evidoria

← Back to browse

Good practice Imported

Lyon Bot — NTU Singapore's AI Virtual Assistant Evolves from Dialogflow to Gemini for Student Services

Singapore · Singapore · See the Singapore profile · See the Singapore profile

Evidence: Descriptive / self-reported Top 89% 27/100 · Ask Evidence Copilot about this practice

Nanyang Technological University's Lyon Bot, built on Google Cloud's Dialogflow, launched in 2020 to handle freshmen queries. A 2025 Gemini-based upgrade now saves NTU staff an estimated 14.5 days of administrative work per month.

~6,000
Freshmen served at launch (2020)
up to 20,000
Query intents handled (Dialogflow Mega Agent) (2020)
~14.5 days
Staff administrative time saved per month (2025 Gemini upgrade) (2025)
Lyon Bot — NTU Singapore's AI Virtual Assistant Evolves from Dialogflow to Gemini for Student Services

Details

Maturity
Established
Promoter
Nanyang Technological University (NTU Singapore), Google Cloud, CloudMile
Period
2020-2025
Keywords
higher education, student administration, generative AI, cloud computing

Context

NTU Singapore, in partnership with Google Cloud, launched the "Lyon" virtual assistant in August 2020 to help roughly 6,000 incoming freshmen navigate course registration, timetables and campus services, becoming the first higher-education institution in Southeast Asia to deploy Google Cloud's Dialogflow Mega Agent.

Objectives

The initiative aimed to give incoming students round-the-clock automated support for administrative queries, and later to move from a purely intent-based chatbot to a hybrid generative-AI approach capable of handling more complex enquiries such as student housing.

Activities

The original Dialogflow-based bot processed up to 20,000 distinct query intents; in 2025, NTU partnered with AI vendor CloudMile to rebuild the chatbot on Google's Gemini large language model.

Results

NTU's CIO reported the Gemini-based upgrade now saves the university around 14.5 days of staff work every month, alongside improved conversational accuracy and reduced error rates.

Conclusions

Public reporting documents adoption scale and operational time savings, but no independent third-party evaluation of response accuracy or student-satisfaction outcomes has been published, so the effect on student learning, as opposed to administrative efficiency, remains undemonstrated.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Long (> 3 years)
Staffing & skills
Google Cloud technical team supporting the original Dialogflow deployment, AI vendor CloudMile leading the 2025 Gemini-based rebuild, NTU IT/CIO office overseeing integration into student services

Conditions for success

  • Enterprise cloud infrastructure (Google Cloud Dialogflow, later Gemini) capable of handling large intent/query volumes
  • An ongoing vendor partnership to migrate from an intent-based to a generative-AI architecture as technology evolves
  • Institutional buy-in from university IT leadership to sustain investment over multiple years

Common failure modes

  • No independent evaluation of response accuracy or student satisfaction has been published
  • Effectiveness is reported by the institution itself (its CIO), not by a third party

Commonly funded by

National / regional programmes Own resources / municipal budget

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful