We examined AI claims with investors

AI features appeared increasingly in pitch decks and portfolio reports. Demonstrations showed successful results but rarely explained the underlying workflow. The investment team could not tell how much manual preparation was needed or who corrected faulty results.

Advisory board meetings also often lacked a shared standard. People used technical terms differently. A model integration could appear to be proprietary technology, while existing process knowledge was barely described. Participants wanted to test business assumptions without having to assess every technical detail themselves.

Our task was to guide the investment team and advisory boards through assessing AI claims using concrete documents. The goal was a shared review framework for management conversations and technical due diligence. Participants needed to recognize which claims already had evidence, which evidence was missing and when to commission a specialist review.

Before the workshop, we collected sanitized excerpts from typical pitch decks and management documents. During the session, participants alternated between management and reviewer roles. This required both sides to explain claims about data, quality and operations clearly.

We initially let a convincing demonstration stand on its own, then added a failure case. This revealed which information had been missing from the first impression. The group asked about handovers to staff, the extent of necessary corrections and updates to the underlying knowledge.

When discussing proprietary technology, we distinguished model access, the data foundation, integration and process knowledge. This helped participants ask more precisely about dependencies and reuse. A model integration received scrutiny before being treated as either unremarkable or a technical advantage.

At the end, the investment team and advisory boards each worked through a complete review assignment. We removed questions whose answers would not change a decision and sharpened the rest. Owners incorporated the framework into their existing templates for management conversations and DD preparation.

The investment team and advisory boards had shared terminology and specific questions for AI claims. When preparing conversations, they could distinguish claims, evidence and outstanding reviews. They could commission deeper technical work around an economically relevant question. The workshop provided a usable framework for further assessments.