Aven sells subscription hardware across eleven markets. Before the project, their support team of 34 worked from four separate tools: a helpdesk for email, a chat widget with its own inbox, a shared phone for WhatsApp, and a spreadsheet for anything escalated. Median first response time was nine minutes; the 90th percentile was over an hour.
The diagnosis
The team's instinct was that they needed more agents. The arrival-time analysis said otherwise: volume was concentrated in two daily peaks, and during those peaks agents spent a large share of their time reconstructing context from other tools rather than replying. The constraint was context switching, not capacity.
What they changed
- Consolidated all four channels into one queue with a single customer timeline, so WhatsApp threads carried the same history as email.
- Introduced a dedicated triage rotation for the two peak hours, with explicit ownership of anything unassigned for more than three minutes.
- Automated three intents end to end, order status, delivery rescheduling and device replacement eligibility, with hard policy limits and immediate hand-off on low confidence.
- Replaced per-channel dashboards with conversation-level reporting, including a percentile view instead of averages.
Results after two quarters
Median first response time fell to 42 seconds and the 90th percentile to under four minutes, which the team considers the more meaningful improvement. Repeat contact rate fell by roughly a third. Headcount was unchanged; two agents moved into a proactive outreach function created with the recovered capacity.
What they would do differently
Aven's operations lead is candid that they automated too broadly in the first month and pulled two intents back after seeing abandonment rates. The lesson they pass on is to launch with fewer automated intents than you think you can handle, and to measure abandonment from day one.