About
Boxed.ai is an early-stage company building the runtime controls and evidence layer that small and medium financial services firms need to adopt agentic AI without losing their second-line oversight.
What we believe
Most of the AI safety conversation a year ago was about what models say. The conversation that matters now is what models do, through the tools they have been handed.
Indirect prompt injection is, per NCSC, not a problem you fix at the model. The execution boundary is where the firm gets to draw the line.
Regulators do not grade firms on a curve based on headcount. Wealth managers and brokers should not have to choose between adopting AI and meeting their obligations.
It is the discipline that lets a firm demonstrate good outcomes after the fact. The product is built around it because regulators, boards and clients reasonably ask for it.
Founders
Co-founder
Background in technology consulting and internal audit / data analytics. Boxed.ai exists because the controls language he spent his career inside — segregation of duties, evidence, exception MI, four-eyes — is the language that should be governing AI agents.
Co-founder
Co-founder of Boxed.ai. Like Rittick, a BlueDot Impact AI safety alumnus. Focused on translating the defensive-layer model into a product that small firms can actually adopt without a 12-month integration programme.
Where we are
Boxed.ai is pre-funding. The founders are still in full-time roles. We are building the MVP in public, talking to design partners and applying to the incubator and grant programmes that exist to back work like this.
We are not pretending to be larger than we are. If you engage with us early, you are buying access to the founders and the genuine ability to shape the product. If that is the wrong tradeoff for your firm right now, we will tell you.
Get in touch