Data Platform (Known-Issue Assist)
Outcome (8 weeks): Time to resolution −22%, rework −25%, first-reply quality up (“helpful on first pass” +18 pts).
Engagement: 60-Day Ops Reset focused on a privacy-safe known-issue assist, ~10–12 hrs/week, remote.
Results (by week 8)
Time to resolution (median): −22% (assist + runbooks cut search time)
Rework: −25% (fewer “that didn’t work” reopenings)
First-reply quality: +18 pts in internal QA (“helpful on first pass”)
Engineering interrupts: −29% for repeatable known issues
Assist adoption: 71% of eligible tickets used at least one suggestion
What changed day-to-day:
Agents started replies with a linked source (article/runbook), not guesswork
Product areas with repeat defects were visible; leaders prioritized fixes by volume + impact
Technical writers had a single queue of article refreshes (oldest/riskiest first)
Starting point
Agents relied on tribal knowledge; solutions scattered across Confluence, Jira comments, internal emails
“Known issues” lived in release notes + Slack threads; hard to find under pressure
Rework common (ticket reopened after first reply); leaders lacked signal on “repeatable fixes”
Engineering interrupted by repeat asks for issues with workarounds
Company snapshot
B2B data platform, Series C, ~250 employees, global customer base
Stack: Zendesk, Jira, Confluence, Notion, Snowflake, Looker
Team: 22 in Support (T1/T2), 4 incident managers, 3 technical writers (part-time to Support)