How Navitas lifted conversion to 43% with AI

The challenge

Navitas works in study-abroad recruitment. It works directly with students and with B2B partners, helping students choose schools, apply and find scholarships.

As enquiries grew, the team faced:

• Consultants stretched thin, answering the same questions about schools, fees and requirements again and again.
• Slow replies, especially outside office hours, when students often look for help.
• Lost students: slow answers meant some went to competitors.
• Peak seasons around intakes and deadlines, which meant hiring or overtime.
• Data ownership: student and school data had to stay with the company, not a third-party tool.

What Bytenamics did

We built an AI chatbot on Amazon Bedrock that helps students from first question to application, and hands them to a consultant when it matters.

1. Bring the knowledge together

School profiles, programs, fees, entry requirements and scholarship details are stored in Amazon S3 and used as the chatbot's knowledge base, updated weekly from the team's school and scholarship database.

2. Choose the right model for each task

The chatbot uses Anthropic Claude and DeepSeek on Amazon Bedrock, matching each type of question to the model that answers it best for the cost.

3. Help students end to end

Students can find schools that fit them, compare schools side by side, check scholarships and how to apply, get school information and ask any other questions — any time of day.

4. Keep data in the company's hands

Everything runs in the customer's own AWS account. Student conversations and school data stay with them and are not used to train public models.

5. Scale with demand

The app runs on Amazon EC2 behind an Application Load Balancer and scales with the number of queries, so peak seasons need no extra staff.

1. Week 2: DiscoveryTop student questions, data sources and success metrics agreed.
2. Week 5: PrototypeChatbot answering school and scholarship questions.
3. Week 8: LaunchLive for students on the website, Zalo and Facebook Messenger.
4. Week 12: First peak seasonHandled intake demand without extra staff.

The results

• Conversion up from 32% to 43%, as students get answers while they are still interested.
• Less consultant time on routine questions, so consultants focus on applications and closing — about 40% less time per student.
• Faster replies to every student — the first reply went from about 4 hours to under 1 minute.
• 24/7 support, improving student satisfaction.
• Scales at peak times without adding people.
• Data stays with the company, keeping it competitive in the market.

What's next

Next, Navitas plans to add more languages and application tracking for students.

Want an AI assistant for your customers? See our generative AI and RAG services.