Fewer questions. Faster calls. More capacity.
How an AI Agent Assist Bot changed the way Humach’s live agents get knowledge and support in real time, cutting handle time 8.5% and recovering 304 hours of capacity in one quarter.
- Comparison
- Same months, year over year
- Solution
- AI Agent Assist Bot
- Team
- Humach live agents
Customer wants to change the billing date on their account. What’s the process?
Illustrative example
Key results at a glance
Humach launched an AI Agent Assist Bot to give live agents answers in real time, instead of through internal chat with peers and supervisors. Compared with the same months the year before, agents posted 46.4% fewer questions, average handle time improved 8.5%, and the operation recovered an estimated 304 hours of capacity in one quarter, while quality and CSAT stayed strong.
Fewer agent questions29,300 posts down to 15,698
Fewer support chat messages164,931 down to 78,759
Lower handle time64 seconds less per call
Capacity recovered in one quarterAbout 7.6 full-time workweeks
Source: Humach internal results, year-over-year comparison of the same months. Capacity recovered is an estimate. AHT = average handle time; CSAT = customer satisfaction.
Every contact center has a hidden second conversation. So we handed it to an AI.
While one agent talks with a customer, another thread runs in the background: a question typed into an internal chat, a supervisor pulled away from their own work, an answer passed along while the customer waits.
Routine questions competed with customers
Agents rely on knowledge and support while they’re live with a customer. When that support comes from peers and supervisors over internal chat, every routine question costs time twice: once for the person asking and again for the person answering. And the load peaks when capacity matters most.
An AI Agent Assist Bot in the flow of work
The bot puts knowledge and support in front of agents in real time. Instead of posting a question and waiting, the agent gets the answer where they work, and supervisors stop being the default source for routine questions.
The questions stopped, and the capacity came back
Same months, before and after Agent Assist. The drop was strongest in peak season.
Questions posted to peers and supervisors −46.4%
Follow-on chat messages −52.2%
Fewer questions and responses
during peak season
Better for customers, lighter for the operation
Faster service, same quality
Agents who spend less time hunting for answers spend less time on the call. Average handle time improved 8.5% year over year, while quality and CSAT stayed strong.
More capacity, no added headcount
Agent questions fell 46% and supervisor support messages fell 52%, freeing supervisors for the work that needs them. In one quarter alone, the bot recovered an estimated 304 hours.
Operational context. These results came alongside 23.9% fewer new hires and improved attrition, from 28.2% down to 23.8%.
The bot didn’t replace anyone. It took routine questions off the plate, gave agents answers in the moment, and returned hours to the people doing the work.
Want to see what AI can do for your contact center?
We’ll walk you through how Agent Assist, digital agents, and AI Whisperers™ could work in your operation.
Explore more: ROI calculatorAI WhisperersDigital AgentsAll case studies
Questions about this case study
Quick answers about the bot, the results, and how they were measured.
What is an AI agent assist bot?
An AI agent assist bot gives contact center agents knowledge and support in real time, in the flow of work, while they’re live with a customer. Instead of asking a peer or supervisor and waiting, the agent gets the answer instantly from the bot.
How much did Humach’s Agent Assist Bot reduce average handle time?
Average handle time improved 8.5% year over year, or 64 seconds less per call, compared with the same months the year before, while quality and CSAT stayed strong.
How much capacity did the Agent Assist Bot recover?
In one quarter alone, the bot recovered an estimated 304 hours of agent and supervisor time, the equivalent of 7.6 full-time workweeks, without adding headcount.
How did the bot affect questions to supervisors and peers?
Agents posted 46.4% fewer questions to peers and supervisors over internal chat (29,300 down to 15,698), and follow-on chat messages fell 52.2% (164,931 down to 78,759). In peak season, questions and responses fell by 60%.
Did the Agent Assist Bot replace agents or supervisors?
No. The bot took routine questions off the plate and gave agents answers in the moment. Supervisors were freed for work that needs them, and the results came alongside 23.9% fewer new hires and attrition improving from 28.2% to 23.8%.
How were these results measured?
These are Humach internal results comparing the same months year over year, before and after the Agent Assist Bot launched. Capacity recovered is an estimate.