Track, Assess and Classify Risks: Keir Henderson on Responsible AI
Keir Henderson, Financial Services Account Executive at Precisely, joins CX Insider to explore the fast-moving risks of artificial intelligence. The discussion covers regulatory developments, bias, unreliable outputs and why robust governance is essential when customer-facing decisions use complex models.
The episode at a glance
- AI oversight should identify risks before tools are introduced into sensitive decisions.
- Keir discusses the EU AI Act and the challenges of cross-border regulation; specific requirements should always be checked against current law.
- Bias and hallucinations can undermine customer trust unless human oversight, data quality and testing are built in.
Reach people shaping the future of customer experience.
Put your brand into the conversations CX leaders are already watching and listening to—through sponsored episodes, video and social content.
Enquire about CX Insider sponsorshipWatch or listen to episode 129
Watch on YouTube
Listen on Spotify
Four insights from Keir Henderson’s episode
Regulation follows the risks of deployment
Henderson discusses the regulatory landscape facing organisations that develop or use AI, including the EU AI Act and implications for UK businesses operating in the EU. The strategic takeaway is to track obligations by market, use case and deployment role rather than assume a single universal rule applies.
Model bias can repeat and amplify unfair outcomes
The conversation raises the concern that biased inputs or historical patterns can produce distorted outputs. In customer-facing decisions, this can affect access, prioritisation or perceived fairness. Risk assessment must therefore look beyond average accuracy to differential impact on user groups.
Hallucinations are a governance issue, not merely a product quirk
Henderson describes AI outputs that may be incorrect or out of context. Where a customer relies on an explanation or recommendation, that failure can become a trust and compliance problem. Appropriate validation, escalation and clear boundaries for automated answers are essential.
Classify decisions before choosing controls
One way to turn the episode into action is to inventory AI uses and separate low-impact assistance from high-stakes decisions. Higher-consequence systems warrant deeper documentation, monitoring and accountable review. The objective is practical risk management that keeps pace with implementation.
A practical takeaway
Create an AI use-case register with decision impact, data sources, affected customers, monitoring cadence and a named accountable owner. Review legal obligations with specialist counsel instead of treating podcast discussion as legal guidance.
ACF Banking Solutions
Explore connected banking journeys and operational visibility as a practical context for responsible AI and customer service decisions.
Explore ACF banking solutionsExplore ACF Technologies
Find practical ways to improve customer and operational journeys.
Explore ACF TechnologiesThree related CX Insider episodes
Want your brand featured on CX Insider?
Reach an audience interested in building better experiences.
Start a sponsorship enquiry