- An IT knowledge assistant finds the runbook step, the past ticket or the postmortem that already answers today's question, and shows the source so the engineer can check it.
- Enterprise h2oGPTe combines lexical and neural search, so exact error codes, hostnames and ticket numbers are found as well as plain-language matches.
- Document AI reads scanned PDFs, diagrams and tables, so old runbooks and vendor manuals become searchable in the same collection.
- The National Institutes of Health runs a knowledge assistant this way for its staff, air-gapped.
Your team keeps solving incidents it already solved
Your runbooks live in Confluence and SharePoint, your incident history lives in the ticketing system, and the lessons from every outage sit in postmortems nobody rereads. When an alert fires overnight, the engineer on call searches three places, then messages the person who fixed it last time.
The cost rarely shows up as a line item. It shows up as incidents re-solved from scratch, tickets escalated because the fix was never findable, new engineers who take months to work alone, and senior engineers answering the same questions every week.
How an IT knowledge assistant works
The pattern is retrieval-augmented generation: the assistant finds the passages that answer a question in your own documents, then writes an answer from them and cites them.
Step 1: connect the corpus
Bring in runbooks, knowledge articles, ticket exports and postmortems from SharePoint, Confluence, cloud storage, file systems or upload, with auto-sync so the assistant stays current.
Step 2: read every page
Document AI reads scans, network diagrams and tables with intelligent character recognition, and handles handwriting and image-heavy pages.
Step 3: find the exact match
Hybrid lexical and neural search finds the error code, the part number and the ticket ID as well as the plain-language description, and a re-ranker puts the best passages first.
Step 4: reason across sources
Agentic retrieval strategies can combine a runbook step, a past ticket and a postmortem finding into one answer when no single document holds it.
Step 5: answer with the source
The assistant writes a grounded answer and shows the document references it used, so the engineer checks the step before running it.
Making every lesson findable
Postmortems hold the root cause, the fix and the follow-up actions for every incident you have already paid for. The assistant indexes them next to the runbooks, so a question about a failing service returns the last postmortem on that service, with the section cited. Collections are private by default, shared with named users or groups, and the corpus is indexed in place, in any of four deployment options.
Results
The National Institutes of Health runs an Enterprise h2oGPTe knowledge assistant across 28 institutes and centers, air-gapped and built in weeks. AT&T added H2O AI Super Agent to power the next phase of its agentic AI initiatives and Ask AT&T, its knowledge assistant.