# Claims Versus Evidence ## The documents of the company do not agree with the pitch The status of the indications follows the grading in [[Evidence Hierarchy]]. The four statuses are SUPPORTED, PARTIAL, CLAIMED, and UNSUPPORTED. The three most important results are the places where the documents of the company do not agree with the pitch. > [!danger] Contradiction 1: "Deploys in days" > The pitch says that a deployment takes days and needs no data-science team. The Q&A of the company says that each proof of concept takes months and costs a low-five-figure sum to put into production. The status is UNSUPPORTED. This result moves the platform-versus-solutions frame ([[Delivery Economics and the Scaling Ceiling]]). > [!danger] Contradiction 2: "Exclusive data in many industries" > The pitch says that the company has exclusive proprietary data in many important industries. The roadmap and the contracts show fewer industries. Partners or customers own the data. The status is UNSUPPORTED. This result moves the moat rating ([[Borrowed Data Is Not a Moat]]). > [!danger] Contradiction 3: "Core framework built in-house" > The Q&A says that the company built the core framework in-house. The roadmap shows a workflow layer on open-source frameworks. The status is UNSUPPORTED until the company gives the bill of materials for the software. This result moves the exposure on licenses and the claim of a proprietary platform ([[Open-Source Dependencies and Licence Exposure]]). ## The audit table | Claim | Status | Information | | --- | --- | --- | | Platform of agents for the manufacturing domain | PARTIAL | The company shows a diagram, demo screens, and signed contracts. It shows no code and no live demo. | | "20+ production-ready agents" | CLAIMED | The count mixes "agents/skills". Most of the indications show proof of concept only. | | Models fine-tuned for each vertical | PARTIAL | The indications cover one sector only. The company does not disclose the model that it fine-tuned. The company does not disclose a test of the model. | | Integration with MES, ERP, SCADA, and IoT | PARTIAL | The company shows ERP and MES. SCADA and IoT are not in the indications. | | Deployment on-premise, in the cloud, and hybrid | PARTIAL | The company gives no topology for each customer. | | Outcome metrics (efficiency, savings, repair time) | CLAIMED | The company gives no source and no baseline. | | 10+ manufacturing ML models | PARTIAL | The research of the founder shows that the company can make such models. The company gives no inventory. | | The company built and deployed one project, and the customer accepted it formally | PARTIAL | The indication is a contract summary only. The customer rejected a subsequent phase. | | The company filed patent applications | CLAIMED | The inspection found none. | | ESG agents monitor emissions | PARTIAL | A different answer says that the company does not measure the reduction of emissions. | | The company is a member of a start-up initiative of a GPU vendor and has a cloud partnership | PARTIAL | The initiatives are open. They are not technical validation. | ## Why it matters None of these results makes the company uninvestable. Together, the results show that the narrative of the platform says more than the indications show. The price of this investment comes from the narrative. ## Related - [[Foreman MOC]] - [[The Evaluation Gap]] - [[Open Questions and Numbers to Reconcile]]