AI self-service in the contact centre: what to automate and what to keep human

Your advisors probably answer the same questions many times a day. When's my bin collection? Has my payment gone through? Has my repair been booked? Each one is quick to deal with, but together they fill the queue, and the people who need more help wait alongside them.

AI-enabled self-service gives your customers instant answers at any time of day and frees your advisors for the conversations where they make the biggest difference. In our experience, it typically handles 30 to 40% of inbound contact volume.

The benefits come from choosing the right enquiries. Automate the simple, high-volume ones, and hand everything else to a person quickly, with the full context of the conversation. That matters even more in public services and social housing, where clear rules set out how some contacts must be handled.

This insight examines the role of AI self-service in the contact centre, highlighting what can be automated, what should remain human-led and how to implement it successfully.

What is AI-enabled self-service?

AI-enabled self-service uses conversational AI to answer customer enquiries without an advisor, across voice, web chat, messaging and email.

Traditional IVR menus ask callers to press 1 for billing or 2 for bookings. Conversational AI understands what people mean in their own words, so it can:

Answer common questions using your knowledge base and policies.
Complete simple tasks, such as taking a meter reading, booking an appointment or updating contact details.
Verify identity before sharing account information.
Route enquiries to the right team when a person is needed.
Pass the full context to an advisor, so customers only explain their issue once.

It works best as part of a joined-up contact centre. When you connect AI self-service to your unified CX platform and the systems your advisors already use, customers move smoothly from the AI to an advisor.

When we moved the DVLA to an AI-enabled cloud contact centre, it saw measurable improvements across every channel. Click here to read the DVLA case study.

Channel or measure	Result
Web chat	300,000 sessions a month, with 30% resolved through self-service
Chat handling time	Around 150 seconds saved per chat
Email	Automated responses to 50,000 emails a month
IVR	Time spent in the IVR cut by 50%
Cost	£8 million saved over the contract term

What to automate and what to keep human

Automate enquiries that are frequent, simple and low risk. Keep people involved where there's urgency, vulnerability, emotion, judgement or a significant decision.

As a simple rule of thumb, involve a person whenever getting it wrong could harm someone, cost them money or break a rule.

Enquiry type	Approach	Why
Opening hours, locations, contact details	Automate	High volume, with fixed answers
Application, order or case status updates	Automate	Simple lookups from existing systems
Booking, changing or cancelling appointments	Automate	Rules-based and easy to confirm
Payments, balances and payment dates	Automate after identity checks	Routine, with secure verification
Password resets and account updates	Automate	Repetitive tasks with a clear process
Reporting a routine issue (missed bin, minor repair)	Automate with triage	AI captures the details and sends urgent cases straight to a person
Complaints	Capture, then hand to a person	Needs empathy, investigation and a proper response
Safety risks (gas leak, flooding, damp and mould)	Escalate to a person immediately	Urgent and legally time-bound
Vulnerable customers or safeguarding concerns	Keep human	Needs human judgement and care
Bereavement, debt or financial hardship	Keep human	Emotionally sensitive and needs flexibility
Decisions with legal or significant effects (eligibility, refusals)	Keep human, or add human review	People have rights to human review
Customer asks for a person	Connect them straight away	Always offer a clear route to an advisor

Good AI self-service plays to its strengths. It handles the routine reliably, recognises when a person is needed and gets the customer to the right advisor quickly.

AI self-service for public sector enquiries

If you're a council, government agency or other public sector body, AI self-service helps you meet rising demand without adding headcount, and keeps your services available outside office hours.

Public sector enquiries: automate or keep human?

Enquiry	Approach	Notes
Council tax balances, payment dates, direct debit changes	Automate	Verify identity first
Bin collection days and missed collections	Automate	High volume, simple lookups
Licence, permit and application status	Automate	Pull updates from case systems
Appointment bookings (registrations, recycling slots)	Automate	Rules-based and easy to confirm
Reports such as potholes, fly-tipping, street lighting	Automate with triage	Capture location and photos, and flag hazards
Signposting to services and forms	Automate	Reduces misdirected contacts
Benefits, eligibility and support decisions	Keep human	Significant effects, so a person reviews them
Safeguarding and vulnerable residents	Keep human	Needs judgement and care
Complaints and appeals	Capture, then hand over	Needs investigation and a proper response
Hardship, homelessness or crisis support	Keep human	Urgent and sensitive

The government's own guidance supports this approach. The AI Playbook for the UK Government says organisations should have "meaningful human control at the right stages", with humans validating high-risk decisions. For chatbots that respond instantly, it recommends building in human control at other stages, including thorough testing, regular checks of the live service and a way for users to report problems for a person to review.

