My biggest prediction for the year is we’re going to see the first AI Retention Agency for eCommerce under the guise of “Service as Software”. The shift to “Service as Software” will be agencies building their own proprietary agents to complete their normal outputs at near zero manual costs.

I believe this will also push many more agencies into not just AI usage, but deeper integration of platforms and even building more of their own tools. Spoiler: I’m already seeing this as I’ve helped a few agencies do it.

To really understand the disruption of the agency model, we have to look at the fundamental tension between labor-based scaling and software-based scaling.

The traditional agency is built on a fundamental conflict of interest: The Billable Hour. In a labor-based model, an agency’s revenue is a function of headcount. To grow revenue, you must either increase the number of people working or increase the hourly rate. This creates a "diseconomy of scale". As the agency grows, its complexity and overhead grow linearly (or sometimes exponentially), while its margins remain stubbornly compressed.

More importantly, the incentive is for inefficiency. If a retention agency can set up a complex Klaviyo or Mailchimp flow in 10 minutes using a highly tuned agent, but they are billing for "Strategic Implementation" that historically took 10 hours, they face a choice:

The "AI Retention Agency" of 2026 chooses a third path: Value-Based Pricing enabled by Zero Marginal Cost. They don't sell hours; they sell the outcome (e.g., "We increase LTV by 20%" or more simply “We’re going to take on X work for you fully and save Y hours”). Because their marginal cost to execute the next unit of work—the next email, the next segment, the next A/B test—is nearly zero (compute costs vs. human salaries), over time their profit margins look closer to a SaaS company than a consultancy.

To understand more deeply what that looks like, we should consider the Smile Curve as applied to services.

We can apply the Smile Curve theory here. In the pre-AI era, value was distributed across the middle: the "doing" of the work (copywriting, technical setup, design, reporting).

In 2026, the "doing" is going to be further commoditized by AI agents. The result is value moves to the edges:

The danger is to the traditional agencies who get stuck in the "sagging middle." They are being hollowed out by software that does the "doing" better, faster, and cheaper. Hence my prediction for the AI Retention Agency.

I know what you’re thinking. “Nathan, you’re just an AI maxi. This won’t happen!”

Ahh, my dear reader, you’re only half right. We only have to go as far back as 2010 for a recent example of this in action for agencies.

To understand how this has happened before, we can look at a major historical shift: Webflow in the 2010’s (If you’re really curious, programmatic advertising is another great example).

In 2010, building a high-performance, responsive website was a Technical Wall. You needed a team of specialized frontend and backend developers to wrestle with PHP, MySQL, and the early, clunky versions of WordPress or Drupal. The "value" was in the building… the sheer labor of making code work across browsers.

Webflow (and to a lesser extent, Squarespace and Shopify) didn't just make it "easier"; they commoditized the middle. They took the technical execution, the part of the project that previously consumed 70% of the budget, and turned it into a visual interface with near-zero marginal cost.

This technical wall is precisely the work we see with emails, flows, segments, on the retention side. It’s also what we’re seeing happen with ads on the performance side.

We can map the Webflow precedent to the current AI trajectory in three distinct phases:

In the Webflow era, the winners weren't the people who "knew how to code"; they were the people who knew how to architect a system within Webflow. They understood SEO, UX, and conversion: the "Strategy" side of the Smile Curve.

Digital technology always follows the path of hollowing out the middle. Webflow hollowed out the "Average Web Developer." AI is now hollowing out the “Average Email Specialist”. The only place left to hide is at the edges: being the Platform (the AI) or being the High-Value Consultant (the human who tells the AI what to do).

1. AI creative breaks out

For the last two years, we’ve lived in the era of the "Horizontal LLM." OpenAI, Google, and Anthropic provided the broad substrate, but for the average marketing operator, these tools remained a "blank canvas" problem. The friction wasn't the technology; it was the workflow.

In 2026, we are going to see the Great Verticalization. This will lead to many ad platforms to stop trying to be "GPT for Images" and have start being "The Automated Creative Director for Performance Creative." While we’ve see marketing around this, I believe this year we’ll actually see SaaS companies verticalized their products in such a way as to where the results will folow.

This will continue the march of the cost of creative trending toward zero.

2. Segmentation finally gets real investment

Still under-invested on the email side. We just started to scratch the surface with platforms like Raleon, but there's so much room for behavioral targeting in email to get closer to what we've come to expect from paid channels. I expect segmentation to move past “engaged in the last 90 days” and into actual predictive signals, like who’s ready to buy again or who needs a nudge. 

3. Agent UX moves more to where operators already live

We’ll start seeing UX that people like, beyond a pure chat interface. Agents inside Slack, inside your ESP, inside the places operators actually spend their day, rather than forcing everyone into another interface. As a part of this trend, I believe we’ll start to see apps within apps. For instance, apps within ChatGPT, or apps within Shopify. To be fair on this one, I had the prediction, but it’s already playing out (looking at you shopify).

4. Shopping discovery shifts into conversation

The most significant opportunity to the traditional e-commerce playbook is the shift in Discovery. Historically, discovery happened on Google (Search) or Meta/YouTube (Discovery). Now, discovery is happening inside the Conversation.

I predict Shopify will both:

If you missed it, this would play quite well into Shopify’s new Product Network.

5. AI ads from mocked to just… ads

We will reach the end of the "AI Uncanny Valley" in advertising. In 2024, an AI-generated ad was a curiosity and often dumped on; in 2026, it will just be what it is… an ad.

The history of media is a history of decreasing the cost of fidelity. We moved from hand-painted signs to photography to digital video. Each transition was mocked for being "fake" until it became the standard. The tool is irrelevant; the Relevance is everything. If a synthetic ad is better targeted and more helpful than a "human-made" one, the consumer doesn't just tolerate it… they prefer it. History has shown this to be true for decades.

And that's it for this week's edition. I hope you all had a wonderful holiday. This will be a great year, and I’m excited to see how much AI continues to help brands and agencies operate more efficiently. It’s only getting easier!

Next week we’ll jump into some image generation model reviews and news!