Customer service
Customers answered in minutes. People on the cases that matter.
At a glance
A customer writes in: their order arrived damaged and they want a refund.
- Deployed AI
About 15 minutes per ticket
The agent does the work
- Operational AI
Minutes, around the clock
AI resolves what policy allows; people take the rest
Illustration
What matters
What customer service teams are judged on.
Time to resolution
Answers in minutes, at any hour, in the customer's language.
Satisfaction
Fewer repeats, fewer transfers, fewer complaints.
Cost per contact
Volume grows without the team growing at the same rate.
Agents who stay
People spend their time on problems worth their skill.
Walkthrough
The same request, three ways.
The request
EmailJ. Ashby, customer
Order 58213 arrived damaged
The lamp arrived with a cracked base.
I would like a refund, please.
Autonomy level 1
Deployed AI
A chat assistant. Someone asks, it answers, people do the rest.
- A person types
Customer says order 58213 arrived damaged and wants a refund. Write a reply.
- AI replies
Dear customer, we're very sorry to hear about your experience. Please send us photos of the damage and we'll look into it as soon as possible.
- Person
Agent. Opens three systems, checks the order and the policy, issues the refund and updates the ticket.
3 steps · one to a person · about 15 minutes per ticket
- How long
- About 15 minutes per ticket
- Who does the work
- The agent does the work
- What keeps it safe
- None: replies vary from agent to agent
- What goes to people
- Everything
Autonomy level 2
Specialised AI
It works from your data and tools and prepares the work. A person approves.
- Trigger
The email arrives in the help desk.
- Tools
orders.get(58213) · policy.refunds("damaged in transit")Returned: Delivered two days ago. The customer's first claim. Policy: refund or replace within 30 days; no photo needed for small items.
- AI replies
Eligible. Proposed: a full refund today, with no need to send the lamp back. The courier claim is drafted for afterwards.
- Person
Agent. Checks the proposal and approves it.
Approved - Action
Refund issued and the ticket closed; the courier claim filed after.
5 steps · one to a person · about 2 minutes of an agent's time
- How long
- About 2 minutes of an agent's time
- Who does the work
- AI prepares, the agent approves
- What keeps it safe
- The policy cited in every decision
- What goes to people
- Every reply
Autonomy level 3
Operational AI
It runs the process within set limits, logs every step and hands people the exceptions.
- Trigger
Every contact, on any channel, at any hour.
- Rule
Customers are told they are talking to AI, and can ask for a person at any time.
Refunds within policy and under the agreed limit go ahead. Repeat claims, safety issues and upset customers go to a person.
- Action
Refund issued and the customer told within minutes; the courier claim filed after.
- Person
Exception: goes to a person
Senior agent. A third damage claim this year from the same account goes to a senior agent, with the history and a proposed resolution.
Escalated - Record
[11:42]Decision, policy clause and evidence saved on the ticket.
- Report
Weekly to the service lead: resolution time, satisfaction, refunds by reason, and gaps in the policy.
6 steps · one to a person · minutes, around the clock
- How long
- Minutes, around the clock
- Who does the work
- AI resolves what policy allows; people take the rest
- What keeps it safe
- Refund limits, AI disclosure, a person on request
- What goes to people
- Repeat claims, safety issues, anyone who asks
Controls
- Owner
- Head of customer service
- Measure
- Resolution time and satisfaction
- Limit
- Refunds under the agreed amount
- Log
- Every decision on the ticket
- Fallback
- Tickets queue for agents
- Stop
- Hand every chat to people in one switch
Illustrative: names, figures and timings are invented.
Where to start
Processes that pay back first.
3Run with exceptions
Triage and routing of every contact
By intent, urgency and value.
2Automate with approval
Replies from your knowledge base
Agents approve until quality is proven.
3Run with exceptions
Refunds and credits within policy
Limits you set; above them, a person.
3Run with exceptions
Quality review of every conversation
Every conversation, not a small sample.
2Automate with approval
Knowledge base upkeep
Gaps found from real tickets.
1Assist
Complaints and vulnerable customers
Always a person; AI prepares the file.
What we build and run
Built in your systems, run with your team.
- Triage and routing across email, chat and phone notes
- Answers grounded in your policy, with sources
- Resolution flows with limits and approvals
- Quality review across every conversation
What we measure
Numbers you already track.
- Time to first response
- Time to resolution
- Satisfaction
- Resolved without a transfer
- Cost per contact
The Academy
Learn to run it yourselves: the Customer service track.
Answer faster with AI, and send the hard cases to people.
Being built