Public services also need to work for everyone. Make your self-service accessible, write it in plain English and keep other ways to get help open alongside it. Some residents will always prefer, or need, to speak to a person.

The DVLA shows what's possible at scale. Its AI-enabled contact centre handles 300,000 web chats and automates responses to 50,000 emails each month, which frees advisors for the complex cases that need them.

AI self-service for housing enquiries

Housing associations and council landlords handle large volumes of repairs, rent and tenancy enquiries. AI self-service can take on much of that work, and set up well, it can also help you meet some of the strictest rules on how certain contacts are handled.

Housing enquiries: automate or keep human?

Enquiry	Approach	Notes
Rent balances, payment dates and confirmations	Automate	Verify identity first
Routine repairs (dripping tap, broken cupboard door)	Automate with triage	Capture details and photos
Repair appointments and reminders	Automate	Book, confirm and remind automatically
Repair progress updates	Automate	Pull status from your housing system
Contact detail changes and tenancy information	Automate	Simple, rules-based tasks
Damp, mould and emergency hazards	Escalate immediately	Awaab's Law timescales apply
Complaints or expressions of dissatisfaction	Capture, then hand over	Complaint Handling Code applies
Rent arrears and financial hardship	Keep human	Needs flexibility and support
Domestic abuse, health concerns, bereavement	Keep human	Needs flexibility and support

The rules that shape housing self-service

Damp, mould and emergency hazards

Since 27 October 2025, Awaab's Law has required social landlords to investigate and make safe emergency hazards within 24 hours. Significant damp and mould hazards must be investigated within 10 working days. The clock starts when the landlord becomes aware of the hazard, so your AI needs to recognise these reports and escalate them to a person straight away.

Complaints

The Housing Ombudsman's Complaint Handling Code became statutory on 1 April 2024. It defines a complaint as "an expression of dissatisfaction, however made", and says residents must be able to complain "in any way". If a tenant expresses dissatisfaction to a chatbot, you need to give them the choice to make a complaint. Stage 1 complaints must be acknowledged within five working days.

Vulnerability and hardship

Rent arrears, domestic abuse, health concerns and bereavement all need a trained person who can listen and respond with care.

Using AI to strengthen compliance

AI self-service can help your housing team meet compliance rules and show how you've met them. It can:

Spot keywords such as "damp", "mould", "gas" or "leak" and route them to an urgent queue.
Timestamp every report, giving you a clear record of when you became aware.
Recognise dissatisfaction and offer to log a complaint.
Create a full audit trail of every interaction.

Find out more about our contact centre solutions for housing and IT services for housing associations.

Getting the handover right

The handover from AI to a person shapes how customers feel about the whole experience. When the advisor already knows what's been said, the customer can pick up exactly where they left off.

Five things make a good handover:

Give customers a way to ask for a person at any point, in plain words.
Pass the full context to the advisor, including the conversation, the customer's details and what's already been tried. Integrated CX puts all of this on one screen.
Hand over when the AI detects urgency, vulnerability or frustration, as well as when it reaches the limit of what it can answer.
Use what the AI has learned to route the customer to the right team or skill first time.
Keep the customer's place in the queue when they move from chat or IVR to a call.

Staying compliant

AI self-service handles personal data, so it needs to meet UK data protection law. Build these steps into your service:

Tell customers when they're talking to an automated assistant.
Verify identity before sharing account details, and collect only the data you need.
Keep people involved in significant decisions. The ICO explains that people have rights to obtain human intervention and to challenge decisions made solely by automated means that have legal or similarly significant effects.
Use the audit trails and activity logs your automated processes create to support complaints handling, regulatory reporting and quality checks.
Check answers regularly, review conversations the AI couldn't resolve and update your knowledge base as policies change.

