Video Interview
Mark Hayward, Head of AI at Lowell discusses the implementation of AI within debt purchase and collections. The discussion examines organisational readiness, risk appetite, regulation, governance, business cases, employee adoption and the limitations created by legacy technology.
Rather than focusing on speculative AI capabilities, the conversation concentrates on deployment experience: automated call summarisation, quality assessment, voice AI, process automation and the need to choose between AI and simpler automation technologies. It also examines how rapid technological change affects investment decisions and why organisations may need to accept shorter technology lifecycles.
A central theme is the emerging balance between people and machines. Routine activities and straightforward customer transactions can increasingly be automated, while human agents remain important for complex conversations, vulnerability and circumstances where customers need meaningful human interaction. The discussion concludes that AI is likely to reshape processes, support functions and employee skill requirements, rather than simply replace existing contact-centre operations.
Find out more about Lowell -> Here.
Key Discussion Points
- Organisations are taking materially different approaches to AI adoption depending on their risk appetite and willingness to deploy technology directly into customer-facing processes.
- Growing familiarity with AI in everyday life is reducing cultural resistance and increasing confidence in workplace applications.
- Financial services firms need to balance innovation with regulation, retaining appropriate human control where AI supports operational or customer decisions.
- Meaningful process redesign can generate greater value than inserting AI into isolated parts of an existing workflow.
- Investment cases need to connect AI initiatives with measurable financial benefits alongside less tangible improvements in customer experience, data and analytics.
- Organisations should actively determine whether a problem requires AI or whether existing automation technologies can solve it more simply and with less risk.
- AI governance needs to build on existing regulatory and risk frameworks while evolving as firms gain practical knowledge of deployment risks and control requirements.
- Employee engagement and early operational participation are important for overcoming concerns about job displacement and generating successful adoption.
- Legacy technology architecture and unrealistic expectations of deterministic performance remain significant implementation barriers.
- The rapid pace of technology development means organisations may need to invest despite knowing that today’s solution could be superseded relatively quickly.
- Customer strategy should focus on delivering answers quickly and cost-effectively while preserving human channels for situations requiring empathy, judgement or vulnerability support.
- Over the longer term, AI could substantially reduce routine back-office and support-service activity, while customer-facing human roles increasingly concentrate on more complex interactions.
Take Aways
- AI adoption in financial services remains highly dependent on organisational risk appetite, with approaches ranging from extensive customer-facing deployment to cautious use in lower-risk internal processes.
- Cultural acceptance of AI is increasing as employees use the technology more frequently in their personal and working lives and experience improving accuracy and usefulness.
- In regulated environments, a pragmatic starting point is to use AI to support decisions while retaining human approval rather than allowing autonomous decision-making.
- The strongest opportunities are not necessarily very small tasks: redesigning meaningful back-office processes can create greater operational and customer value.
- Successful implementation favours a small number of carefully selected and controlled deployments rather than large portfolios of loosely governed pilots.
- AI investment needs a tangible business case. Financial return remains particularly important in organisations where investment must demonstrate benefits quickly, although improved data and analytics can provide additional value.
- AI should be treated as another transformation tool rather than the default solution. Conventional automation or RPA may be simpler, cheaper and lower risk for many processes.
- Governance should evolve alongside practical experience. Existing policies, Consumer Duty, DPIAs and second-line controls provided an initial framework, with AI-specific governance strengthened as deployments matured.
- Operational adoption improves when frontline managers, team leaders and engaged users participate early, provide rapid feedback and develop ownership of the solution.
- Legacy technology remains a major barrier. AI should be integrated into the technology architecture rather than bolted onto existing processes in ways that create future re-engineering requirements.
- Organisations need to accept that AI is probabilistic rather than deterministic; a solution delivering approximately 95% acceptable performance may still create substantial value where exceptions can be managed appropriately.
- AI is expected to change the mix of work rather than simply eliminate human interaction: routine transactions can increasingly be automated while human agents concentrate on more difficult, sensitive and vulnerable-customer conversations.
Innovation
- AI-generated call wrap-up: Automatically structuring and summarising customer conversations, with human approval retained before completion.
- Automated call quality assessment: Moving quality assessment away from humans manually listening to calls towards automated analysis.
- Blended transformation architecture: Combining AI, traditional automation and robotics within a single step-by-step process rather than treating AI as a standalone technology.
- Voice intelligence: Using AI for interpretation, transcription and summarisation, areas identified as natural applications of the technology.
- Dynamic customer routing: Using customer history and contextual information to anticipate why a customer is making contact and direct them towards the fastest appropriate journey.
- Real-time escalation from virtual to human agents: Monitoring what customers say and how they say it so potentially sensitive or vulnerable situations can be transferred quickly to a human agent.
- Conversational digital journeys: Replacing rigid sequences of webpages, forms and drop-down menus with conversational interfaces that gather information dynamically and generate the appropriate customer journey.
- Highly tailored general-purpose AI: Increasing capability within mainstream AI products to understand business context, produce presentations, create user interfaces and work in an organisation’s preferred style, potentially reducing the need for proprietary LLMs.
Key Statistics
- The discussion referred to AI development changing materially within approximately three to four months.
- Microsoft Copilot was reported to have improved significantly over the previous six months.
- Some organisations had been using AI-based call wrap-up for approximately 12 months.
- Human approval of AI-generated call wrap-up was described as being accepted in the high 90% range.
- Agentic AI had been a major industry theme approximately six to nine months earlier.
- The interviewee referred to deploying two or three significant AI-supported process changes rather than very large numbers of pilots.
- An alternative approach cited was organisations running around 60 AI pilots simultaneously.
- Approximately 50 business processes had been reviewed to determine whether AI or conventional automation was the appropriate solution.
- Operational roll-out examples involved moving from 10 users to 30 and then 70 once early testing demonstrated value.
- AI solutions were described as potentially being acceptable at around 95% good rather than requiring deterministic perfection.
- New AI investments were discussed in terms of potentially having an effective life of around 12 months before newer capabilities leapfrog them.
- The discussion reflected on approximately 15 years of digital change within the credit market and the continuing requirement for human contact centres.
#Lowell
RO-AR.com contact list
