Will customers know they are talking to AI?

Yes, because pretending otherwise is both unethical and quickly obvious. Disclosure does not hurt satisfaction when the resolution is fast and correct — what hurts satisfaction is a bot that loops, cannot escalate, or hides the route to a person.

What stops it doing something expensive and wrong?

Action-level limits. Refunds above an agreed threshold, account closures and anything irreversible route to a human queue regardless of confidence. The agent's authority is configured deliberately, not inherited from an API key.

How does it handle a question it has never seen?

It says so and hands over with the conversation summarised. An agent that guesses to avoid escalating is worse than no agent, and we tune the confidence threshold against real traffic to get that boundary right.

How long before it is actually useful?

Four weeks for one well-defined ticket category in production, which is the Automation Sprint. Expanding to more categories is faster because the integration and evaluation scaffolding already exists.

By use case

Customer Support Automation

Agents that resolve tickets, not summarise them

Scope this for us Model the savings

Usually owned by

Customer Support
Customer Experience
Operations

55-70%
Tier-1 tickets fully resolved: < 30 sec
First response, 24x7: < 2%
Escalation-after-resolution rate

The change

What actually differs afterwards

Not a maturity model. The concrete difference in how the work happens.

Today

Agents copy the same answer from a wiki forty times a day
First response times stretch overnight and over weekends
Tier-1 volume grows linearly with customers
Every action requires a human to click through three systems
Quality varies by which agent picked up the ticket

Afterwards

Routine tickets resolved automatically, including the action
First response measured in seconds, at any hour
Volume grows without headcount growing with it
Agents handle the genuinely hard 20% with full context
Consistent answers, with an audit trail on every action

Start with one ticket category

The instinct is to build an assistant that handles everything. It produces a system that handles nothing well.

Pick the highest-volume category where the correct answer is knowable from data you already have — order status, refund eligibility, password and access issues, subscription changes. Automate that completely, including the action in the underlying system. Then take the next one.

Resolution means taking the action

The distinction that determines whether this is worth doing:

Assistance

Drafts a reply for an agent to review and send
Tells the customer what a human will do next
Summarises the ticket for the queue

Resolution

Issues the refund in the payment system
Updates the shipping address and reconfirms delivery
Cancels the subscription and sends confirmation

Assistance saves an agent a few minutes. Resolution removes the ticket. The engineering difference is integration and permissions — which is most of the work, and the reason demos are easy and production is not.

What we measure

Resolution rate without human involvement, escalation-after-resolution rate (did the customer come back because we got it wrong), time to first response, and cost per resolved ticket. Satisfaction is tracked but it lags; the escalation-after-resolution number is the honest early signal.

FAQ

Questions we get asked

Let's find out what is actually automatable

Bring a process that annoys you. In 30 minutes we will tell you whether AI helps, what it would cost, and where it would fail — even if the answer is don't bother.

Book a consultation Get a quote

Or email [email protected] · we reply within 1 business day