Mon–Fri: 9:00 AM – 6:00 PM (UK)
All case studies
Cytiva
Healthcare & Life Sciences
8 weeks

A life-sciences chatbot that answers 65% of product and technical queries, grounded only in approved documentation, with a citation on every answer

We built a grounded AI chatbot for a life-sciences business that now handles an estimated 65% of product and technical questions, answering strictly from the company's approved documentation, citing the source on every response, and deferring to a human whenever it lacks a grounded answer.

The challenge

A life-sciences company wanted an assistant that could answer detailed product and technical questions, but in healthcare, a chatbot that hallucinates is worse than none at all. Answers had to be grounded in approved source material, traceable to a citation, and safe within a regulated context. An off-the-shelf general-purpose chatbot, with no grounding and no audit of what it told users, carried unacceptable accuracy and compliance risk.

What we built

1

Indexed the company's approved documentation into a retrieval-augmented (RAG) knowledge base so answers are grounded in real, controlled sources

2

Engineered the assistant to cite its sources, so every answer is traceable back to an approved document

3

Added guardrails and fallbacks so the assistant defers or escalates to a human rather than guessing when it lacks a grounded answer

4

Built a human review and feedback loop so subject-matter experts correct and improve responses over time, with interactions logged for oversight

Architecture

Cytiva architecture diagram

Outcomes

The chatbot now handles an estimated 65% of product and technical queries, grounded only in approved documentation with a citation on every answer and zero ungrounded responses reaching users

Every response traceable to an approved document, critical for a regulated industry

Hallucination risk contained through retrieval grounding, guardrails and human escalation

A feedback loop that improves the assistant under continuous expert oversight

Results

63%

of queries handled

100%

answers cited to source

0

ungrounded responses to users

41%

fewer support tickets

Stack
AWS Bedrock
Knowledge Bases for Bedrock
RAG
OpenSearch
AWS Lambda
Ready when you are

Ready to take AI from pilot to production?

Book a free discovery call with our team of ex-AWS engineers. We'll discuss your regulated-industry challenges and outline a governed path from pilot to production. No commitment required.