How to get started with AI self-service

Start with the enquiries you understand best, then build from there.

Analyse your contact data to find the 10 to 20 enquiry types that make up most of your volume. These are your starting point.
Sort them using the automate or keep human test, and begin with the simple, low-risk ones.
Agree your escalation rules, including the keywords, topics and customer signals that trigger an immediate handover. Pay particular attention to safety and vulnerability.
Get your knowledge base in order. Your AI draws its answers from this information, so keep it accurate and up to date.
Connect self-service to your CRM, housing or case management systems so it can complete tasks as well as answer questions.
Run a pilot and track containment rate, customer satisfaction, handover reasons and repeat contacts. Use what you learn to refine the service, then expand to more enquiry types.
Involve your advisors in design and testing. They know the questions customers ask, and they'll handle the contacts the AI passes on.

What to measure in your pilot

Measure	What it tells you	What good looks like
Containment rate	Share of enquiries resolved without an advisor	Rising steadily for automated enquiry types
Handover reasons	Why the AI passed contacts to a person	Mostly planned escalations
Customer satisfaction (CSAT)	How customers rate the experience	Equal to or better than advisor-handled contacts
Repeat contact rate	Whether customers come back about the same issue	Falling over time
Average handling time	How long advisors spend on handed-over contacts	Lower, because context is passed across
Escalation speed for urgent issues	How fast safety or vulnerability cases reach a person	Immediate, with a timestamped record

If you're also reviewing your wider platform, our guide to contact centre as a service (CCaaS) explains how cloud contact centres support AI. If your teams already use Microsoft Teams, see how to turn Teams into an AI-powered contact centre.

Find the right balance for your contact centre

Our AI-enabled self-service helps you automate routine enquiries and keep people at the heart of your service. We'll help you:

Work out which enquiries to automate and which to keep human
Set up clear escalation rules for urgent and sensitive contacts
Connect self-service to your existing contact centre and business systems
Create full audit trails to support compliance
Configure, deploy and support the service from day one, and adapt it as your needs change

We work with a range of contact centre platforms and have delivered AI-enabled contact centres for organisations including the DVLA.

Frequently asked questions

What is AI self-service in a contact centre?

AI self-service uses conversational AI to answer customer enquiries and complete simple tasks without an advisor. It works across voice, web chat, messaging and email, and hands over to a person when needed.

What percentage of contact centre enquiries can AI handle?

In our experience, AI self-service typically handles 30 to 40% of inbound contact volume. The right level depends on your enquiry mix, and simple, repetitive questions are the easiest to automate.

Will AI self-service replace contact centre advisors?

It works alongside them. AI takes on routine enquiries so your advisors can spend more time on complex, sensitive and high-value conversations. It can also help you handle growth and seasonal peaks without adding headcount.

Which enquiries should stay with a person?

Safety risks, safeguarding and vulnerability concerns, complaints, financial hardship, bereavement and decisions with legal or significant effects. These need human judgement, and in some sectors, there are rules on how they're handled.

How does AI self-service handle damp and mould reports in social housing?

It should recognise them straight away and escalate them to a person. Under Awaab's Law, landlords must investigate and make safe emergency hazards within 24 hours, and investigate significant damp and mould hazards within 10 working days of becoming aware of them.

Can a chatbot accept a housing complaint?

It can capture one and pass it to your complaints team. The Housing Ombudsman's Complaint Handling Code says residents can complain in any way, so a chatbot should recognise dissatisfaction, offer to log a complaint and hand it to a person.

Is AI self-service suitable for the public sector?

Yes, for routine enquiries such as payments, status updates, bookings and signposting. Government guidance calls for meaningful human control, so high-risk decisions should stay with people, and residents should always have other ways to get help.

Is AI self-service GDPR compliant?

It can be, when it's designed with data protection in mind. Tell people they're using AI, verify identity before sharing personal data, collect only what you need and give people a route to human review of significant decisions.

How long does it take to set up AI self-service?

It depends on the number of enquiry types and the systems involved. A good approach is to start with a pilot covering your highest-volume, simplest enquiries, then expand from there.