Marketing has brought capabilities in-house before. AI is making the case for doing it again — faster, cheaper, and at a scale previous waves couldn’t match.
The problem is what happens after the tools are adopted. Previous in-house pushes exposed hidden costs around talent, culture, and technology. This time, the test could come when boards ask whether all that AI-driven efficiency is actually producing better marketing results.
As the famous warning goes, “Those who fail to learn from history are doomed to repeat it.” Marketing is entering its third major push toward in-housing, and the first two offer a warning: The promise of efficiency can outpace the reality of building and managing the capability yourself.
The first two in-house waves came with hidden costs
The first came during the Great Recession of 2008-2009. Facing budget cuts, companies like Intel brought services, especially media, in-house to preserve budget, hiring creative talent to execute campaigns directly.
The second came during the mid-2010s digital boom. As transparency concerns grew, especially with social platforms, CMOs lost trust in their data and programmatic media partners. As internal teams’ skills grew, so did CMO confidence in them. The ANA reported that the share of members with in-house agencies jumped from 42% to 78% by 2018.
10X your SEO with Semrush for Enterprise.
The world’s most powerful SEO platform, purpose-built for Enterprise.
Request demo
CMOs assumed that competitive salaries would lure top agency talent in-house, and some did jump. But especially in B2B, creative talent got bored. Agency life offered variety: a scrappy startup one week, a Fortune 500 rebrand the next. Brands struggled to build that creative culture, and much of the talent drifted back within a year.
In-house agencies also never got a seat at the leadership table. They became order-takers stuck on repetitive production work.
Then there was the cost. CMOs saw agency hourly rates and assumed they could save millions, overlooking what it actually takes to run an internal department. Software licenses and martech stacks that agencies had spread across a whole client roster now had to be absorbed by a single company.
AI is creating a different in-house problem: proving performance
AI is changing the in-house equation. Pressure from boards and executive teams is pushing CMOs to integrate AI into everything, recognizing that work once delivered by agencies can now be done in-house. Generative tools make it easy to develop content, build ABM programs, and handle more creative internally.
Agentic AI is streamlining workflows and absorbing mundane tasks. But we’re about to reach the point where executives start asking about performance.
The efficiency case for AI in marketing isn’t in question. Content moves faster, campaigns launch quicker, and mundane work disappears. But efficiency was never what CMOs promised the board. They promised results. That’s where the third wave starts to echo the first two.
Start with performance. Duke University’s 2026 CMO Survey asked marketing leaders to rate their organizations’ marketing technology activities on a 7-point scale. No martech activity scored above 5, including “generating ROI from marketing technologies.” Adoption has outrun the organization’s ability to turn that investment into measurable results. The tools are running. The proof isn’t there.
That gap shows up sharply at the board level, too. Comviva’s 2026 Global CMO Survey found that 86% of marketing leaders had been asked to justify AI spending at the board level, while only 16% felt confident defending those investments with clear business evidence. That’s a remarkable admission: most CMOs pushing AI into their organizations can’t yet show their own leadership what it’s buying them. This isn’t a messaging problem — it’s an accountability gap opening up in real time, with boards now asking CMOs the question CMOs used to ask their agencies.
Then there’s the number that should worry marketers most. The GenAI Divide: State of AI in Business 2025 report, produced by MIT NANDA, found that 95% of organizations were getting no measurable return from enterprise genAI, despite $30 billion to $40 billion in enterprise investment.
Buried in that finding: sales and marketing absorbed the largest share of that budget, precisely because it was the easiest use case to pitch internally. Operations and finance pilots, which got less funding and less attention, actually produced better returns. Marketing got the biggest slice of AI spend and the weakest evidence to show for it.
The third wave will be judged on proof
Put those findings together, and the pattern is unmistakable. Spend is up. Confidence is down. Proof of performance is still weak. This isn’t primarily a talent problem like 2009, or a trust-in-vendors problem like 2016. It’s a proof problem — and proof problems don’t resolve themselves just because the tool gets more powerful.
Which brings us back to that warning. The lesson of the first two waves wasn’t that in-housing was wrong. Intel’s move in 2008 and the ANA’s members in 2018 both had real reasons behind them. The lesson was that moving capability in-house without solving for culture, standing, and true total cost just relocates the same failure to a new address.
AI doesn’t change that math. It raises the stakes on it. The CMOs asking “Can we prove it’s working?” are the ones who’ll avoid becoming the third case study in a pattern historians will have no trouble recognizing.