated tools. They're the ones building the fastest feedback loops — running more experiments, learning faster, and improving the system every time. That speed of learning compounds. And it comes from building something your team owns and trusts, not something that was handed to them.
Our Marketing System Design solution can help.
The urgency underneath all of this.
We kept the talk honest, because the room deserved that. The fear of displacement isn't irrational. AI is genuinely changing what skills are required and how much headcount a team needs to do good work. That's real.
But here's what's also real: the in-house agencies that figure this out first become more valuable to their organizations, not less. They become the ones who know how to build AI-powered brand systems that actually work — because they understand the brand from the inside in a way no outside partner ever will.
The ones that don't figure it out will find themselves defending their existence against a question they could have made irrelevant.
The Futureproof Project is a community of 350+ CMOs and marketing leaders working through exactly this — together. Come build it with us or request a live demo of TractionOS in action at futureproof@tractionco.com.
Adam Kleinberg has been CEO and a founding partner of Traction since 2001. He has written over 100 articles in publications like AdAge, Adweek, Fast Company, Forbes, Mashable and Digiday and spoken at dozens of industry conferences. He's led Traction to win Agency of the Year awards from AdAge, ANA B2 Awards, CampaignUS, and in 2025, he was recognized as one of the Campaign 40 Over 40 game-changers in marketing and advertising.
Every marketer has access to the same AI tools. And here's what that's quietly doing: it's making everyone faster at producing the exact same average. It’s diluting brands — just as brand has become more important than ever. The winners of the next few years aren’t going to be the marketers who adopted AI fastest. It'll be the ones who figured out what to do with it that no one else can copy.
Most brand strategy fails before it starts. Not because the strategists aren't smart, or the brief wasn't thorough, or the research wasn't conducted. It fails because the listening stopped too early.
Yesterday, LinkedIn sent me a notice. I've been the CEO of Traction for 25 years. It feels trite to talk about how much has changed since then. In 2001, there was no Facebook, no iPhones, no Uber, no Zoom. But as I sat and reflected last night about spending the last quarter century leading this business, what came to mind wasn't the things that had changed — it was the things that stayed the same.