Case studies

Case study · B2B SaaS · Customer support

3 June 2026

An AI support agent that handles 85% of their tickets

Their support inbox was overwhelmed and the obvious fix was hiring. In three weeks we built them an AI agent instead — it now resolves 85% of every ticket that comes in, and the two new hires never happened.

  • AI support agent
  • Knowledge base
  • Human handover
85%of tickets handled
3 wksto build & go live
£6k/moin hires avoided
24/7always answering

Their support inbox was drowning. Tickets were coming in faster than the team could clear them, and the only obvious answer was to hire.

This is a fast-growing B2B SaaS company, and support volume climbed with every new customer they signed. The realistic next move was two more support staff — somewhere around £5,000 to £6,000 a month between them. Instead, they put that money into an AI agent that scales without a salary, so they could keep growing without hiring indefinitely.

What we built

An AI support agent — we called her Amelia — that reads, understands and answers support tickets in the company's own voice. We built her and got her live in three weeks.

One inbox for everything

First we gave her a dedicated support inbox and forwarded all of their existing support addresses into it with ImprovMX, so she monitors every inbox from one place. Then we wired her into their website, so the moment anyone raises a support ticket it lands with Amelia first.

We taught her everything they knew

Her knowledge base is built entirely from the company's own documentation. Every doc, guide and learning resource they'd written, we uploaded and turned into knowledge she can actually use:

01

Their services

What the product does, how every feature works, plans, limits and the details customers ask about daily.

02

Resolving issues

Their playbooks for the common problems — the fixes, the settings, and the step-by-step answers their team already knew.

03

How-tos & onboarding

Every guide and learning resource, so she can walk a customer through anything from setup to advanced use.

Here's Amelia in action

A customer raises a ticket from the help centre we built — the same docs and guides that were loaded into Amelia's knowledge base:

A customer raising a support ticket on the company's help centre, with the knowledge-base docs above the form
The help centre and the "Talk to support" form. Every ticket lands with Amelia first.

Seconds later, here's the reply Amelia sent back — no human touched it:

She answers first, humans catch the rest

Every ticket goes to Amelia before anyone on the team sees it. If it's within her scope — about 85% of the time — she answers it and closes it out. If it isn't, a human-handover alert pings the support manager, who sends the reply themselves. There's always a human backup, and nothing falls through the cracks.

1Ticket inFrom the website or any inbox
2Amelia reads itChecks it against the knowledge base
3In scope?Yes → she resolves & replies
Human handoverOut of scope → alerts the support manager, who sends the reply. Always a human backup.

She flags what's missing

This part is quietly valuable. When a ticket surfaces a problem or a gap the documentation and product don't cover, Amelia doesn't just muddle through — she flags it as a suggestion. Over time the support inbox becomes a product feedback loop, surfacing exactly what customers keep getting stuck on.

The maths

85%by Amelia

Amelia handles 85% of every ticket that comes in. The other 15% are the genuinely unique cases a human should see — and they go straight to one.

The two support hires they were about to make, at £5–6k a month, never happened. And because software doesn't need a second desk, they can keep scaling support without hiring indefinitely. The return on a fraction of one salary is enormous.

The result

Built and live in three weeks. Amelia now clears 85% of their support load, around the clock, for a fraction of what two hires would have cost. The team is off the support treadmill and back on the work that actually grows the business — more of their time and their capital going into marketing and sales.

They were about to hire two people just to keep up. Instead they got an agent that handles 85% of it on day one, and a team free to go after growth.

Is support eating your team's week?

Book a free AI Blueprint call. We'll find what's draining your hours and map out exactly what to automate first.

Get a free AI Blueprint