Blending AI and people in a contact centre: a UK guide
By Keith Hamilton · · 5 min read
Most contact centres do not need to choose between AI and people. Blending AI and people in a contact centre means letting AI handle routine conversations while your team handles the ones that need judgement, care or authority. This article covers which conversations suit AI, how to escalate with context, what should stay with people, how to monitor quality and how staff roles change.
Why blend AI and people rather than choose one?
AI is good at being available, consistent and quick on simple, repeatable conversations. People are good at empathy, judgement and dealing with the unexpected. A contact centre that uses only one of them usually ends up either stretched too thin or frustrating its customers.
A blended model puts each conversation in front of whoever is best placed to handle it. Customers get faster answers on simple questions, and your advisers get more time for the conversations that really need them.
In our own work, a blended contact centre service has AI taking calls and answering queries, while clients' own human teams handle escalations and regulated business. You can read more in the blended contact centre case study.
Which customer conversations suit AI?
Start by listing your contact reasons and roughly how many of each you get. Conversations that tend to suit AI share a few features:
- They happen often and follow a similar pattern each time
- The answer is held in a system the AI can reach, such as an order, booking or account record
- A wrong answer would be inconvenient rather than harmful
- Most customers would be happy to sort them out without speaking to a person
Typical examples include order and delivery updates, appointment changes, opening hours, simple account questions after proper security checks, and gathering details before passing a customer to the right team.
How should escalation work?
Escalation is where blended contact centres succeed or fail. A customer who is passed to a person and then has to start again from scratch will feel let down, however good the AI was.
Hand over the context
When the AI passes a conversation on, the adviser should see who the customer is, what they asked, what checks have been done and what the AI has already said or done. A short written summary at the top of the record saves time on both sides.
Escalate on signals, not only on request
Customers should always be able to ask for a person. The AI should also escalate when it notices:
- It has misunderstood the customer more than once
- The customer is upset, or says they want to complain
- There are signs the customer may be vulnerable
- The request falls outside what it has been set up to handle
Which conversations should stay with people?
Some conversations should stay with your team, even if AI could technically take part in them.
Regulated conversations are the clearest example. Firms in regulated sectors, for example financial services, must follow their regulator's rules, including treating customers in financial difficulty fairly. Advice, affordability discussions and similar conversations generally belong with trained people, with AI at most helping to gather information beforehand.
Sensitive conversations also need a person. These include bereavement, health concerns, safeguarding worries, serious complaints and anyone who seems vulnerable. In these moments the customer needs to feel heard, and your business needs a person exercising judgement.
How do you monitor quality across AI and people?
Hold AI conversations to the same standard as human ones. In practice that means:
- Reviewing a regular sample of AI transcripts and recordings, not only the ones customers complain about
- Tracking resolution, transfer rates and the reasons for each transfer
- Watching for repeat contacts, which often show an answer did not really solve the problem
- Collecting customer feedback on AI and human conversations in the same way
- Feeding what you learn back into the AI's information and the team's guidance
Be open with customers that they are speaking to AI. Recordings and transcripts contain personal data, so UK GDPR applies to how you store, use and keep them.
What happens to contact centre staff roles?
Roles change rather than disappear. As AI takes on routine contacts, the conversations that reach people tend to be harder, more emotional or more complex. Your advisers become the specialists customers are escalated to.
New or larger responsibilities often appear, such as:
- Reviewing AI conversations and flagging wrong or unclear answers
- Keeping the knowledge the AI relies on accurate and up to date
- Handling escalations and regulated or sensitive cases
- Supervising the blended operation and watching the measures above
Harder conversations all day can be tiring, so training, support and attention to wellbeing matter more, not less. Involving advisers early, and listening to what they know about customers, also makes the AI better.
How do you get started?
Keep the first step small:
- Map your contact reasons and pick one or two that are high volume and low risk
- Design the escalation route and the hand-over summary before anything else
- Record today's measures so you have a baseline
- Run the AI on those contact reasons with close monitoring
- Review the results with your team, then decide whether to extend it
Frequently asked questions
Will customers accept AI in our contact centre?
Many customers are happy to use AI for simple questions, especially when it answers quickly and they can reach a person easily. Problems usually come from poor escalation or from AI being used for the wrong conversations. Being open that it is AI also helps.
Can AI handle regulated conversations?
Regulated firms must follow their regulator's rules, so regulated conversations generally stay with trained people. AI can often help around the edges, for example by gathering details or routing the customer, but the regulated part should sit with your team and be checked by your compliance lead.
How many of our contacts could AI handle?
It depends on your mix of contact reasons, the quality of your data and how well your systems connect. Automation typically removes somewhere between a third and two thirds of a repetitive task, never all of it, so plan for people to remain central.
Where Forwardcycle fits
Forwardcycle helps contact centres decide which conversations suit AI, design escalation that works and keep quality high. Find out more about our customer experience services, take the free AI Readiness Assessment or arrange a free call.