What You Will Learn
✅ How Universitas Terbuka handled 200,000+ student conversations in just 3 months with Qiscus AgentLabs, without scaling headcount
✅ How AgentLabs handles 92% of student inquiries end-to-end, with only 6.5% requiring human escalation
✅ How UT cut its first response time by 98%, from 10-15 minutes to 20 seconds, across four channels, 24/7
✅ How a 760,000-student university optimized its service team from 20 agents down to 10, with the rest moved to higher-value academic functions
About Universitas Terbuka
Universitas Terbuka (UT) is Indonesia’s national distance learning university, serving 760,000+ active students across every province in the country. Unlike conventional campuses where students gather physically and access services in person, UT’s student body is national, distributed, and often non-traditional. Working professionals, parents returning to study, and learners in remote areas balance education with the rest of their lives.
For a university operating at this scale, the contact center isn’t a support function. It is, effectively, the campus. Every interaction a student has with the institution, whether asking about graduation dates, course codes, PMB registration, or payment deadlines, happens through a digital channel. If that channel is slow or unanswered, the institution itself starts to feel distant.
That structural reality is what made UT’s student service operation the highest-leverage problem to solve. Improving it didn’t just mean better service scores. It meant making a 760,000-student university feel responsive and present.
The Challenges Universitas Terbuka Faced
Before adopting Qiscus AI, UT’s student support ran almost entirely on two channels: phone and ticketing. Both were saturated.
Phone volume reached 500 to 600 calls per day, and a meaningful share were abandoned simply because the team couldn’t pick up fast enough. Tickets routinely breached the 3-day SLA, with end-of-day backlogs as a daily reality. First response time during peak periods stretched to 10 to 15 minutes, long enough for a frustrated student to give up and try again later, doubling the volume.
Even with a 20-person service team, the operating model couldn’t absorb the load. Night-time traffic was significant, predictable given that many UT students work or study after hours, but had no coverage at all. The team’s day started behind, and ended behind.
UT had tried automation before. A previous bot deployment underperformed, with a low resolution rate that pushed almost everything to human agents anyway. A separate AI test for Instagram comment replies was shut down because the questions were too unstructured and the pricing didn’t scale.
The conclusion was clear: the problem wasn’t effort, and it wasn’t team size. It was that the operating model itself couldn’t grow with the student body.
What Universitas Terbuka Aimed to Achieve
UT didn’t approach Qiscus with a vague mandate to “improve student services.” The team had specific, measurable problems to solve, and the objectives were defined accordingly.
1. Make AI the default first responder across all student channels.
The first step was acknowledging that phone and ticketing alone could not serve a 760,000-student body. UT needed to meet students where they already were, on WhatsApp, live chat, Instagram, and TikTok, and put AI in front to handle first response and resolution wherever it could. The goal wasn’t just to add channels. It was to make sure that opening one didn’t mean adding another set of overwhelmed humans behind it.
2. Automate the repeatable layer of student inquiries.
A large share of student questions at any university are predictable: when do I graduate, what’s my course code, how do I register for PMB, and when is the payment deadline. These are information-based questions, not judgment calls. UT wanted AI to absorb that entire layer so the human team could focus exclusively on cases that required system access, escalation, or empathy. The mental model was simple: humans for human problems, AI for everything else.
3. Deliver a sub-30-second first response, consistently.
The 10 to 15 minute peak wait time wasn’t just a service quality issue. It was actively creating duplicate volume, as frustrated students gave up and re-contacted later. UT wanted near-instant response as the baseline, 24/7, regardless of time of day, conversation volume, or sudden demand spikes around exam periods and registration deadlines. The target was structural reliability, not occasional speed.
4. Hold answer quality as the primary KPI.
UT’s operating principle is that fast-but-wrong is worse than nothing. A wrong answer about a graduation date or registration deadline costs the student real time and can erode trust in the institution. So while speed mattered, it could never come at the expense of accuracy. The AI had to be measurably correct before it was allowed to be fast.
