Updates

• We'd like to feature some agency leaders in a new interview series related to our Digital Agency Growth Guide. Spots are first-come, first-served, so be sure to fill this out soon if you're interested in participating.
• We're going live with our Q3 agency pulse survey next week. If you lead an agency and want a real answer to "Is it me or the market?" join Delphi and find out.
• ​We work with SaaS/PaaS companies to fund agency research. We have new research topics we would like to publish that are awaiting sponsors (2026 Salary Survey, Pricing & Packaging, Agency Profitability Report). If you know anyone, I'd really appreciate help connecting with marketers or agency-partner leads at SaaS/PaaS companies.

TL;DR

• The "all-in on AI" wave in agencies is receding. Our Delphi pipeline surveys, conversations with SaaS/PaaS teams, and agency clients are all getting more critical of how AI is being used.
• AI's "jagged frontier" never fully flattened. The valleys where AI fails just got shallower.
• AI can't replace an agency, but it can seriously enhance one. Success lies in matching the tool to the job, not forcing every task through it.
• Using AI for strategy is the real mistake. LLMs converge on whatever's fashionable, aka "strategy trendslop" (HBR, March 2026), and convergence is poison for strategy.
• Where AI belongs: structure, first drafts, research (augmented with the real thing), idea generation, coding, documentation. Keep humans on the strategic calls for your agency and your clients' work.
• Build the guardrails now. You'll out-position the flood of trendslop your competitors are about to ship.

The Mad Rush to AI Everything in Agencies is Waning.

We’ve seen this from a few angles:

• We run a quarterly pipeline survey with digital agency leaders (DELPHI), and we have watched AI sentiment shift over the last year.
• We work with marketing and product teams at various SaaS/PaaS companies, and many are questioning their “All-in on AI” stance that they established earlier this year.
• From our consulting work with agency leaders, we’re hearing that their clients have become more critical of the agency’s use of AI.

What we’re witnessing is an entire global market testing and learning the limits of a new technology. Within this testing and learning, there are pockets of right-fit applications and a broad swath of failures.

The Shape of Capability

LLM capabilities were described as having a “jagged frontier” where they would excel in certain areas and perform horribly in others.

As LLMs improved, the peaks and valleys of that frontier shifted. It was mostly a smoothing, in which many of the valleys became shallower, leading many to believe the valleys had disappeared.

They hadn’t.

We’re at a spot where AI companies, and the ones that benefit from them, have a massive incentive to overstate their capabilities.

“Our next model is so damn powerful it will topple empires! Also, here’s our S-1 for our upcoming IPO.”

I’m not qualified to evaluate the validity of their claims. Mythos may be legitimately powerful, and in a year from now, we could all be enlisting for our favorite overlord in the corpo-water wars of 2027.

The thing is, I don’t think any of us actually NEEDS to validate their claims right now. Clients still have aspirations that their internal teams can’t cover, and while that may change in the future, currently, AI isn’t at a point where it can replace an agency.

But, AI IS at a point where it can significantly enhance an agency. We just have to be aware that we have a hammer that must find a nail, not a screw.

Using AI for Strategy is Dumb AF

This has been obvious to most of us for a while, but now there’s data to back it up.

It’s fundamentally an issue of convergence and confidence.

We’re all well aware of how confidence can mask competence, so we’ll focus on the convergence issue here.

Even outside of LLM usage, convergence is a huge issue for strategy teams.

I’ve been glued to the World Cup lately, and because of that, my wife has seen more soccer in the last week than she ever has. She commented on how cool it was that FIFA had the players in pink shoes, unaware that it was a brand decision rather than the overall event colors. In these games, a huge number of players are wearing bright pink cleats. They’ve essentially become the unofficial uniform of the 2026 World Cup, with nearly every major brand independently releasing their own "Electric Fuchsia" colorways.

Adidas, Nike, Puma, and New Balance all predicted that the vibrant hue would help their brands stand out, as neon pink contrasts best with the green field.

In reality, it did the opposite.

Their thinking converged around this strategy, and now, everyone has pink shoes, and there’s zero brand differentiation. They spent millions of dollars on these campaigns that are now fully interchangeable.

Convergence is poison for strategy.

LLMs Have a Convergence Problem

In a March 2026 HBR article, Angelo Romasanta, Llewellyn D.W. Thomas, and Natalia Levina tested leading LLMs on common strategic trade-offs: differentiation versus commoditization, automation versus augmentation, competition versus collaboration, centralization versus decentralization, and others.

The models did not behave like neutral strategy partners.

Across thousands of simulations, the models converged and repeatedly favored the answers that sound fashionable in current management culture: differentiate, augment, collaborate, decentralize, think long term.

The authors call this pattern “strategy trendslop”: advice that sounds polished and current, but is driven more by business fashion than by the specific logic of the situation.

So is the takeaway simply “Don’t use AI to dictate your agency’s strategy?”

I mean, yes, that’s definitely it, but it goes deeper than that.

Also, please tell me you weren’t doing that…

Applying AI to Agency Work

If every agency were to dump its entire workflow into AI, the convergence issue would be as bright as Erling Haaland’s pink cleats.

This helps answer many questions leaders have about where their agency employees should use AI. The answer appears to be anywhere BUT strategic decision making.

Use AI for the structure, drafts, some research (but augment this with real research), topic ideas, coding, documentation, references, etc.

Essentially, use experts (humans) to make the strategic decisions for both your agency and for your client’s work, and use AI to augment the work around it.

This also applies to how you use AI in your agency’s growth story. We detail how agencies are applying AI inside their Revgen Systems in Chapter 9 of our Digital Agency Growth Guide.​

Everyone’s parroted the line “keeping the human touch,” but until now, we haven’t been explicit about what that means.

Use this to build some guardrails around AI use in your agency, and hopefully you’ll be able to out-compete the trendslop your competitors will undoubtedly be putting out there over the next year.

Until next time!

Leaders are Loving Our Growth Guide

You might like it too.

"As usual, excellent stuff. Easy to recognize our agency in many of the observations you make in the report. Thank you." --- "Thank you so much for all the research that you've published. I found it to be very clarifying, and I'm so grateful to have come across it." --- "I have read the summary on your website, can't wait to dig in deeper."