Last week we talked about how AI is coming for your agency’s margin. This week, I’ll offer some ways you might be able to hold onto it, or at least reallocate some of it, to serve you and your clients better. Some of this feedback came from you, the readers, for which I’m grateful.

Some of the efficiencies are load-bearing

Michelle Tresemer noted that calling it “Donkey Work” may undervalue what these rote tasks create. She reminded me of the Doorman Fallacy:

The bean counters thought they would save a ton of money by installing automatic doors. [But when they did that, the] property’s value tanked because the doorman also greeted everyone and made them feel welcome. The doorman kept transients away from the building, thus protecting its safety and value. 

While it’s easy to automate meeting notes and contact reports, those reports may be about more than reminding everyone what was said and agreed to. One value of meeting summaries is to reassure the client that they were heard, and an LLM spitting back a meeting summary doesn’t do that. 

Nick Richtsmeier has been talking about this idea by saying, “Some inefficiencies are load-bearing.” Lots of things that take time hold the roof up. When we try to “optimize,” we may destroy what matters.

The problem with removing the “load-bearing” elements is that it takes time to notice that they are gone. The costs tend to show up 6-12 months later, when we finally “miss” the thing they were holding up! Which is why identifying where the value is created, and monitoring to make sure we’re still delivering it, is key.

Where does some of this rote, repetitive work create real value, and where is it record-keeping? What risks get mitigated, or value created if humans do that work? 

Clients aren’t clamoring for cheaper advice, but they do want more value.

The Agency Core 2026 report released this month offers a counterpoint to my assertion about what clients want. They asked 400 client-side decision-makers what they want from agencies using AI. 

The answers were remarkably consistent. They don’t want AI-generated outputs dressed up as strategy. They want the human thinking they were already paying for, and they want it to be better because AI exists, not cheaper.

When clients were asked, “What do you want your agencies doing with AI?” less than a third said, “Reduce fees for us.” It was in the bottom quartile of answers. 

What came out at the top? 52% of respondents mentioned increasing the effectiveness of our marketing, improving the quality of the strategy, and generating better new ideas. 

Clients are asking their agencies to produce better thinking, not cheaper thinking. They don’t want you to undermine that with AI.

Some firm owners I’ve talked to are reallocating the savings from using AI for the Donkey Work to increase the budget for their human work. By putting more budget into deeper research, more strategic work, and more time with the client, the efficiencies gained through automation now create value and build a stronger relationship with your client.

Have you talked to your clients about automation and how you are using it? They want to know, and being explicit about what you are doing and why they can trust you builds their confidence in your recommendations. This might be the most important conversation you’re not having.

How do we make new experts?

A lot of that rote, repetitive work has another value. It builds experience and judgment. 

If a firm’s most valuable work is done by experienced folks with deep expertise, how will we build the next generation of those people? What is the role of the work that junior-level folks are doing in building their expertise? In the past, firms paid for that by billing clients for those hours. How can we give them “reps” without running up the client’s bill? This is another place where that reallocated budget can serve you and your clients.

One assumption I’ve been working under is that assigning things to AI means I won’t grow my skills in that area. Like an unused muscle, my skills will atrophy when I use AI. 

But this week I ran into research that challenges this notion. Angela Duckworth, a respected university researcher and Author of “GRIT,” and her colleagues found that, for many people, AI provided positive examples they could learn from, and it made them better. The best results came from seeing the AI sample and then writing themselves.

If the robot does the rote work, the junior person reviews and critiques the result. The feedback they get on those critiques from more senior folks creates the “reps” needed to develop the junior person’s judgment. The danger is that automation makes creating the deliverable so easy that the senior person either creates it themselves or never returns it to the junior person to improve, denying them the feedback they need to learn from.

If we choose to embrace this new tool, we have to know where the value is today and monitor our effectiveness with and without it.

Are you having those conversations in your firm? If so, hit reply and tell me what you are learning.