AI won't replace your marketing team. It'll replace the one doing the wrong things.

Adoption is near universal. Value is not.

The headline numbers look like a stampede. McKinsey's State of AI found 78% of organisations now use AI in at least one function, with 88% reporting regular use by late 2025. That is the part everyone quotes. The part that matters to a CFO sits a few lines down. Only about 21% have redesigned any workflows around AI, roughly 39% report a measurable EBIT impact, and only about 5.5% qualify as AI high performers. Adoption is everywhere. Return is rare.

The gap between those two facts is where most of the bad decisions are being made. Buying a licence is not the same as redesigning the work. The productivity story bears this out: bolting AI onto existing process tends to deliver 10 to 15% gains, while the 25 to 30% step change only shows up when you redesign the full lifecycle. Most firms have done the bolting and skipped the redesign, then wondered why the saving never landed.

The headcount is being cut on a story, not a result

Here is the most uncomfortable finding for any board considering an AI-led restructure. Research from Davenport and Srinivasan found that roughly 60% of companies cut headcount anticipating AI value, against about 2% reporting cuts driven by value actually realised. That is a thirty-fold gap between the story and the result. The polite term is AI washing. The honest term is that AI has become a convenient cover for cuts leadership wanted to make anyway.

The agency world is feeling it directly. Forrester predicts about 15% of US agency roles will be eliminated in 2026, after roughly 8% in 2025. The roles going are not random. They are the ones doing work AI now does at speed: first-draft copy, routine production, the repeatable middle of the process.

The cautionary tale everyone cites, told properly

Klarna is the case study held up as proof that AI replaces people. Read it closely and it argues the opposite. The company compressed its image production cycle from six weeks to seven days, ran roughly 80% of copywriting through AI, cut agency spend by about 25%, and roughly halved its marketing team. Aggressive, and on the execution layer it worked. Then it went further, replacing around 700 customer-service roles with AI, and by 2025 began rehiring humans, citing quality. The lesson is not that AI failed. It is that AI is brilliant at the repeatable and poor at the judgment-laden, and the company learned the boundary the expensive way.

Automate the work, protect the judgment

The useful distinction is not human versus machine. It is repeatable versus accountable. Production, variant generation, first drafts, tagging, reporting scaffolds: automate them. Positioning, the strategic bet, the call on what is on-brand, the read on what the number actually needs: protect them. Agentic tools are already shipping into this space, with Salesforce launching agentic marketing through Marketing Cloud Next and Agentforce in late 2025. The execution layer is being automated whether you plan for it or not. The only choice is whether you redesign deliberately or let the org chart get gutted on a hunch.

That is why the roles gaining value are the ones AI cannot stand in for. The strategist who decides what to make. The editor who owns the brand line. The senior partner who can stand in front of the CFO and tie the work to revenue. When execution is commoditised, accountability is the scarce asset. A stack of channel specialists each defending a task is exactly what AI erodes. One accountable partner who owns the number, sets the strategy and uses AI to compress the execution underneath it, is exactly what it does not. That is the shape we are built for at Ruckus.