The conversations showed that data and AI leaders are moving beyond automation by default towards deliberate orchestration by balancing cost pressure, customer expectations, and brand identity in an increasingly commoditised AI landscape.
The discussion focused on where removing humans improves efficiency and where it damages trust, experience, or brand differentiation, often quietly. This comes at a time when AI investment is under tighter scrutiny, and the conversation has shifted from experimentation to outcome-led decisions grounded in cost, risk, and customer impact.
Questions Explored
- Should we remove humans from the loop, and if so, where?
- How do we balance cost reduction with customer experience and brand perception?
- What are the risks of over-automation, and who owns them?
- How should organisations prioritise AI use cases and investments?
- What role do data foundations and governance play in successful automation?
- How is automation reshaping roles, skills, and organisational structures?
Start with the outcome and not the technology
The strongest consensus was that automation decisions should begin with a clear articulation of the desired outcome, not a drive to deploy AI. As one participant put it: “What’s the outcome we’re trying to achieve? Removing humans, or delivering a better customer experience?”
This reframing is shifting how initiatives are evaluated, especially as automation is increasingly treated as a means of delivering business strategy rather than being a strategy itself. That distinction matters, particularly when cost pressures push organisations towards automation that risks undermining long-term customer value.
There is no correct level of automation
Leaders were clear that the human versus automation debate is not binary and that it depends heavily on the customer, the context, and the moment. Routine, transactional journeys, such as changing bookings, are often improved by seamless automation. But high-value, sensitive, or complex interactions still demand human involvement and was particularly prominent in industries that handle sensitive personal data.
The real work lies in rigorous segmentation by mapping journeys to identify where automation enhances experience and where it erodes it. Without that discipline, organisations risk applying blunt efficiency logic to nuanced customer needs.
Reinvent the loop, but don’t remove it
Rather than eliminating humans, many organisations are redesigning their existing interaction model. This means using AI to reduce cognitive load while retaining human oversight. One retail example highlighted algorithm-led markdown decisions for perishables which removed manual effort but kept human execution.
This so-called “guided autonomy” model is emerging as a more realistic end state as automation handles scale and speed, while humans provide judgement and accountability.
Poor escalation design quietly destroys customer trust
A recurring frustration felt across all organisations was the failure of many automated systems to recognise when a human is needed.
“Customer frustration often stems from feeling trapped by limited bot options and escalation to a human must be easy and clear.”
This directly affects brand perception. When customers feel blocked or ignored, even efficient systems can generate disproportionate dissatisfaction. It is now the case that designing clear, responsive escalation paths is becoming as important as the automation itself.
Cost pressure is real, but so is the risk of false economies
As one leader noted: “In our world, the cost of our call centre is absolutely eye-watering… we have to make that smaller to improve.”
However, the discussion exposed tensions around over-indexing, particularly on cost reduction, as this can lead to degraded experiences, reputational damage, and higher long-term costs. Balanced scorecards that combine cost-to-serve, NPS, and operational metrics are increasingly used to navigate these trade-offs to achieve optimal value.
Brand differentiation lives in the human layer
For customer-facing businesses, human interaction remains a key differentiator. One example highlighted a deliberate choice to preserve human touchpoints as part of brand identity by using automation to support staff rather than replace them.
“Customer-facing brands see the human touch as their core differentiator; automation frees humans to do what only humans can.”
The risk is that over-automation drives commoditisation which is flattening experiences across competitors and eroding brand distinction.
AI failures are often organisational
When automation goes wrong, the issue is usually the surrounding system and not the model itself. Participants gave examples of failures being traced back to unclear ownership, weak governance, or poor alignment with business priorities, and questions such as “who owns the outcome?” remain unresolved in many organisations.
There was broad agreement that accountability should sit with business leadership (often the customer function) and not purely within data or technology teams as risk becomes diffused and harder to manage.
Data foundations are the gating factor
While AI tooling continues to advance, its effectiveness is still constrained by data quality and accessibility.
“Garbage in, garbage out” remains the defining limitation, as one participant described, with many early proof-of-concepts failed because underlying data around contracts, records, customer information was fragmented or unreliable.
The current wave of AI investment is forcing organisations to address long-standing foundational gaps that were previously deprioritised. Until this is done, successful scaling cannot happen.
Automation is reshaping roles
There is growing concern across the industry about the creation of low-value oversight roles, with one participant asking: “Are we going to create some really boring jobs?”
At the same time, demand is increasing for translator roles to bridge business needs, data capabilities, and technology delivery.
The implication is a widening gap: fewer entry-level roles, but greater need for broad, adaptive skillsets that combine technical understanding with commercial and communication strengths.
Practical Next Steps
- Map customer journeys by value and complexity. Identify where automation improves experience versus where human interaction remains critical, then design around those distinctions.
- Ensure automated systems can recognise friction and route to humans quickly and transparently, especially in sensitive or high-value interactions.
- Evaluate automation using a mix of cost, customer outcomes, risk, and brand impact.
- Tie every AI initiative to a business owner and outcome. Require clear P&L sponsorship and measurable objectives to avoid fragmented or low-impact investments.
- Invest in data foundations before scaling AI. Prioritise records management, data quality, and access to canonical sources to avoid repeated POC failures.
- Clarify ownership and governance for AI risk by defining who is accountable for customer outcomes.
- Use pilots to validate impact, particularly in customer-facing processes where failure carries reputational risk.
- Protect human-led moments that define the brand. Be explicit about where human interaction is a differentiator and design automation to support those moments.
- Plan for fewer transactional roles and greater need for translators, orchestrators, and cross-functional thinkers.
