AI systems that reach production, and leaders who stay in control of them.

I’m Alpesh Doshi. I build agentic AI for enterprises, advise the boards that sponsor it, and speak about what I learn in both rooms.

The work

I founded Fintricity in 2001 as a data and technology consultancy. For most of its life the job has been the same: take an organisation’s messy reality and build something that actually runs on it. Big data from 2011. My first machine learning project, for Danone, in 2014.

Since then the work has narrowed and deepened. Today it is almost entirely agentic AI: systems where software takes actions rather than producing summaries. I run Kendra Labs, which builds the platform for that work.

I still write code. That is deliberate. The distance between a boardroom decision and a working system is where most AI programmes die, and it closes faster when the person advising the board can also read the pull request.

What change actually costs

The technical work is usually the easier half.

At a FTSE 250 company I led the transformation of 700 engineers and cleared a two-year delivery backlog. The technology was never the binding constraint. The constraint was people deciding whether this change was for them or against them. Every organisation I work with now faces a version of that question, and no architecture diagram answers it.

How I work

Three modes, often in the same week.

• Advisory. Boards and executive teams on AI strategy, operating model and governance. Usually a standing relationship rather than a one-off deck.
• Build. Designing and shipping agentic systems with in-house engineering teams, from architecture through to production.
• Speaking and teaching. Keynotes, panels and executive workshops. More than 90 sessions since 2022.

What I believe

• Demos are cheap. Working assets are not. The gap between a prototype and a system that survives real data, real regulation and real users is where the value sits.
• Start from the outcome, not the model. Pick the process you want to change, then choose the technology. Not the other way round.
• Humans above the loop, not in it. People should set intent and judge outcomes, not supervise every step. That has to be designed in.
• Governance is the brake that lets the car drive faster. Controls are not a tax on progress. They are what makes progress safe to accelerate.
• The people problem is the real problem. Re-skilling, role redesign and an honest conversation about what changes. Skip it and the system works while the organisation does not.

At a glance

• Founder, Fintricity (2001)
• Founder & CEO, Kendra Labs
• Host, Innovation Briefing podcast
• 25+ years in enterprise technology across financial services, telecoms and media
• London based, working internationally
• Represented by specialist speaker agencies in the UK and US