SaaS and technologyAI and data
Turning support tickets into answers, and cutting resolution time 41%
Halcyon's support team was drowning in tickets. We built a practical AI assistant on their own data, governed and measurable.
41%
faster median ticket resolution
32%
of tickets deflected by self-serve answers
4.7 of 5
customer satisfaction, up from 4.1
The challenge
Where they started.
Halcyon, a fast-growing SaaS company, was adding customers faster than it could hire support. Answers existed across docs, tickets, and code, but nobody could find them quickly, and quality varied by agent.
The stack
Google CloudVertex AIBigQueryLangChainPythonLooker
What we did
- We built a retrieval system over Halcyon's own documentation and resolved tickets, with clear guardrails and source citations.
- An assistant drafts answers for agents to review, so a person stays in the loop and quality stays high.
- We instrumented everything: accuracy, deflection, and resolution time, so the value is measured, not assumed.
- The data foundations were cleaned up first, which made the AI reliable rather than a demo.
More work
Other engagements.
Lower cost, lower risk, one partner
Let's write yours.
Tell us where the business hurts. We will map a pragmatic next step and a plan you own.