Conversational AI Manager

Who you are

• A track record of owning a conversational AI service, in a high-volume regulated environment, with results you can speak to precisely: the baseline, what changed, and how it was proven
• Experience managing a delivery partner you depended on for execution, where you held someone to a commitment and kept the relationship intact
• Hands-on experimentation on live client journeys: sizing samples in advance and having abandoned your own idea when the numbers did not support it
• A data-led approach: comfortable in contact analytics, measurement design and structured root-cause analysis, and hard to sell to
• Product management craft: outcome ownership, prioritisation by value, backlog discipline, in-life iteration
• Enough technical fluency to be a peer to an engineering-minded manager: you can discuss what an API can realistically return and push back on a proposed design
• Awareness of IG’s regulatory context, including Consumer Duty outcomes and the needs of vulnerable clients, in how automated decisions are designed and measured

What the job involves

• This role owns the performance of our live AI agent, how well it serves clients today, and how much better it serves them each week. The technology is operated by a platform partner who also builds the procedures, but everything that is live belongs to this role
• Your manager owns new capability, strategy, the API integration roadmap, and expansion into channels not yet running. You own the running system: performance, optimisation, configuration changes, simple new procedures, thresholds and escalation design. You are also the platform partner’s primary day-to-day contact, the person who holds them to their commitments and reads their reporting sceptically
• This is not a vendor management role. You are accountable for how well the agent serves clients, week by week, and for proving it
• Own live performance: resolution rate is our primary metric with repeat contact and journey CSAT tracked on every change, so removed contact is a solved problem rather than a displaced one
• Own the ongoing optimisation roadmap: a prioritised backlog of configuration changes and new procedures, sequenced weekly and defended on evidence
• Diagnose underperformance to root cause using your own case sampling, referred and declined case analysis, complaint themes and advisor insight, rather than the partner’s reporting alone
• Build and run the experimentation practice: hypotheses, primary and guardrail metrics, sample sizes set before launch, stopping rules, and results that actually change what we do
• Ensure any implementations are/remain compliant with the relevant regulatory framework
• Own and confirm the eligibility rules and confidence thresholds applied to automated decisions, and justify to Compliance if required
• Act as the platform partner’s primary counterpart holding them to committed delivery timelines and escalating within their organisation when work is not moving
• Take ownership of each new capability at go-live, having set the client requirements and accuracy standards for it before build started
• By the end of year one:
• Optimisation runs as a weekly rhythm rather than a series of reactions, with a backlog that always holds the next evidenced change
• Experiments land on a regular cadence and have changed decisions, including at least one change killed because the data did not support it
• Every live threshold has a justification you can give from memory, and the escalation and challenge routes have survived audit scrutiny
• The platform partner treats you as the decision-maker and is candid about problems rather than defensive

The application process

• End Date: October 31, 2026