# Roadmap Budget Realism ## The roadmap budget is many times too small for the unit costs of the company The roadmap has five tracks. The tracks are more domain databases, more domain-specific models, a larger agent catalog, a stronger agent workflow platform, and more ML modules. The roadmap has budgets. The roadmap has no dates. > [!danger] The numbers of the company contradict each other > The company plans to go from about 20 agents to 100 agents or more. The stated agent budget gives a cost for each agent. This cost is **a small fraction** of the stated cost to put one agent into production. > > One of three claims is incorrect. The claims are the agent count, the budget, and the unit cost. This affects one decision. The decision is how much of the investment the company needs for work on the platform layer and the agents. The other part of the decision is how much remains for expansion. This item goes to [[Open Questions and Numbers to Reconcile]]. - **Domain models**: The budget is sufficient only for parameter-efficient fine-tunes of open-weight models. - **Domain models**: This does not agree with the claim of domain-specific foundation models. - **Domain databases**: Each domain database depends on a new institute partnership. The company cannot control the timing. - **Deck-only themes**: The themes are physical AI, digital twins, and agentic AI. The budget does not include the themes. The detailed roadmap does not include the themes. - **Deck-only themes**: The deck and the Q&A are weeks apart and give different emphasis. This suggests that the roadmap is a narrative and not an operating plan. ## The chain of dependencies The chain has four steps in this order: institute partnership, domain database, domain model, and validated vertical agents. The company must also resolve the license and select the tenancy architecture before the first hosted tier. Sales of semiconductors and overseas sales cannot start without certification. > [!tip] A test for the roadmap > The test is a roadmap for 18 months, with dates and resources. In the roadmap, each new agent is a configuration of a template that the company has. The company gives examples as proof. ## Why it matters The stated investment is not sufficient for the roadmap. This shows one of two things. The first is optimism. The second is a different business model below the stated model, with templates and not bespoke agents. The answer clarifies [[Platform or Solutions - The Central Question]]. ## Related - [[Foreman MOC]] - [[Team Depth and Key-Person Concentration]] - [[Digital Twins]]