Skip to content
upshift
Marketing Agencies

Your competitors are offering the same work for less. AI gave them the margin to do it.

Agencies facing cheaper rivals compete on business results instead of rates. PMG ties part of its fee to clients’ business targets and grew revenue 38% in 2024, according to Digiday. Accenture Song sells data architecture, platform integration and workflow redesign, and doesn’t bid for agency-of-record creative work. Neither one competes on who makes the creative for less.

Book 20 minutes

What’s happening

Omnicom is on course for a roughly 15,000-role reduction in 2026, from about 120,000 at year-end 2025 to about 105,000, with $900 million in cost-reduction synergies this year after the IPG combination (John Wren, Q3 2026 earnings; Angelastro at Communacopia, September 2026). Wren told investors in October the integration is “substantially completed.” When a holding company can produce the same work with fewer people, they can undercut independent agencies on every proposal.

S4 Capital executive director Wes ter Haar told Digiday that “AI is eating the agency business.” Headcount was 6,156 at 30 June 2026, down 10.5% year-on-year, and about 6,100 by the third-quarter update. The AI-forward agencies aren’t only working faster. They’re restructuring their cost base to offer lower prices and still maintain margins. The agencies that haven’t restructured are competing against a fundamentally different cost structure.

The pricing pressure comes from every direction. Large holding companies are using AI to consolidate and cut costs. Small boutique agencies are using AI to punch above their weight. Freelancers with AI tools are competing for work that used to require a team. The middle, full-service agencies charging full-service rates without an AI-augmented model, is the most exposed position in the market.

The newest competitor is a single operator with a stack of agents. Experienced practitioners describe building agents that let them make sophisticated calls the work used to require a team for, and they draw the obvious conclusion out loud: if one veteran can distill years of judgment into a workflow, so can others, and that’s who the agency is bidding against on the next proposal. The undercutting is no longer only the holding company with a lower cost base. It’s a lone specialist whose costs are close to zero on the work you still price as a team effort.

Where this sits: The Three Shifts

Why the obvious responses don’t work

“Match competitor pricing”

Margin death. If your cost structure hasn’t changed but your pricing has, you’re subsidizing every engagement. Matching AI-enabled pricing without an AI-enabled cost base is not competitive. It is unsustainable.

“Differentiate on service quality”

Hard to prove before the sale. Quality is demonstrated in the work, but you need to win the work first. When a competitor offers the same scope at 40% less, “we do better work” rarely survives the procurement conversation.

“Specialize in a niche”

Buys time but doesn’t solve the pricing problem. Niche expertise is valuable, but AI-forward competitors will specialize in the same niches. Specialization is a positioning strategy, not a business model.

What’s working instead

PMG ties fee portions to hitting client business targets and grew revenue 38% in 2024. Tinuiti, the largest independent full-funnel performance agency in the US, tripled revenue under a performance-driven model, managing $4 billion in digital media. Accenture Song, which hit $20 billion in revenue in fiscal 2025 (Accenture), doesn’t compete for agency-of-record creative budgets at all. It sells data architecture, platform integration, and workflow redesign. These firms aren’t competing on who can produce creative cheaper. They’re competing on who can drive measurable business results, and that’s a price conversation clients are willing to have.

The way out is to stop optimizing the old model and sell new services built on what AI can’t do. We discover what your clients will pay for, then build the AI that delivers it. Start with a pilot: in two weeks, a working first version tried on your own client work, then a decision on production. Or start with a plan for the whole firm.