●
SaaS
·
AI support desk
How a software company answers most support tickets before a person opens them.
Client
Timeline
1-week audit, 4-week build
Tools
Zendesk, Notion, Slack, an LLM agent
The problem
A four-person support team was answering the same twenty questions all day, while urgent tickets waited in the queue.
Password resets, billing questions and how-to requests made up most of the volume. The answers already existed in the help centre, but nobody had time to link them, and first replies were taking most of a working day.
What we built
An agent that reads each new ticket, finds the right answer in the help centre and replies within a minute.
If the agent isn't confident, or the customer mentions a refund, an outage or cancelling, the ticket goes straight to a person with a drafted reply and the sources it used. The team reviews a sample of automatic replies every Friday.
Before
Ticket arrives → waits in queue → agent reads it → searches help centre → writes reply → tags and closes. Several hours to a first reply.
After
Ticket arrives → agent finds the answer → reply sent in under a minute → tricky tickets routed to a person with a draft ready.
Tickets answered automatically
First reply time
Paid back in
“Our team finally works on the tickets that need a human. Customers get an answer in seconds, and we stopped dreading Monday mornings.”
Priya Natarajan (placeholder)
Head of Support, Example Software Ltd.
AI agents, automations and integrations, built inside the tools you already pay for. Fixed price, agreed before we start.