5. Reshape the operating model, not just cut costs.
The most important objective wasn’t headcount reduction. It was unlocking the existing team to do work that mattered more. UT wanted AI to absorb the repetitive, high-volume layer so the service team could be redirected into roles where their academic knowledge and judgment actually moved the needle, not into work that AI could do equally well.
Solutions Implemented
UT partnered with Qiscus to rebuild the student support operation around AI as the primary layer, with humans concentrated on the cases that genuinely required them. The implementation wasn’t a single feature switch. It was three deliberately chosen products working together to handle different parts of the same problem.
AgentLabs (AI Agent)
AgentLabs is the engine that powers UT’s AI-first student support across all active channels. It handles information-based queries end-to-end, including graduation dates, schedules, course codes, and PMB registration, without any human involvement. Crucially, AgentLabs doesn’t just answer. It resolves, meaning the conversation closes inside the AI layer without a handover, escalation, or callback. The result is that 92% of student conversations now reach completion without a human ever touching them, and only 6.5% require agent intervention.
Omnichannel Chat
Even with strong AI, a fragmented inbox would have undone the gains. Before Qiscus, UT’s student communication was spread across siloed surfaces, with no shared context between channels. Omnichannel Chat centralizes live chat, WhatsApp, Instagram (DM and comments), and TikTok (DM and comments) into one unified inbox. When AI does need to hand off to a human, the routing is instant, the context is preserved, and the agent picks up where AI left off without asking the student to repeat themselves.
WhatsApp Business API
WhatsApp anchors UT’s second-highest-volume student channel and is the clearest example of how AI transforms a single touchpoint. With WA Business API powering the channel and AgentLabs handling first response, 92.7% of all incoming WhatsApp conversations are now fully resolved by AI, without any human involvement at all. For a channel where students used to wait through long, code-based bot flows, the experience is now a direct conversation that ends in a real answer, usually within seconds.
How Universitas Terbuka Leveraged the Solution
The shift in UT’s operating model isn’t just visible in the metrics. It’s visible in how the team’s day looks now compared to how it looked two years ago.
Before Qiscus, the service team’s day was reactive from the moment it started. The phone rang 500 to 600 times before lunch. Tickets stacked up faster than they could be resolved. Twenty people worked at capacity, and by end of day, the backlog hadn’t shrunk; it had just rolled into tomorrow. Night-time hours were a black hole. There was no coverage, no AI, and no continuity. Students who reached out after 6 PM essentially had to wait until the next business day, by which point they had often given up.
After Qiscus, the same operation looks fundamentally different. AgentLabs is the first point of contact across every channel, every hour of every day. In just three months of implementation, AgentLabs handled more than 200,000 student conversations, with 92% resolved by AI end-to-end and 84.8% completed without ever requiring agent escalation. The seven human agents who remain on the team don’t deal with the routine layer anymore. They focus exclusively on the small slice of cases that need system access or judgment, which means their work is more meaningful and significantly less repetitive than before. Thirteen of their former colleagues have been moved to other functions inside UT, where their academic knowledge does work that AI cannot.
| Before | After | |
|---|---|---|
| Primary channels | Phone + ticketing only | Live chat, WhatsApp, Instagram, TikTok |
| Total conversations | Capacity-limited, manual | 200,000+ in 3 months, AI-managed |
| AI handling rate | None | 92% |
| WhatsApp AI resolution | N/A | 92.7% fully resolved by AI |
| Handover to human agents | Majority of conversations | Only 6.54% |
| AI accuracy (resolved without escalation) | N/A | 84.8% |
| Service availability | Business hours | 24/7 automated |
| First response time | 10 to 15 minutes at peak | 20 seconds |
| Service team size | 20 agents, still overwhelmed | 10 agents (10 moved to other functions) |
| Ticketing SLA | Regularly missed (3-day target) | Same-day resolution |
What’s notable about this transformation isn’t just that the team got smaller. It’s that the work got bigger while the team got smaller. Channels expanded from two to four. Total conversations crossed 200,000 in just three months, with no ceiling visible. First response time collapsed from 15 minutes to 20 seconds. And yet the active service team shrank to seven, because most of what the original twenty were doing simply no longer required a human. The capacity wasn’t removed. It was redirected to where it mattered more.
