Pre-funding diligence for an AI startup proposal targeting generative media
A private client approached me with an early-stage startup concept proposed by a prospective cofounder. The pitch: build a generative AI platform that allows users to create personalized video content featuring real-world creators–streamers, influencers, adult entertainers–by training AI models on their likeness and mannerisms.
The idea was ambitious and loosely inspired by recent advances in generative diffusion models. But the founder's proposal lacked technical specifics and included several exaggerated claims about feasibility, especially around the speed and cost of training proprietary AI models.
My client had no technical background, so my role was twofold:
Assess the technical plausibility of the proposal–what could be done today, what might be viable in the near future, and what was pure vapor.
Evaluate the credibility of the proposer's background, architecture, and ability to execute–even if the vision were sound.
I reviewed the architecture outline, ran a feasibility risk analysis, and assessed whether any off-the-shelf models or platforms could support even a simplified MVP. I also questioned assumptions about dataset availability, latency tolerances, model training costs, and ethical compliance with platform policies (e.g. impersonation, monetization of likeness, etc.).
One red flag stood out immediately: when asked how the team would handle foundational model access, the founder replied, “If nothing works, we'll just build our own.” That alone indicated a significant mismatch between ambition and reality.
Outcome
Delivered a clear go/no-go recommendation: not technically or financially viable with current tools or team
Helped the client avoid committing funds to an overpromised, under-validated pitch
Created a short feasibility memo with recommendations for what would make the space viable in the future–and under what conditions
This kind of short-term engagement is typical of how I work with early-stage decision-makers: cutting through hype, identifying real bottlenecks, and surfacing practical paths forward–or a clean, early no.
