Before its intelligent virtual agent went live, a multi-site early childhood education provider was losing nearly one in three chat conversations. Of 5,230 chats in a 77-day window, 926 ended with a family leaving before anyone answered: a 30.6% abandonment rate.
The problem wasn’t effort. The same team handled family inquiries across dozens of school locations, and after-hours conversations had no path to an answer. The chat channel had hit its human capacity ceiling.
The provider added GuruConnect as the first point of contact for every inbound chat. Over the next 77 days, the virtual agent resolved 71.7% of chats on its own, abandonment dropped to zero, and the same team absorbed 58% more volume. Here’s how it worked, what the numbers show, and what to measure if you’re considering a virtual agent of your own.
What Is an Intelligent Virtual Agent? How GuruConnect Works
An intelligent virtual agent is an AI-enabled assistant that understands what a customer needs, completes tasks, and brings in a human agent when the conversation calls for one. In contact centers, it works across voice and digital channels and goes well beyond a fixed list of answers.
Intelligent virtual agent vs. chatbot
A basic chatbot follows a script. It matches keywords to prewritten answers and tends to stall when a question falls outside that script.
An intelligent virtual agent identifies the intent behind each message. It retrieves information, completes workflows like scheduling, and hands off the conversation to a live agent with context when the conversation needs human judgment.
GuruConnect is our intelligent virtual agent solution for voice and digital channels. It’s built to:
- Answer questions and retrieve information through natural, conversational exchanges.
- Complete supported workflows, such as verification and appointment scheduling, through configured integrations.
- Connect customers with a live Guru when a conversation needs judgment, empathy, or complex problem-solving, with interaction context where integrations support it.
- Be tested before launch with GuruLab simulations that evaluate responses, workflows, and escalation scenarios.
- Improve over time through automated QA and conversation intelligence that track containment, failed interactions, and transfer reasons.
Our team also helps identify which inquiries are right for automation in the first place, using real interaction data instead of assumptions.
How the virtual agent was set up in this deployment
The provider deployed GuruConnect on chat, covering the same channels and schools its agent team already handled. The virtual agent became the first point of contact for every inbound conversation.
It was configured for the full range of family inquiries: FAQs, pricing questions, program and services information, enrollment requests, and tour scheduling through self-serve booking links. In each conversation, it identified the reason for the contact, collected family information, and qualified whether the family was ready for the next step.
The system transferred families ready to schedule a visit to an agent, pre-qualified and with full context. The virtual agent handled everything else, including after-hours inquiries.
We measured results across two equal 77-day windows: April 6 to June 21, 2026, before launch, and June 22 to September 6, 2026, after.
The Results: 77 Days Before vs. 77 Days After Launch
The virtual agent took chat containment from 0.5% to 71.7% while total volume grew 58%.
| Metric | 77 days before | 77 days after | Change |
| Total chats received | 5,230 | 8,281 | +58.3% |
| Chats resolved by the virtual agent | 25 | 5,938 | +5,913 chats |
| Containment rate | 0.5% | 71.7% | +71.2 pts |
| Chats handled by an agent | 4,055 | 2,008 | −50.5% |
| Chats abandoned | 926 | 0 | −100% |
| Abandon rate | 30.6% | 0% | −30.6 pts |
| Turnaround rate | 61.6% | 80.07% | +18.5 pts |
| Lead created rate | 49.77% | 67.13% | +17.4 pts |
More volume, fewer chats for agents: total chat volume grew 58%, yet the chats that reached an agent fell by half, from 4,055 to 2,008. The same team handled it all: four dedicated agents and two floaters.
Every family got an answer: abandonment fell from 30.6% to zero, well past the 5% target. Turnaround rate rose to 80.07%, against a 60% target.
Conversion crossed its goal for the first time: the lead creation rate climbed 17.4 points to 67.13%, past the provider’s 65% goal.
After-hours demand, captured: the virtual agent handled about 190 after-hours family contacts a week, answering questions and sharing a self-serve link to request a tour. Across seven weekly recaps, its resolution rate held between 95% and 98%.
See it in action. Watch how the virtual agent handled thousands of family conversations, from the first question to a scheduled visit.
Containment by Design: Why a Quarter of Chats Still Reach a Person
A 71.7% containment rate means the remaining conversations still reach people, and that’s by design. In this deployment, 2,008 of 8,281 chats, about a quarter, went to an agent on purpose.
Those transfers were the point. The virtual agent sent families who were ready to schedule a visit after identifying their needs and collecting their information. Agents picked up each conversation with full context instead of starting from zero.
That changed the shape of the agents’ day. Routine questions no longer competed with enrollment conversations for attention. Every chat that reached a person was a high-intent conversation worth having.
The lesson applies to any deployment. Containment rate alone doesn’t tell you whether a virtual agent is working. The better question is whether the conversations reaching your team are the ones that need a person.
What to Measure in Your Own Virtual Agent Deployment
A virtual agent changes how work flows through call centers and support operations, so no single number shows whether it’s succeeding.
| Metric | What it tells you | What to watch |
| Containment rate | The share of conversations resolved without a transfer | Read it alongside handoff quality, never on its own; chatbot containment rate on a script tells a different story than an intelligent virtual agent’s |
| Abandon rate | How many customers leave before getting an answer | The clearest signal of demand your operation is missing |
| Handoffs with context | Whether the right conversations reach people | Transfers should carry customer data and intent |
| Conversion metric | Whether automation moves business outcomes | Use the one your business runs on: leads, bookings, or sales |
| After-hours contacts | Demand you used to miss | Make sure self-serve actions taken after hours are tracked |
| Agent handle time | How agent workload is shifting | Read it alongside containment and conversion, since agents now focus on higher-value conversations |
| Intent mix | What customers actually ask about | Use it to tune content, configuration, and training |
Intent data deserves special attention, because the virtual agent generates it from every conversation. In this deployment, FAQs made up 70% to 79% of weekly threads, and pricing accounted for 42% to 51% of activity. That signal led us to tune pricing answers by age group, giving supervisors a continuous view of what families need.
Frequently Asked Questions
A basic chatbot answers set questions from a script. An intelligent virtual agent identifies the intent behind each message, completes tasks like scheduling or information retrieval, and hands the conversation to a live agent with context when a person is needed.
Yes. It captures and qualifies every inquiry, not only the ones agents have time to answer. An intelligent virtual agent identifies the reason for each contact, collects key details, and qualifies intent, so teams receive leads ready for the next step at any hour.
When the conversation needs human judgment, empathy, or complex problem-solving, or when the customer is ready for a high-value next step, like booking or buying. The best handoffs pass along the customer’s details and intent, so the agent can pick up without asking the customer to repeat anything.
Yes. A virtual agent can answer questions and complete supported tasks outside staffed hours. It can also capture contact details or share self-serve links, such as booking requests, so overnight demand isn’t lost.
No, it changes their workload. A virtual agent resolves routine, repetitive inquiries, so live agents spend more of their time on complex, high-intent conversations that need human judgment.
See Whether a Virtual Agent Fits Your Volume
A virtual agent delivers when it matches the right conversations. This deployment worked because routine inquiries went to automation, high-intent conversations went to people, and context traveled between them, creating faster, more consistent customer experiences.
If routine questions are crowding out your highest-value conversations, a virtual agent may be the fastest way to give your team that time back. When you’re ready to review your volume, after-hours demand, and conversion goals, our team can help you see where a virtual agent fits.
