Case Studies / Enterprise AI Hub

CASE STUDYHEALTHCARE ENTERPRISEAGENTHUB

An enterprise AI hub that reviews contracts, not just stores prompts.

Built for a global healthcare company's internal innovation program under the Maslow AI brand: a department-organised prompt library as the front door; behind it, a Statement-of-Work review agent with graph-augmented retrieval, field-level citations, and a chat that answers in charts, tables and drafted documents instead of walls of text.

generative-UI widgets the agent can answer with

93%

intent-routing accuracy on a 28-case live test suite

SOW corpus, chunked by document structure

departments in the curated prompt library

THE CHALLENGE

Two problems wearing one interface

Enterprise knowledge workers had no shared home for the prompts that actually work, and no fast, trustworthy way to interrogate dense Statement-of-Work documents. "What are the termination terms?" meant an afternoon of reading. "Where's the duplicate-spend risk across these vendors?" meant nobody checked.

So AgentHub is two products in one: a curated, community-style prompt library organised by department as the front door, and an agentic SOW reviewer as the engine, grounded in the documents themselves.

THE FRONT DOOR

Department prompt library

Eleven colour-coded departments, real-time search, likes and comments, public/private prompts, per-author categories. Every prompt opens straight into the AI chat; the library is how people discover what the agent can do.

MarketingLegalR&DFinanceSupply Chain+6 more

THE ENGINE

Agentic SOW reviewer

Ask a contract corpus questions in plain English (“compare the day rates across these SOWs”) and get grounded answers with citations to the exact field in the source document, via parallel dense-vector and knowledge-graph retrieval.

▸ VECTOR SEARCH · top-5 · 3072-dim embeddings▸ KNOWLEDGE GRAPH · entities + one-hop neighbours▸ CITATIONS · down to the exact jsonPath field

DEEP DIVE · HYBRID RAG

From 50 contracts to field-level answers

Parallel dense-vector and knowledge-graph retrieval, fused as grounded context, with every answer citing the exact field it came from. If the graph is unavailable, retrieval degrades gracefully to vector-only.

01INGEST

Structure-aware chunking

Contracts are chunked by document structure (not fixed token windows), with breadcrumb metadata (section path, field names) preserved on every chunk.

02EMBED

3072-dim embeddings

Each chunk is embedded and stored behind a native vector index: cosine similarity over the full corpus.

03RETRIEVE ×2

Two legs in parallel

Dense vector search (top-5) runs alongside a knowledge-graph query for matching entities (vendors, clients, deliverables) plus their one-hop neighbours.

04FUSE

Grounded context fusion

Both result sets enter the prompt as labelled context sections, each vector hit carrying its similarity score and source path.

05GENERATE

Agentic tool loop

The model answers in a multi-round tool loop, streaming tokens, reasoning and widgets live over server-sent events.

06CITE

Field-level citations

Drafted documents cite per-section back to the exact field of the source SOW, auditable by a procurement or legal reviewer.

PROVENANCE, VISIBLERetrieval telemetry surfaces as data-source badges on every answer: “Vector Search · 5 results · top score 0.82”, “Knowledge Graph · 12 entities”; a live Activity Panel shows the agent's searches, tool rounds and prompt modifications as they happen.

DEEP DIVE · INTENT ENGINEERING & GENERATIVE UI

The agent decides what kind of answer to give, before it answers

Every message is routed through a deterministic intent layer: tool request, document drafting, data-heavy, narrative, off-topic or conversational. That classification gates which tools the model is even offered. Say hello, and it can't hallucinate a chart at you.

SEVEN WAYS TO ANSWER

show_chartBar, line, area and pie charts in brand colors

show_data_tableSortable column/row comparisons

show_statsKPI grid with deltas and trend arrows

show_planVertical stepper with task states

show_optionsClickable choices that feed back into the chat

ask_questionsMulti-step clarifying wizard before drafting

generate_documentEditable sectioned docs with per-section citations

TESTED, NOT VIBED

26 / 28

intent→tool cases pass through the live streaming pipeline (~93%), asserting the first tool the agent reaches for.

“put the comparison in a table” → show_data_table

“draft a PRD” → ask_questions first, then the document

“what's the weather in Boston?” → text only, no tool

BUILT FOR A REGULATED ENTERPRISE

✓Explainable by design: the Activity Panel makes retrieval and reasoning visible, not a black box

✓Field-level citations: auditability that procurement and legal reviewers can check

✓Graceful degradation: graph down means vector-only, never a silent failure

✓Tool-hallucination guardrails: function-calling disabled for conversational turns

✓Real test coverage: the 28-case intent harness plus browser test specs, run against the live pipeline

UNDER THE HOOD

Frontend: React + TypeScript, Tailwind, assistant-ui with a custom streaming runtime, Recharts

Data: Convex reactive DB with native vector index · Neo4j knowledge graph · Clerk auth with post-sign-in intent replay

AI: Gemini 3 Flash (chat, thinking, tool-calling) · gemini-embedding-001 at 3072 dimensions · optional MCP tool discovery at runtime

Delivery: serverless; the whole chat/RAG/tool loop is one streaming action; no separate RAG service to operate

Precisely stated: field-level-cited, graph-augmented SOW review with an intent-routed generative-UI agent. Dense + graph retrieval fused as grounded context; intent routing is engineered orchestration, not an ML classifier.

SERVICES USED

This engagement, as catalog services

Knowledge foundation build Custom agentic harness Skills authoring Team enablement

“

The prompt library is the front door. The engineering story is a contract-review agent that shows its sources, down to the exact field.

Live demo deployed · 11-department library · 50-SOW corpus · adoption metrics in progress

Have documents your teams are afraid to ask questions of?

We build grounded, citable, explainable review agents on your data, behind your walls.

BOOK A CONSULTATION

← Infinite AI OS All case studies