Enterprise AI · Agentic systems · Europe & UK
AI agents in production. Not another pilot.
I design and ship AI agents for European enterprises. Sixty to ninety days from stalled pilot to production. Founder of Helionics. Twenty years at Shell, Freshfields and JP Morgan.
Confidential. No pitch. Two new engagements per quarter.
Services
Every engagement ends in production. Pick where we start.
AI fails in execution, not theory. The gap between a working model and a working system is governance, workflow integration and ownership. That gap is where I work.
Flagship
Agentic AI Build
60–90 days
- AI agents that do real work, mining email and document archives into structured, verifiable records
- Human-in-the-loop by design, recommendations not autopilot, nothing ships without sign-off
- Agent architecture, workflow integration, production rollout
- Governance your risk team and the EU AI Act will accept
Before you invest
AI Opportunity Sprint
10 days · €15,000 fixed
- Scored AI use-case backlog, ranked by value and feasibility
- Risk analysis across data, governance and operational readiness
- Two execution candidates with implementation roadmap
- Executive pack for the investment decision
When delivery stalls
5-Day Delivery Reset
€12,000 · fixed
- Root-cause diagnosis, the 2 to 3 blockers actually stalling delivery
- Stakeholder alignment map, who decides what by when
- Board-level decision memo, what to do Monday morning
- If I cannot diagnose the issue, you pay nothing
Evidence
Four programmes. Four hard numbers.
Flagship · Founder proof
My own company reads 500,000 emails and proves what happened
Every claim I make about agentic AI, I have had to defend in my own company first. At Helionics, Helios reads the purchase orders, certificates and deviation emails an industrial supply chain already produces, and its AI agents turn them into evidence an auditor can check: one tamper-evident record per physical asset.
Email mining. Proven on a public corpus of 500,000 real emails. A pattern problem, not a volume problem.
The intervention board. A dashboard shows data. An intervention board tells the right person what is broken, what it costs today, and what to do about it.
Human-in-the-loop. Recommendations, not autopilot. Nothing goes out without human sign-off.
EventChain. Open-source verification layer. Every record tamper-evident, time-anchored, independently verifiable.
500k
Emails mined, public-corpus proof
60 days
Kickoff to verifiable production record
Live
Manufacturer module in production
×35
Faster record-to-invoice, pilot-evidenced
Method
How we work
01
Diagnose
Surface the real constraint. Not symptoms. Causes.
02
Design
Agent architecture, operating model, governance the EU AI Act will accept.
03
Ship
Production, not pilots. Human-in-the-loop from day one.
04
Embed
Install rhythm and control. Exit when momentum is self-sustaining.
“I do not bring a standing army. I lead yours.”
No bench. No juniors. No discovery phase that becomes a programme. I drop into your teams, align your people and your vendors around a production outcome, and leave when it holds without me.
Private advisory
Stress-test your situation before we speak.
Ask Rakesh is not a chatbot. It is a private assistant trained on twenty years of programme recovery and AI delivery. Use it to stress-test your situation, explore options, or prepare questions before we speak. No transcripts saved. No data shared. No follow-up unless you request it.
Questions
Asked before every engagement
What does an enterprise AI consultant actually do?
Takes AI from ambition to production: prioritising use cases by value and feasibility, designing agent architectures and operating models, building governance that regulators and risk teams accept, and shipping working systems into real workflows. I do all four. I founded and shipped a production vertical-AI platform, and led GenAI to production across 2,500+ lawyers in 28 offices at Freshfields.
How long does it take to get AI agents into production?
With a scoped workflow, a decision-maker in the room and human-in-the-loop governance: 60 to 90 days from kickoff. At Helionics we deploy in 60 days. At Freshfields, adoption went from 40% to 80% in six weeks once the approach shifted from tech demos to workflow-based rollout.
Do you work across Europe and the UK?
Yes. I am based in Amsterdam and operate as TwentyThree Projects B.V. across Europe and the UK, on-site and hybrid. Engagements are contracted as outcome-based consultancy: fixed scope and deliverables, not day-rate capacity.
What is the 5-Day Delivery Reset?
A fixed-scope diagnostic for enterprise programmes under pressure: €12,000, five working days, no overruns. Root-cause diagnosis, stakeholder alignment map, board-level decision memo. If I cannot diagnose the issue, you pay nothing.
How is this different from hiring a big consultancy?
I do not bring a standing army. I lead yours. No bench to keep billable, no discovery phase that becomes a programme. I drop into your teams, align internal people and external vendors around a production outcome, and exit when momentum is self-sustaining. Two new engagements per quarter, by design.
What does an engagement cost?
Two fixed-price entry points: the 10-day AI Opportunity Sprint at €15,000 and the 5-day Delivery Reset at €12,000. Agentic AI Build engagements are scoped against the value at stake, typically €60,000 to €150,000 for 60 to 90 days, contracted on fixed scope and milestones rather than day rates.
Can you help with EU AI Act readiness?
Yes. Risk classification, governance design and compliance roadmapping are part of every engagement. High-risk system obligations land from August 2026, and governance that exists only as a slide deck will not pass. I build governance into the delivery pipeline, not alongside it.
Fifteen minutes. No pitch. An honest answer.
Confidential, and it stays that way. You leave with three answers: what is really blocking delivery, the fastest path back to control, and where AI earns its keep. If I cannot help, I will say so.
Currently accepting two new engagements per quarter.