It support · Production

NIH deploys H2O.ai h2oGPTe Business Assistant to deflect 10,000 IT service requests annually

The problem

NIH support teams fielded a high volume of repetitive 'how-do-I?' questions while staff and researchers struggled to find answers scattered across hundreds of SharePoint documents. Search was slow and maintaining current institutional knowledge consumed valuable time.

Workflow diagram · grounded in source

SharePoint document ingestion

integration

“Indexes every user guide, training deck and policy in SharePoint”

Employee submits question

trigger

“Business Assistant chatbot is embedded directly in the SharePoint intranet”

RAG surfaces policy answers

ai_action

“surfaces precise policy and procurement answers in seconds”

Grounded response with source links

output

“Summaries are returned with live hyperlinks to the original source”

Context-aware session personalization

feedback_loop

“Remembers previous questions to personalize follow‑ups”

Reported outcome

The h2oGPTe-powered Business Assistant serves 8,000 NIH employees and deflects up to 10,000 annual service requests, freeing support teams to focus on higher-value tasks, with all data remaining within an air-gapped NIH network.

Reported metrics

NIH users served8,000

Service requests deflected per year10,000

Support team time savingsSupport teams focus on higher-value tasks

FAQ handling time freedfreeing our team to concentrate on mission-critical improvements instead of repeating FAQs

Reported stack

h2oGPTeOracle Cloud Isolated Region (OCIR)SharePoint

Source

https://h2o.ai/case-studies/nih/

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The h2oGPTe-powered Business Assistant serves 8,000 NIH employees and deflects up to 10,000 annual service requests, freeing support teams to focus on higher-value tasks, with all data remaining within an air-gapped N…

What tools did this team use?

h2oGPTe, Oracle Cloud Isolated Region (OCIR), SharePoint.

What results were reported?

NIH users served: 8,000; Service requests deflected per year: 10,000; Support team time savings: Support teams focus on higher-value tasks; FAQ handling time freed: freeing our team to concentrate on mission-critical improvements instead of repeating FAQs (source-reported, not independently verified).

How is this it support AI workflow structured?

SharePoint document ingestion → Employee submits question → RAG surfaces policy answers → Grounded response with source links → Context-aware session personalization.