UT Students’ Journey
Universitas Terbuka’s adoption of Qiscus didn’t start as a full student-services overhaul. It started narrowly, using Qiscus purely as a tool to close marketing leads with prospective students via WhatsApp. The early integration was deliberately small in scope, focused on giving the admissions team a faster channel to convert PMB inquiries into enrolled students. That phase did two things at once: it proved the channel worked, and it gave UT a working understanding of how WhatsApp at scale actually behaves in their environment.
In practice today, the operation looks almost unrecognizable from where it started. Live chat carries the highest traffic overall, with WhatsApp close behind, and Instagram and TikTok rounding out the active channels. AgentLabs autonomously resolves the 92% of inquiries that are information-based, including questions about graduation, schedules, course codes, and registration, while human agents handle only the 6.5% of cases that require system access. A three-person QC team samples 100 to 150 AI conversations per agent per day to maintain quality.
The turning point came when UT realized the seven-person team was handling more total volume than the twenty-person team ever had. The math had inverted: more channels, more conversations, faster responses, smaller team. That was the moment the operating model had visibly changed.
Measurable Outcomes
200,000+ conversations handled by AI in just 3 months.
Within just three months of implementing Qiscus AgentLabs, the AI handled more than 200,000 student conversations across live chat, WhatsApp, Instagram, and TikTok. Without AI, managing this volume manually would require substantial operational resources, increased staffing capacity, and significantly higher service costs. Instead, the institution scales with student demand instead of against it.
92% AI handling rate across all student conversations.
AgentLabs autonomously handles more than nine out of every ten student inquiries end-to-end, without any human involvement. For an institution of UT’s scale, this is the difference between a service desk that can absorb growth and one that can’t.
84.8% of AI conversations resolved without agent escalation.
Of every conversation AgentLabs takes on, the vast majority are fully understood, answered, and closed by AI without ever needing a human to step in. Only the remaining slice gets handed over, which means the human team’s bandwidth is reserved for the cases that genuinely need it.
98% drop in first response time, from 10-15 minutes to 20 seconds.
Students used to wait long enough that many gave up and re-contacted, doubling the volume. AgentLabs now delivers a first response in 20 seconds on average, consistently, regardless of time, volume, or demand spikes. For students working or studying after hours, the difference between waiting overnight and getting an answer in under half a minute is the difference between feeling supported and feeling stranded.
92.7% of WhatsApp conversations are fully resolved by AI.
WhatsApp is UT’s second-highest-traffic channel, and more than nine out of every ten incoming inquiries on it are now resolved by AI without any human involvement. This is a structural shift in how UT operates one of its busiest student touchpoints.
The service team was optimized from 20 agents to 7.
The remaining 13 team members have been moved to other functions where their judgment and academic knowledge create higher value. This isn’t a cost-cutting story. It’s a talent reallocation story, where the same people are now doing work that actually requires them.
Same-day ticket resolution, replacing chronic 3-day SLA breaches.
UT’s ticket SLA used to be missed routinely, with backlogs carrying overnight as a daily condition. Tickets are now resolved the same day they come in, with zero end-of-day backlog.
Hear from Universitas Terbuka
| “We’re not a conventional university. With 760,000+ students spread across Indonesia, our entire campus runs through digital channels. Before Qiscus, our team was working hard but always behind. Now AgentLabs handles the volume that used to overwhelm us, and our service team focuses on the cases that genuinely need them. The success we’ve seen at UT Pusat is now being looked at by other UT branches, who are exploring Qiscus for their own operations.”Kukuh Nur Ihsan, Universitas Terbuka Pusat |
Achieve Similar Success with Qiscus
UT’s transformation didn’t come from a single feature. It came from a deliberate sequence: start narrow, prove the channel, then expand AI coverage as the knowledge base matured. For institutions managing hundreds of thousands of students across multiple digital channels, that sequence is repeatable.
If your service team is firefighting volume across phone, WhatsApp, and social channels, and adding headcount no longer solves it, there is a different operating model available. Talk to our team and see how Qiscus can work for your institution.
