Home / C.A.T

Service area · 03

Tiered support, reinvented with AI.

The C.A.T —our Technical Support Centre— resolves with an AI agent on the 24×7 front line and, behind it, people in tiers (N1–N2–N3). It resolves over the service's historical knowledge and, when it escalates, does so with the full context. Already in production in regulated healthcare.

24×7AI agent on the front line
≥ 65%resolved at first contact (N1)
40–60%deflection: resolution with no human
In productionin regulated healthcare
What we solve

What we hear from teams running support under strain.

“When a peak hits, support saturates and response times blow up.”

AI handles the front line 24×7: common cases are resolved on the spot and people focus on what needs judgement.

“The person who knew how to fix this is gone, and their knowledge left with them.”

The history is the memory: every ticket, document and past resolution feeds the agents, so the knowledge stays.

“I don't want a bot that goes in circles; I want it to resolve or put me through to someone.”

Clean escalation by confidence level: the agent resolves when it is sure and hands over to a person when it is not, always with the full trace.

“Our data cannot leave the building.”

On-premise deployment for data-sovereignty environments: the agents work where the information is, not the other way round.

The service

How the C.A.T resolves, level by level.

AI handles the front line; people step in when the case calls for it.

AI

AI agent · 24×7

A front line that resolves what can be resolved and escalates with judgement and full context. Already in production in regulated healthcare.

N1

People + AI

A single point of contact and resolution with a living knowledge base. Target: ≥ 65% resolved at first contact.

N2

Specialists

Deep diagnosis and problem management: eradicate the cause, don't repeat the patch. A monthly top-10 of recurring incidents.

N3

Engineering

Code, architecture and vendors. Escalation arrives with the full context: nothing is lost along the way.

Entry points

Three ways to start, depending on where you are.

All with a clear deliverable and scope. You decide how far to take the automation.

A

Support diagnosis

You don't yet know where to start

We review volumes, repetitive cases and your history. Deliverable: where AI fits and what stays resolved by people.

2–3 weeks
B

Front-line pilot

You have a clear domain to try

An agent resolving over your history on a bounded scope, with escalation to a person. Deliverable: an agent in production and its measurement.

4–8 weeks
C

Managed C.A.T 24×7

You want the full tiered service

N1–N2–N3 support with AI on the front line, SLAs and traceability. On-premise if your data cannot leave.

Ongoing service
The numbers, straight

AI, with the numbers straight.

Real deflection, where the market is heading and why a good service ends up shrinking its own volume.

i.

Deflection measured properly

Full resolution, with no human and no reopening — not auto-replies. Best-in-class 2026: 40–60%.

ii.

Where this is heading

Gartner: by 2029 agentic AI will resolve 80% of common incidents on its own, at −30% cost.

iii.

Shift-left

Every resolution that moves down a tier costs 3 to 10 times less. A good service shrinks its own volume.

iv.

ENGRAM

The service's knowledge outlives staff turnover: every resolution leaves accumulated judgement.

Other areas

The rest of the house.