Key takeaways

McKinsey's 2025 survey found workflow redesign had the biggest effect on EBIT impact from gen AI out of 25 attributes tested. It's a correlation, not proof of cause.
Only 21% of gen AI users in that survey said they'd fundamentally redesigned at least some workflows.
BCG's 2025 study of 1,250 executives put 5% of companies as future-built and 60% as lagging on AI capabilities.
Gartner's 2026 survey of 401 marketing leaders found 15.3% of budgets going to AI while only 30% report mature readiness.
These firms sell transformation advice, so treat the numbers as direction. Their shared message, fix the workflow first, matches what I did at DemandNexus.

I keep hearing the same sentence in marketing meetings about AI: “We’ve rolled out the tools.” Usually it’s true. The licences are paid for, the training session happened and there’s a shared prompt library.

Then someone asks what has changed in how the team works, and the room goes quiet. Often the answer is nothing. Same steps, same hand-offs, same approvals, with a chatbot added at step three. The tools are the easy part.

What three surveys agree on

McKinsey, BCG and Gartner all sell transformation advice. Their surveys are self-reported and most findings are correlations. So I read them for the pattern they share.

McKinsey. In March 2025 it tested 25 attributes. Redesigning workflows had the strongest link to seeing an effect on EBIT from generative AI (an R-squared of 0.20). Yet only 21% of companies using it had fundamentally redesigned any workflow. Its 2026 report, as I read it, has 37% saying AI has helped EBIT, about the same as a year earlier. I couldn’t open that report in full, so check the figure at the source.

BCG. It surveyed 1,250 senior executives in September 2025. It found 5% of companies “future-built” and getting substantial value, with 60% lagging behind. The leaders paired a multi-year ambition with an operating model designed around people and AI working together, a talent plan and a sound data foundation.

Gartner. Its 2026 survey of 401 marketing leaders found CMOs putting 15.3% of their budgets into AI. Only 30% report mature readiness to scale it. Its analyst Ewan McIntyre warned of buying tools faster than the data, processes, governance and people needed to use them.

The bottleneck is process, roles and governance.

At DemandNexus I changed one workflow this way: the weekly research. It went from four hours to one because we wrote the work down before automating it. One workflow proves nothing general. I haven’t run a transformation at the scale these surveys describe. But the order fits.

A sequence for a social team

This is my opinion, built from those sources and my own work.

Measure the hours. For two weeks, log time by task: listening, drafting, approvals, replies, reporting. Without the baseline you can’t prove anything later.
Redesign one workflow. Remove steps and change who decides and where a person checks. The weekly listening digest is a good first choice, because it feeds every later decision.
Name the owners. An editor for voice and claims, a community lead for escalations, an analyst for measurement and listening, and one person, perhaps part-time, for the automations.
Write the rules before scaling. A voice sheet, a claims and disclosure rule, and a reply ladder for what AI may answer, draft or leave alone. Working versions are in my AI workflows for social media teams.
Decide where the saved time goes, or it vanishes into more posts. I’d spend it on fewer, better posts, more listening and more original proof.
Measure outcomes. After hours saved, look at first response time, qualified conversations, saves, shares and sales influenced. If the hours fell and none of these moved, the time went somewhere that doesn’t matter.
Review each quarter. Drop the workflows that don’t pay and widen the ones that do.

How it usually fails

Pilots with no owner and no measure.
Five tools doing overlapping jobs, with nobody sure which is the source of truth.
Saved hours with nowhere to go.
Skills gaps left alone. Gartner’s June 2026 media-spend release, as summarised in search results I haven’t checked at the source, names lack of in-house AI expertise as the top barrier for 38% of CMOs.
Governance left until after a bot has answered a customer wrongly.

The platforms are changing the work from the other side too. In paid media, Google and Meta now automate much of the buying. AI makes the doing cheap, so the value moves to deciding, checking and understanding.

Sources

Frequently asked questions

What does AI transformation mean for a marketing team?

Changing how work gets done, as well as adding tools. In practice: redesign a workflow (what goes in, who decides, who checks), define roles and governance, and move the saved time into higher-value work. McKinsey's 2025 survey found workflow redesign had the biggest effect on EBIT impact from gen AI among the attributes it tested.

Where should a social media team start with AI?

I'd start with one repeated task that eats hours, such as the weekly listening digest or the monthly report. Write the steps down with the person who does it. Run the AI version beside the old way for a month. Only widen it once the first workflow shows a measured gain.

Can you trust consultancy surveys on AI value?

As direction, yes. As proof, no. McKinsey, BCG and Gartner sell advice on transformation, and most findings are self-reported and correlational. I'd read them for the pattern they share and test the idea on one workflow of your own.

Three-minute lesson on this Build a weekly research brief Marketing course, advanced. Free, no sign-up.

More from the blog

All articles →