Firefish Sets Out Framework for AI Governance

London-based insight consultancy Firefish has launched a set of AI Governance Principles, overseen by an AI 'Council', and based around five foundational principles intended to promote 'safe, responsible and client-centred AI innovation.'

The initiative builds on the firm's connection with QualifyAI, a separate company launched in April 2024 by Firefish execs Jem Fawcus and Richard Owen, with a mission 'to automate qualitative research without stripping it of its humanity'. Firefish says it believes the primary challenge in integrating AI across the industry is no longer slow adoption, but 'overconfidence' in unvalidated outputs. The new framework is designed to prioritise 'the integrity, transparency, and human oversight required for high-stakes strategic work,' without stifling AI innovation.

The five foundational principles are:

• Human oversight: AI serves to augment, rather than replace, the human judgment required to interpret complex cultural data.
• Accuracy and accountability: all tools are risk-assessed, continuously monitored, and rigorously validated and fact-checked.
• Transparency and explainability: AI-generated outputs are clearly identifiable, with full visibility into how conclusions are reached.
• Data ethics and privacy: strict compliance with data protection regulation, lawful data use, anonymisation, and proactive bias management.
• Fairness and non-discrimination: systems are designed and tested to eliminate bias across protected characteristics.

Owen, who is the firm's Head of Innovation and Technology, says AI governance represents 'ongoing sense-making in a world of uncertainty.' He explains: 'You cannot predict every downstream impact of AI, but you can build an organisation that learns quickly, notices harm early, and adapts. Strong governance is what makes innovation sustainable.' CEO Fawcus adds: 'In a market full of bold claims and implied inevitability, our job is to be the client advocate. We want our clients to benefit from the advantages AI brings, but they must have the confidence that their decisions are built on a grounded, culturally informed, and safe context.'

Web site: www.firefishgroup.com .