• by Maarten Albarda , Featured Contributor, Today

Tis the season: “pumpkin-everything” time, but also,for many marketers, 2027 budget planning season. Marketing and procurement teams are sharpening their pencils with a fresh set of demands.

So far, I am seeing a clear trend toward more output for less money. The future is highly uncertain: tariffs, inflation, a lingering war, elections… none of these provide clear direction or future clarity. So marketers are battening down and planning very conservatively.

But for next year, procurement has a shiny new weapon handed to them directly by agency pitch decks: the promise of AI-driven efficiency. Over the last 18 months, agencies and media platforms aggressively marketed a narrative built on being faster, cheaper, and fully agentic. They promised automated workflows and smaller account teams. From WPP Open and Omnicom Omni to Publicis Marcel and Dentsu.Connect, every agency pitch now centers on a proprietary, "end-to-end, agentic AI" platform.

Then there is Meta Advantage+, Google Performance Max (PMax) & AI Max, Amazon Marketing Cloud & Automated Ad Suites, Salesforce (Agentforce), Workday, and Microsoft (Copilot Studio)… the list goes on and on. Procurement heard those pitches, took them at their word, and is now using those exact same claims to justify slashing agency retainers across the board.

But the entire conversation is built on a misunderstanding of productivity data.

When marketers and procurement leads push for fee reductions, they often cite headline figures like McKinsey’s widely quoted estimate that generative AI could automate work consuming 60% to 70% of employee time. But nobody spends 100% of their workday on tasks that can be handed off to AI. Even large-scale self-reported surveys reflect this reality. When you look at empirical studies, a different operational picture emerges.

A BCG/Harvard study showed consultants using AI completed specific tasks 25% faster, with 40% higher quality. MIT research demonstrated a 38% performance improvement for skilled workers on tasks within AI's capability range, with lower performers gaining even more. Microsoft’s GitHub Copilot study showed developers completing specific coding tasks 56% faster. In BCG’s 2026 AI at Work study, roughly 40% of white-collar non-managers reported saving about a full workday per week, roughly 20% of their total working time.

With these numbers in mind, procurement might go in with a 30% reduction based on agency promises. The agency accepts the cut, reduces headcount for that client -- and soon both sides realize that AI cannot write strategic briefs, navigate internal client politics, or manage complex stakeholder alignment. Operational quality collapses, the relationship sours, and both sides end up frustrated.

Fixing this problem requires shifting the negotiation from theoretical efficiency to operational scope. Financial savings are the natural result of fixing processes and scope, not the starting point.

I would recommend not asking your agency for a blanket AI efficiency discount. Instead, require an operational audit that breaks work into buckets: strategic work, repetitive implementation work, and a bucket for complex tasks that most likely requires a heavy human touch. You should see less cost for the implementation bucket, where AI actually saves time but protect the budget for strategic judgment.

Attempting to measure AI efficiency through agency timesheets is not going to give you the clarity you need. Migrate your agency contracts toward deliverable-based pricing or fixed retainers tied to specific business outcomes. Paying for deliverables forces the agency to adopt AI to protect its own margins, while ensuring you pay for the value of the output rather than the hours spent sitting at a desk, or churning tokens.