Where business reasoning gets lost on the way to AI
Business defines the need. Transformation formalises it. Architecture translates it. AI and data teams build it. At every handoff, part of the reasoning disappears.
Perhaps the handoff between business expertise and AI expertise is where part of the AI value gets lost. Productivity gains and successful pilots are everywhere. Yet fewer than four in ten companies report an EBIT impact from AI, while only 26% have embedded AI into broader business transformation.
One reason may be organisational. Business defines the need. Transformation formalises it. Architecture translates it. AI and data teams build it. At every handoff, part of the business reasoning can disappear: the nuance behind a judgment, the exception an expert never documented, the context behind a decision.
A dual capability in one person
This is why I believe a specific dual capability may need to sit in the same person: someone able to extract how experts actually think, interpret and arbitrate, while understanding AI deeply enough to abstract and redesign those mechanisms into human-AI systems.
The capability sits directly on the business mechanism: understand it, abstract it, redesign it with AI, then carry that intent through execution with technical teams.
The market may already be moving in this direction. OpenAI and Anthropic are expanding forward-deployed and applied AI roles, while Palantir addresses a related challenge through its Ontology.
Small teams close to the CEO
My hypothesis is that companies will build small teams of these profiles close to the CEO, responsible for maintaining coherence across business knowledge, AI capabilities, decision mechanisms and performance KPIs.
Competitors can buy the same models. They cannot buy an intelligence built around a company's own business, decision mechanisms and accumulated knowledge.
McKinsey, The State of AI, on EBIT impact and embedded transformation.