Four New Speakers Join AI Strategy & Practice in London
Four more speakers are joining the AI Strategy & Practice Conference on 21 October 2026 at Ham Yard Hotel, London.
The new additions bring experience across machine learning engineering, responsible AI, governance, business strategy and enterprise deployment.
Meet the new speakers
Joining the programme are:
Huma Lodhi — Principal Machine Learning Engineer at Sky
Jovita Tam — Business Strategy & Execution Operator
Phoenix Williams — Responsible AI Specialist at Lloyds Banking Group
Richard Sargeant — Founder of Agentometry and AI Advisor to BCG
They join our previously announced speakers:
Richard Masters — Head of AI Strategy at Ankura and formerly Vice President, Data & AI at Virgin Atlantic
Mira Pijselman — Managing Director, AI Governance Engineering at Ankura
Stuart McMeechan — EMEA Applied AI Engineering Leader at OpenAI
Practical questions for organisations deploying AI
AI Strategy & Practice is designed for people responsible for turning AI plans into working systems.
The conference will examine the decisions organisations face as they move from experimentation towards deployment:
Where is AI delivering meaningful value, and how can organisations measure it?
How can teams move quickly while maintaining quality and appropriate oversight?
What evidence should organisations require before deploying AI systems?
How should engineering, business and compliance teams work together?
What should organisations prepare for over the next three to five years?
One track. Honest discussion. No vendor pitches.
Limited half-price tickets
A limited allocation of half-price tickets is still available.
Use the code AISP2650OFF at checkout to receive 50% off your event pass while the allocation lasts.
Further reading from Safe Intelligence
AI safety needs professional standards and continuous validation
Safe Intelligence has welcomed the BCS’s call for more rigorous safety testing and clearer accountability in the deployment of AI systems.
As agentic systems gain greater autonomy, evaluation needs to cover the complete system: the model, its tools, its operating environment and its interactions with other systems.
Our response explains why AI safety should be treated as a continuing engineering discipline, supported by explicit requirements, reproducible testing and evidence that can be independently scrutinised.
Formally verifying the robustness of object detectors
An object detector can perform accurately on its original test images and still fail when brightness, contrast or motion blur changes slightly.
Our latest publication introduces IoUCert, a method for formally verifying the robustness of anchor-based object detectors such as SSD and YOLO.
IoUCert can prove that a detector continues to identify the correct object, class and location across a defined range of image changes. When the detector is not robust, the method can produce a concrete counterexample that helps researchers understand the failure.
