How to Build AI Governance for Better Risks, Innovation and Return on Investment
AI governance at Innovate UK’s Bridge AI
The equitably.ai team was at the BridgeAI Annual Showcase 2026 exchanging the lastest on AI governance.
Hosted by Innovate UK with sponsors and attendees including The Alan Turing Institute, the event was a brilliant pulse check on how the UK’s high-growth sectors, everything from digital technologies, financial services, to the creative industries, are deploying AI.
Pragmatic optimism. Most of the conversations were past the “AI hype” of the last couple of years and focused on deploying securely with governance. At equitably.ai, we’ve always said that the bridge over this gap is built on professional quality competencies. Principles that not only have governance, but also context (e.g. regulated sectors), leadership (needed for change management), improvement (measuring impact), assurance (realiable output), AND governance.
The sessions were a great chance for us to share some insights on how a quality-first approach to AI actually helps get AI projects to balance innovation, risk and impact.
AI Governance: Why Does it Matter and How do we Make It Work?
The session on AI Governance was fantastic! In summary, the consensus among the experts on the panel from Lloyds Banking Group, British Standards Institution (BSI) and Kanios was clear: AI governance is the mechanism for getting a balance between innovation, risks and impact. AI governance is not just about putting in place restrictive policies but about being proactive to build internal trust necessary to fully integrate AI into core operations and maximise opportunities and return on investment.
Effective AI governance ensures AI deployments remain reliable, auditable, and compliant – particularly given growing regulations (EU AI Act and others).
The Alan Turing Institute’s AI Governance Framework
What frameworks can we use for AI governance? The Alan Turing Institute has defined a framework for AI governance based on a comprehensive survey of the UK business ecosystem, feedback from early adopters and taking into account evolving AI regulations and standards.
The Core Elements of the Alan Turing AI Governance Framework:
- Multi-Dimensional Profiling: Categorises use cases by organisation, system, data, and sector
. - Risk Taxonomy: Identifies 27 failure points across the AI value chain
- Product Mitigations: Technical “first lines of defence,” including data debiasing and adversarial training
. - Management Oversight: Institutional controls like AI Impact Assessments and Ethics Boards.
- Information Transparency: Uses “System Cards” and rigorous documentation to ensure accountability.
- Cultural Alignment: Fosters a “safety-centric” environment where staff are empowered to halt risky deployments.
- Data Governance: Strict protocols for personal data protection (DPIA) and IP safeguards
. - Human-in-the-Loop: Strategic human oversight to override or validate autonomous outputs
. - Trustworthy Procurement: Vets third-party vendors via specialised contract clauses and safety audits
. - Lifecycle Monitoring: Continuous post-deployment reviews to detect “model drift” or emergent biases
.
During the session, Arcangelo Leone De Castris from The Alan Turing Institute shared that based on experience in deployments within organisations and enterprises, the fusion of Top Down AI Governance (policies, standards, risk assessments) and Bottom Up AI Governance (technical safeguards and engineering controls) achieved the best outcomes. And that safety must be baked into AI system architecture, not just written in a policy manual.
For further information about the Alan Turing AI Governance framework, see here.
How to Build an AI Governance Team?
Lloyds Banking Group Approach to AI Governance
AI Governance Case Study: Lloyds Banking Group
Claire Schrader from Lloyds Banking Group gave an insightful overview of AI governance at Lloyds Banking Group, considered a leading deployer of AI governance in financial services.
Key elements of Lloyds Banking Group’s AI governance approach included:
Balancing Innovation: AI governance must support, not stifle. “Invisible Governance” for innovation without friction. Lloyds aims to make processes nearly invisible by automating workflows and integrating them into existing change management frameworks. This allows AI governance to scale alongside innovation rather than slowing it down.
Proactive Assurance: They utilise internal and external “red teaming” (also used in cyber security) and penetration testing to identify vulnerabilities before they reach production. This proactive assurance model also supports responsible AI governance.
Scalability: By creating repeatable, streamlined processes, they ensure that the “right thing” is done consistently across thousands of employees and diverse departments. Scalable AI governance frameworks are essential for large enterprises, like Lloyds Banking Group, who need to deploy AI across multiple business units.
AI Governance: Practical Lessons from Deployments
Practical Lessons: Lloyds Banking Group’s AI governance
Breaking Silos: They moved from fragmented, departmental practices to a collaborative, organisation-wide model.
Center of Excellence (CoE): A centralised hub that brings together experts in data, legal, security, and architecture. Many organisations now adopt AI governance centers of excellence to coordinate responsible AI development.
Iterative Frameworks: Their assurance model is not static; it is continuously updated to address emerging risks.
Single Source of Truth: Establishing a unified view of all AI models and use cases to ensure group-wide consistency.
AI Governance Best Practice
Agentic AI Governance
Standards for AI Governance
Practical Lessons: AI Governance Implementation Considerations
For firms looking to start their journey, the panel provided a clear 3-step AI governance implementation framework:
| Step | Action | Objective |
| 1. Define Principles | Identify and rank your organisation’s ethical values. | Create a rubric for making difficult trade-offs (e.g., Innovation vs. Security). |
| 2. Harms Workshops | Brainstorm “what could go wrong” for specific use cases. | Proactively map unintended uses and implement mitigations early. |
| 3. Establish a Task Force | Form a cross-functional committee (Legal, Tech, Ethics). | Centralise expertise and ensure departmental buy-in. |
Pillars of Responsible AI
Theresa Yurkewich Hoffmann from Kainos outlined a framework for AI goverannce built on:
Ethics & Human-Centered Culture: Ensuring AI supports human workflows and societal values.
Trust (Bias & Fairness): Constant algorithmic monitoring to prevent AI bias and discrimination.
Governance & Security: Maintaining operational integrity and AI sovereignty by understanding infrastructure, hardware origin, and regulatory jurisdiction.
Leveraging International Standards for AI Governance
Dr. Laura Bishop from BSI also highlighted ISO/IEC 42001: Artificial Intelligence Management System (AIMS), widely considered the global gold standard for AI governance and AI management systems. She shared how using these ISO standards for AI governance is like following a proven recipe rather than guessing and taking risks.
The AI ISO standard ensures interoperability with other ISO standards critical for AI, like ISO/IEC 27001 for cybersecurity, allowing organisations to integrate AI governance and benchmark against global peers.
Summary: AI Governance Provides Trust
Most enterprises aren’t just putting Copilot in the hands of staff anymore, but designing and deploying AI programmes with robust AI governance that provides trust. Trust for users to use it. Trust by management in return on investment with acceptable levels of risk.
During the conference, equitably.ai team discussed a whole range of AI challenges and opportunities with presenters and attendees, ranging from AI governance for return on investment to addressing risks of growing Agentic AI with AI governance to using ISO standards for regulations to capturing best practices.
We would love to share what we know from the market and our work with you. Are you ready to move your AI strategy from pilot to production? Contact us to discuss how a quality-first AI governance framework can help you.