The AI predictions I'd put my money on this year

I jot down observations about AI in my Notes app almost every day. Here are five I've thought of the most this month. Mostly about marketing.

1. Agents will converge into a super-agent that's used organization wide. The era of naming Openclaw agents and giving them cute personality quirks (though I love to see it!) will end nearly as soon as it's started. An org-wide super-agent will be implemented because 1. It'll have access to the knowledge base of all work at a company, making its inputs exponentially more useful, and 2. It will be used to measure individual employees' performance. Many teams measured exclusively on revenue like sales are already doing this. But it will seep into roles that have soft metrics too, like design or account management.

There are two ways to react to this. One is to be upset and worry that we'll end up in a Big Brother dystopia. But the other, more helpful perspective to have is, what if this helps teams better manage their time? For example, let's say that an agent documents an Art Director spending four hours in meetings this week and according to meeting notes, essentially gave the same walk through four times on their latest ad creative. Someone can now review this and say, "hmm, maybe some of these meetings could have been combined/an email" and now—four hours of meetings have gone to 1.5.

2. Agencies will experience a renaissance as strategic and critical growth partners to businesses. Companies do not hire extra people internally to sit around and think about AI. Internal staff are hired to think about the business needs of today, which means hitting goals 6–12-months out. Learning AI can often compete with that.

Advertising agencies who can implement AI ways of working to help companies hit certain milestones (then end their engagement with a handoff) will be in high-demand. But only the best will survive; companies will also need to have a high degree of trust in these agencies because of how much tools and data they will likely require for their practices to be effective. It's never been more possible to analyze data at scale. Inputs like call transcripts, sales data, emails or event data that were too cumbersome to comb through in the past can finally be analyzed an applied to actual, real-time work.

But the number of tools, MCPs and logins that make this orchestration possible is…a lot. Companies will not willingly hand these over unless they have a great deal of faith in the partnership. So agencies that position themselves as strategic, AI-builders with a handle on security are much more suited than let's say, a solo practitioner who has never worked with that many tools before.

3. Someone will take a serious crack at the brand-spend-to-impact problem, and it'll be good. I've worked on large scale brand marketing campaigns, and the question of "how do we actually measure this" always comes up. Companies know they need to make some investment, and they implement some measurement (internal models built by their data science teams, marketing research firms, incrementality testing, surveys, etc.), but it's still often unclear what amount will move the needle.

Here's how I think AI changes this. The most challenging part of brand measurement was never gathering data, it was the counterfactual. Teams couldn't see the version of the world where they didn't run a campaign. AI will model that counterfactual at a granularity traditional MMM, current consumer insights companies, or incrementality testing can't compete with.

Some internal teams with talented data scientists will be able to build this in-house. But I think companies in the data insights space will come up with models of this this year that will work very well. Better than they've ever worked before. We'll see at least one headline about a company being able to effectively and accurately do this before 2026 ends.

4. B2B companies that can make their less-technical prospects feel like the "AI guy" of their company will have a serious advantage. Here's how I think products with AI-powered features, especially for B2B SaaS companies, should be marketed: customers should feel incredibly special if they've discovered a new AI tool or more efficient way to use AI in their org. Companies who over-invest in training, onboarding and exclusive content for existing customers will have an advantage here. Lifecycle and customer marketing teams who can effectively run less polished resources like public Discord channels or private slack groups will win.

Now, are these channels already table stakes at many companies? Yes, especially those in a highly technical space that market to developers or IT professionals already. But there is room here to grow or companies who sell into marketing, sales, operations or customer teams.

5. On a philosophical note, UBI will enter the political conversation and be taken seriously this time. I can't stop thinking about AI has been trained on all our collective work. And while the share of what its trained on is proportional to impact (the more contributions you personally make, the more you've influenced the training data), all humans have likely contributed in some way. Even if someone isn't a published author, nearly everyone else has left a review of a product, posted on Facebook or even clicked a CAPTCHA. So it does seem fair that at least some of this wealth is redistributed. Or that we at least talk about it.

I can't wait to read these again in six months and see if I'm right or wrong.