Paul Roberts

AI fails without context. Building the fix @ Edra, the context layer for the enterprise (Sequoia-backed) | ex-Palantir

Edraex-Qlikex-FrontRange Solutions (Centennial Software)

“

Every failed enterprise AI initiative I've seen shares the same root cause: the AI doesn't understand the business. The models aren't the problem.

More

The missing piece is context. Agents don't know that your Tokyo office processes returns differently. That Sarah in compliance has a workaround for that system bug. That your "standard" process has 47 exceptions that actually matter. I've spent the last five years inside these deployments. Four at Palantir, where I helped build the UK business from the ground up and watched forward-deployed engineers solve the context problem the hard way — embedded for months, brilliant, effective, but impossible to scale to every company that needs it. That's why I joined Edra (Sequoia-backed). We build the context layer for the enterprise. Edra learns from your operational data and experts, captures how your organisation actually works, and makes that knowledge available to every AI system you run. Some of the world's leading companies are already automating workflows that previously needed their most senior people. Most weeks you'll find me inside a live deployment. I write about what I see there — what works, what breaks, what surprised us. Follow along if you're trying to separate what's real in enterprise AI from what's noise. Outside of work you'll usually find me with my family or indulging my inner petrolhead. I mentor through the Social Mobility Foundation and invest in and advise early-stage AI companies.