AI spending continues to accelerate, but the ROI story has not kept pace. This week, Ray Rike and Peter Buchanan dig into the proposed fix that has the market talking: turning frontier AI labs into professional services firms through the forward-deployed engineer (FDE) model.
The role is not new. Palantir built the FDE function in 2005 to embed technical teams directly within customer environments, and by 2016, it had more FDEs than software engineers. What is new is the capital. Five ventures from the major AI labs and top three hyperscalers have committed more than $10 billion, betting that the bottleneck to enterprise AI value is not the model; it is getting that model wired into a customer's data, processes, workflows, and compliance requirements.
Ray and Peter connect this moment back to the ERP era, when SAP and Oracle needed four to five dollars of services for every dollar of software, and explain why the same people, process, and services reality is playing out again with agentic AI.
What the episode covers:
Why OpenAI's $4 billion DeployCo, with a guaranteed 17.5% investor return, is the most aggressive and most financially puzzling bet of the five
How Anthropic (Ode), AWS, Microsoft, and Google each took a different structural path, from joint ventures to capital-light partner ecosystem plays
Why the MIT 95% pilot failure stat and McKinsey's finding that two-thirds of organizations have not started scaling AI make this a real problem, not a fringe one
How incumbent consulting firms are playing every side at once to protect their AI practices
Alex Karp's argument that the AI industry broke its own business model, and the irony of him making it
What CFOs and GTM leaders should take away:
Ask any FDE partner for two or three production use cases with measurable outcomes before expanding scope
Start narrow with one win, not four or five simultaneous projects
Price for outcomes up front, before mid-project renegotiation
Model the ongoing maintenance cost, since 20 to 40% of the initial investment often goes to keeping it running, and an outside FDE team at $300 to $800 an hour is an expensive long-term maintenance line item
Weigh the neutrality trade-off honestly, since lab-backed ventures get you the deepest model roadmap access but also the deepest lock-in
Lawyer up on data privacy and IP, because these engagements can feed your proprietary workflows back into the next model release
Ray's read: independent consulting firms are the likely long-term winners, with hyperscalers close behind. The AI labs are, in Peter's words, still leaving the Shire without their full posse ready to go.
Listen to the full breakdown, and for the deeper analysis behind each big story, subscribe to the AI to ROI newsletter at ai2roi.substack.com.
Podden och tillhörande omslagsbild på den här sidan tillhör
Ray Rike. Innehållet i podden är skapat av Ray Rike och inte av,
eller tillsammans med, Poddtoppen.
