Meet the AI Builders #15 — Thomas Spitz — CEO & co-founder @ AI Partners
AI is easy. Adoption is not.
Most conversations about AI obsess over what models can do.
Thomas Spitz is obsessed with something else: why organizations still fail to use their potential at scale.
Today, he’s the founder of AI Partners, a firm that made a deliberate choice: focus exclusively on generative AI and agentic AI—nothing else.
But his story doesn’t start with large language models (or even with code). It starts with data—and a shift from being a consumer of algorithms to working behind the scenes of how they shape behavior.
“I realized I wasn’t just seeing ads online,” he told me. “I was helping decide which ads people would see—and why.”
🔥 Introduction: From data pipelines to behind-the-scenes algorithms
Thomas says he’s “always done data,” but one early milestone still anchors his worldview: working in the innovation hub at Warner Bros. Entertainment Group of Companies as a data project manager.
There, he worked on a Data Management Platform that connected campaigns across platforms like Meta and YouTube—with a very tangible outcome: smarter tracking and retargeting, and a big lift in conversion rates.
But the real impact wasn’t just performance. It was perspective.
He describes it as a “crossing”: from being a user of the internet to becoming an actor who understood what happens “behind the scenes”—how the ads everyone sees every day are decided.
After a consulting stint at Accenture, he joined Tradelab—now Jellyfish—right as the company entered hypergrowth.
When he arrived, the company was around 60 employees. When he left nearly six years later, it was closer to 3,000.
He grew with that curve, eventually becoming Director of Data & Analytics, leading a ~20-person team across three countries on topics like measurement, scoring, algorithms, and audience modeling.
He also worked internationally—he even name-drops projects spanning places like Cape Town and Rio.
And then, at the end of 2022, came the moment that triggered everything.
⚡The ChatGPT “slap”
Thomas first tested ChatGPT to prepare a pitch.
“The output was better than what I would have done myself,” he said. “That was a slap in the face.”
Three months later, he resigned. Not because he feared being replaced — but because he realized that the real challenge wasn’t technology anymore.
It was transformation. He tried launching an internal GenAI task force. The inertia was massive. Security concerns, organizational friction, decision paralysis. That failure clarified his next move.
🛠️ What Thomas is building at AI Partners
AI Partners was founded with a radical constraint: only generative AI, only agentic AI—no side offerings.
The goal is equally specific: operational productivity and efficiency through real adoption—not demos, not “AI theater.”
Their work rests on three pillars:
Organization — redesigning processes so AI can actually fit
Competence — building skills and reflexes across the company
Technology — connecting data and building agents inside workflows
A key point in Thomas’ method: technology comes last—because you can’t automate chaos.
They often start with audits and roadmaps: identify use cases, define deployment paths, and build internal capability so clients don’t stall after the first “pilot.”
He also has a practical filter for “good” use cases. He looks for tasks that are:
time-consuming
high-frequency
low decision intensity
Those are the sweet spots where assistants and agents can create immediate value—without requiring heroic change management.
And yes, AI Partners is a technical shop: 12 out of 18 people are data scientists/engineers.
But Thomas keeps bringing the conversation back to humans: training, leadership alignment, and the reality of day-to-day workflows.
He has a strong thesis here: start by training leaders, because strategy and resources get decided at the top. In practice, that becomes workshops, webinars, hackathons, and e-learning—often packaged as concrete “top 10 use cases” by function (marketing is a common entry point).
🧩 AI Challenges & Pain Points: the adoption paradox
When I asked Thomas about the biggest blockers, he didn’t talk about model quality. He talked about people.
His adoption pattern is blunt—and familiar: 10% early adopters, 70% cautious middle, 20% resistant.
The biggest blocker, in his view, isn’t a technical gap—it’s a shared understanding gap.
“Beyond skills, you need circles of discussion,” he told me. Because AI isn’t just a tool: it triggers political—and even philosophical—reactions inside companies. Everyone has an opinion, especially those who are skeptical. So you have to give them space, listen, and get them talking, instead of trying to push adoption through.
For Thomas, real adoption only works when it’s driven from both directions: top-down (a clear vision, clear objectives, and real resources—often backed by someone at the COMEX) and bottom-up (use cases that emerge from the field, where the work actually happens). And to connect the two, you need a clear owner—someone who “holds the topic” and stays agile and close to the ground, because use cases won’t be designed in the boardroom.
And then there’s the security paradox—one of the sharpest points he made:
Companies that delay adoption “because of security” often end up with more security issues, because people use tools anyway—Shadow AI—outside governance.
🚀 A concrete use case: augmenting sales, not replacing it
One project with Cuisine Schmidt captures AI Partners’ approach perfectly.
They started with training, then the client wanted to go further—specifically into improving the sales process with AI. Instead of building from a slide deck, AI Partners spent two full days in-store, selling kitchens, to understand the job, the brand, and the workflow.
That immersion surfaced an important operational truth:
Sales happen in two steps.
The first appointment is discovery and information.
The second is where conversion happens.
And preparation for that second meeting is critical.
But sellers juggling many first meetings struggle to retain details, write clean notes in the CRM, and prepare consistently—especially juniors. Seniors often don’t need much help; they’ve built intuition over years.
So the solution was augmentation.
AI Partners built an agent embedded in the workflow: it asks the salesperson 7–8 structured questions, then generates guidance for the second appointment—talking points, next steps, and even elements like the customer’s temperament.
The intent is explicit: support the salesperson, not replace them.
💭 Reflection: what changed in 2025 — and what 2026 will demand
Looking back, Thomas sees a clear evolution in how companies approach AI:
2024 was the year of training — understanding the tools, building basic literacy.
2025 became the year of audits and RAG — stepping back, mapping workflows, and grounding AI in real company knowledge.
2026 will be the year of agents — and the moment where ROI becomes unavoidable.
He’s also unusually direct about the current state: in many organizations, adoption is still shallow—despite all the buying, selling, and FOMO. There are few standards, limited measurement, and real frustration when “AI initiatives” don’t translate into day-to-day change.
That’s why he doesn’t believe AI will magically transform companies. In his view, transformation requires clear vision, organizational courage, and relentless focus on real work.
And he keeps returning to the same line—because it’s the one most teams try to avoid:
AI isn’t the hard part. Changing how people work is.
🔗 Follow Thomas
Thank you for tuning in to this edition of the AI Builders Interview series! 🎙️
🔗 Dive deeper into the minds behind AI innovation—explore more conversations with the builders shaping the future, and check out 150+ real-world AI case studies in our growing library:
✨ We hope this dialogue offered fresh insights and sparked new ideas!
🛠️ Know an AI builder doing great work? Nominate them for a future interview!
Disclaimer: This content was (obviously 😉) built with the assistance of AI.