§ 01 A workshop for boards & leadership teams

Your team is making decisions with AI.
Are those decisions on solid ground?

A half-day workshop for leadership and strategy teams on how to evaluate AI claims — the ones from vendors, the ones from internal champions, and the ones embedded in the tools you have already bought. Grounded in formal research on what reasoning actually is.

Typically a reply within two working days
Format
Half-day · live · in-person or remote
Group
Up to 20 participants
Starts at
£1,500
Lead
Logician, Southampton & UCL
As featured in / affiliated with —
University of Southampton UCL Times Literary Supplement The Conversation Graham Hoare Prize 2025 Anthropic Certified
§ 02 Four situations you have probably already been in

None of these is hypothetical. Each is a specific kind of mistake — and each is avoidable, once your team has the vocabulary to see it coming.

01
Procurement

A procurement team approves a tool because its demo was fluent — without asking whether fluency is evidence of reliability.

02
Boardroom

A board signs off on an AI strategy built on vendor claims that no one in the room had the tools to interrogate.

03
Governance

A governance document specifies "human oversight" without defining what oversight means when the system's reasoning is opaque.

04
Reporting

A 97% accuracy figure is treated as evidence of readiness, with no one asking what happens in the 3%.

§ 03 What participants leave with

A durable shift in how your team thinks — not a briefing that fades by Friday.

Participants leave with a shared conceptual vocabulary for evaluating AI claims, immediately usable in vendor meetings, board conversations, and internal strategy sessions. A short reference document is provided.

  1. I.

    Distinguish robust capabilities from brittle ones

    Know which questions reliably tell them apart — and which questions vendors are trained to deflect.

  2. II.

    Interrogate AI claims with rigour rather than instinct

    Vendor pitches, internal champions, embedded tools. A shared method, not a gut feel.

  3. III.

    Surface the assumptions your strategy rests on

    Identify where your current AI roadmap depends on things the technology cannot formally support.

  4. IV.

    Frame governance around what AI can be held to

    Make accountability decisions on the basis of what a system is — and is not — formally responsible for.

§ 04 Format & logistics

Built around your team's calendar, not mine.

Duration
60–75 minutes (briefing) · 4 hours (half-day) · plus optional 90-minute follow-up at 6–8 weeks
Group size
Up to 20 participants for the workshop · up to 10 for the board briefing
Delivery
In person at your office, off-site location of your choice, or remote — equal effectiveness either way
Materials
Workshop reference document delivered post-session · written debrief on the follow-up tier
Lead time
Typically 3–6 weeks from enquiry to delivery, depending on your scheduling
Audience
Boards, strategy teams, procurement, AI governance committees, senior leadership
§ 05 Three ways to engage

Choose the engagement that fits your team's situation.

Prices exclude VAT and travel where applicable. A limited number of engagements are taken each quarter.

Single slot

Board Briefing

£1,500

A focused briefing for a single board slot or leadership offsite — not a full workshop.

  • 60–75 minutes, live
  • Up to 10 participants
  • In person or remote
  • Introduction to evaluating AI claims
  • No reference document or follow-up
With accountability

Workshop + Follow-Up

£6,500

For teams with active AI procurement decisions who want accountability after the session, not just before.

  • Everything in the Half-Day Workshop
  • 90-minute follow-up at 6–8 weeks
  • Review applied to your real decisions
  • Written debrief after the follow-up
§ 06 The facilitator
Portrait of Alexander V. Gheorghiu
Alexander V. Gheorghiu — Southampton & UCL

A logician at the University of Southampton and UCL whose research is in proof theory — the formal mathematical study of valid inference and what it means to reason. His work establishes precisely where current AI systems fall short of genuine reasoning.

AI is being adopted faster than most organisations can evaluate it. The systems do genuinely reason in some respects; the impressive demonstrations are real. But there is a precise gap between what AI does well and what it can be relied upon for — and the consequences of that gap fall on the people making procurement, governance, and strategy decisions, not on the vendors selling the tools.

This workshop translates that research into something operational: a framework your team can use, on Monday morning, to interrogate the AI claims in front of them.

Full academic profile & publications — alexandergheorghiu.com ↗

Position
New Frontiers Fellow at Southampton; Honorary Research Fellow at UCL.
Recognition
Graham Hoare Prize 2025 (Institute of Mathematics and its Applications) for the essay High School Algebra and the Limits of AI.
Industry
Anthropic Certified. Active engagement with industry on AI evaluation.
§ Free resources — try the thinking before you book
Interactive · 4 minutes

The AI Claims Diagnostic

Seven claims drawn from real vendor materials and board decks. For each one, decide whether the claim is robust, misleading, or unfounded — then see what a logician makes of it.

Take the diagnostic
Research briefing · 15 sources

Using AI Safely: A Framework for Decision-Makers

Four principles — grounded in peer-reviewed research and the philosophy of science — for distinguishing brittle capabilities, interrogating vendor claims, and building governance that holds.

Read the briefing
§ 07 Questions

The questions teams ask before they book.

  • No. It is designed for people making AI decisions, not building AI systems. The ideas are rigorous; the framing is not.
  • The workshop does not replace your strategy. It stress-tests the assumptions it rests on. Most AI strategies are built on claims about what the technology can do; this workshop gives your team the tools to distinguish the claims that hold up from the ones that do not.
  • It is — which is why the workshop is built to produce a durable framework, not a one-off briefing. Teams report using the reference document in vendor meetings months later. Measured against the cost of a single bad AI procurement decision, half a day is small.
  • Yes, with equal effectiveness. Remote sessions are live and interactive, not pre-recorded.
  • Yes. Mutual NDAs are standard for engagements that touch live procurement decisions or unannounced strategy. I'll send mine, or sign yours.
  • The framework is sector-agnostic — the examples are tailored. Tell me what's on your roadmap and the worked examples in the session will draw from your domain.
§ 08 Begin a conversation

A limited number of engagements are taken each quarter.

Fill in the form and I will be in touch within two working days. If you'd rather write directly, that works too.

Reply within two working days · Mon–Fri, UK time
Thank you — I will be in touch within two working days.
Sorry — your enquiry could not be sent right now. Please email me directly at .