(AI made this image, after a lot of prompting as you would a toddler, and then rounds of error-checking)

I’ve been reflecting recently on a paradox I've witnessed in my professional life versus my personal life. At a high-level it's something many of you may have witnessed: that AI can add various conveniences to one's personal life, but to date adds few meaningful, let alone cost-effective, benefits in the workplace.

What I want to discuss today is more specific, and perhaps more pernicious, for you the MarTech leader. It's a phenomenon I first started noticing towards the end of the summer and has only gotten worse since then: Most of you are getting sandwiched at work between executive requests to generate value from AI, versus many your own staff failing to find sufficient value from early prototypes, especially around Generative AI.

The View from Below

Each August my wife and I spend a New Jersey beach-house week with our Kid and a half-dozen of their friends coming down from New York City. It's a lot of fun, because they're an engaging and perceptive group of mid-20s folk with degrees in STEM or Humanities, working entryish-level positions in Publishing, News Media, Higher Ed, Law, and Consulting -- all knowledge-intensive fields.

Of course this year I asked them about how AI was impacting their workplace. It solicited a collective groan that could serve as a kind of audio track to all the survey-based evidence about failing pilot projects. Just like you, they are not Luddites. Each could foresee how AI might make them more efficient or effective.

It's just that they found workplace attempts to do this almost comically ineffective. AI classifiers generating inaccurate or simply absurd metadata that require major human clean-up. Summary routines that miss key findings. Having to swivel-chair from an AI platform output to a workplace application and back. No ability to tune. Systems that don't learn, as if stubbornly proud of their own ignorance. If this all sounds like an incompetent new co-worker you'd like to see consigned to a different department....well....that's pretty much how they described it.

Greetings from Asbury Park, NJ, where some of your employees are wondering if workplace AI is a pyramid scheme...

As someone likely closer to the action, you my reader could apply labels to many of the underlying problems above (wrong LLM, no RAG, poor guardrails, silly embedded prompts, no workflow integration, poor memory persistence, no learning loop,etc.), but remember these employees are not AI experts. They work in the real world. The larger point here is that converting these prototypes into useful applications would require substantially more and better enterprise expertise, experience, business analysis, tooling, effort, and testing. In other words: AI in production wants more of your attention, resources, time, and funds. I'll come back to this later.

The View from Above

For a daily perspective on how corporate America perceives new trends, there's no better lens than the Wall Street Journal. Scrolling through the WSJ app these days, you'll absorb myriad advertisements from tier-one consulting firms on how they can untie the knot of AI value by applying better business rigor. (Disclosure: At Real Story Group we argue this too; more about that later.)

Interestingly, when you read actual articles reported by WSJ journalists, you'll discover vignettes of C-suite whisperers from those same consulting firms advising your execs and board. A common message delivered to your leadership is that they face a stark strategic choice: obtain dramatic new valuation multiples by going all-in on AI, or face ignominy if they take a more measured approach. Sound familiar? From there, the downward pressure builds.

The MarTech Leader in the Middle

All too often you the MarTech leader get sandwiched in the middle. In theory AI -- especially Agentic AI -- could automate your content and creative supply chain, accelerate campaign production and execution, and bring more attribution clarity, all while scaling and optimizing targeting decisions.

In practice, this typically requires threading together multiple steps and types of AI, with each node fragile and unreliable in its own way. Insider experts reply, "it's still possible!" Yet you who has to make it happen increasingly perceives two things:

1. Going from sloppy AI prototypes to effective production services requires complex engineering executed over time
2. If your content, data, and governance houses aren't in order, AI will simply magnify your internal disorder, rather than correct it

What You Should Do

This all takes us to some tough conversations, and demands a more discerning approach to what and where you prototype opportunities. At Real Story Group we have some specific viewpoints on these dilemmas. First, invest in AI as a layer to ensure durability, compliance, and coherent customer experience. Then, don't over-index on GenAI when Insights and Decisioning AI may offer more value, and should drive your GenAI services in a kind of flywhweel effect.

And what about Agents in MarTech? The long-term prospect remains very exciting. Near-term prototypes around high-value, "orchestration agents" are proving difficult to scale effectively. Today RSG's private Council of MarTech Stack Leaders is meeting together to share some wisdom on Agentic AI.

In the meantime, feel free to reach out to RSG for feedback on your team's efforts. At this still-early stage, we want to see you directionally on the right path. At least know that, if you're getting sandwiched here, you're not alone....