01
Make decisions continuous
Re-evaluate wait time, queue health and expert availability throughout the conversation—not only at arrival.
← Alipta BallavIndividual Contributor
03
Microsoft · Dynamics 365 Contact Center · 2022–2026
Reducing Average Handling Time through Conversation Orchestration
An AI-native coordination layer that reduces avoidable waiting, transfers and repeated context—without treating shorter calls as the only measure of success.
Business goal
Reduce Average Handling Time
My role
Individual contributor · Product design
Design shift
Static routing → continuous orchestration
Evidence
700+ preview customers · 8–10 → 2 days
01 · Business problem
AHT was being lost in the system, not only on the call.
Average Handling Time combines talk time, hold time and after-call work. Static routing increased avoidable waits, transfers and repeated explanations before an agent could resolve the issue.
AHT = (talk + hold + after-call work) ÷ interactions
Design could influence hold time and talk time by getting each conversation to the right resource with context intact.
02 · Design challenge
Translate an operational target into better decisions.
The design responsibility was not to rush agents. It was to remove the orchestration friction that made customers wait, repeat themselves or move between teams.
Goal
Reduce AHT
Friction
Waiting · transfers · lost context
Design
Continuous, explainable decisions
Outcome
Faster resolution with quality intact
03 · Key decisions
Three moves connected design to the business goal.
02
Make AI logic inspectable
Use grounded variables, live previews and explicit fallbacks so admins can trust what will happen.
03
Make authorship accessible
Replace brittle rule trees with natural-language playbooks, templates and reusable patterns.
04 · Product evidence
The system observes, evaluates and acts.
Playbooks continuously coordinate priority, overflow and reconnection as customer and operational conditions change.
If VIP + wait > 30 sec
→ Escalate priority
If No agent available
→ Route to overflow
If Repeat customer
→ Reconnect previous agent
05 · Outcome and evidence
Separate what shipped from what it is designed to move.
No unsupported AHT reduction is claimed. The business measure remains explicit; product and delivery evidence show the mechanism and adoption signal.
Evidence
700+
Customers in public preview
Evidence
8–10 → 2 days
End-to-end prototype integration
Intended impact
Lower AHT
Less waiting, transferring and repetition
Intended impact
Faster setup
Natural-language playbooks replace rule-heavy configuration
06 · Craft takeaway
Design the mechanism behind the metric.
“Reducing AHT is not about making conversations shorter. It is about removing the system friction that delays resolution.”