Service 01

RAG Assistants & AI Agents

An assistant that has actually read your files, and an agent that can act on what it finds — with retrieval built so the answer is traceable back to the document it came from.

All services

Who this is for

Teams sitting on a body of private knowledge — precedents, policies, contracts, past matters — that is technically searchable and practically unsearchable.

LangChainLangGraphPineconeChromaFAISSFastAPI

answer

retrieve → reason → act

The problem

General models know everything except the thing you need: what is in your files.

Naive retrieval returns plausible passages rather than correct ones, and a confident wrong answer is worse than no answer at all.

Once more than one client's data is involved, retrieval has to prove it cannot reach across the boundary — filtering results after the fact is not an answer.

How it's built

Five stages, and a human at the end.

Each stage is something you can watch run. Nothing here is a black box you're asked to take on faith.

Ingest

Documents are parsed, chunked with structure preserved, embedded, and written to a vector store scoped to the tenant or matter they belong to.

Isolate

Access is enforced at the storage layer, not by post-filtering — the same pattern I worked out in the open-source multi-tenant RAG build.

Retrieve

Queries are expanded and reranked so the passages that reach the model are the ones that answer the question, not merely the ones that share its vocabulary.

Ground

Answers cite the source passage. If the retrieval is weak, the assistant says so instead of filling the gap with fluent invention.

Act

Where it's warranted, the assistant becomes an agent: updating the CRM, triggering a workflow, opening a task — with the boundaries of what it may do written down and enforced.

What you get

A document ingestion pipeline you can re-run
A vector store with per-tenant or per-matter isolation
A retrieval layer tuned and evaluated on your own questions
Citations on every answer
Defined action boundaries for anything agentic
An interface your team will actually open — chat, or inside a tool they already use

Proof

No client case study is published for this service yet. The engineering behind it is public, though — the multi-tenant retrieval work is open source and you can read the code.

See the build

Next service

Document & Intake Automation

Assistant over internal documents

Interactive demo

Answers drawn only from approved internal documents, with the source shown — and a refusal when the answer isn't in them.

The three documents below are the assistant's entire knowledge base, invented for this demo. No account needed, no personal data collected.

Knowledge base — 3 fictional documents

Internal · last updated 12 Jan
Before kickoff
Signed order form filed against the account. Access request sent for the client's mail and CRM. A named approver is recorded on the account before any workflow is switched on — no build starts without one.
Week one
Kickoff call, then process mapping with the people who do the work rather than their manager. The output is a written map the client signs off before anything is built.
Going live
Workflows run in review-only mode for the first two weeks. Nothing sends automatically during that window.
Internal · last updated 3 Feb
Legal — email automation
US law firm. Routine client, court and counsel email went from 7–10 minutes per reply to 1–2 minutes to review and send. Approved for external use.
SaaS — not yet available
No approved SaaS case study is currently in the library. Pending client sign-off.
Internal · last updated 20 Jan
Engagement shapes
Fixed price per build, or hourly for ongoing work. The figure is agreed before work starts.
Scope changes
A change in scope is re-quoted in writing before it is started, never absorbed silently.

Ask a question

Parse the question
Search approved documents
Check the passages support an answer

Pick a question. One of the three has no answer in the documents.

Walkthroughs

Watch this one run.

The same workflow as a walkthrough. Written out in full underneath, so you can read it instead.

2 min

Assistant over internal documents

Recording in progress — read the walkthrough below

Assistant over internal documents

Staff get answers with sources, and a clear refusal when the answer isn't there.

1.An employee asks a question in plain language.
2.The system searches only the documents you have approved for it.
3.The answer comes back with the passage it came from, named and quoted.
4.The employee opens the source document to check it.
5.Asked something the documents don't cover, the assistant says so instead of inventing an answer.

Book a workflow assessment Try the interactive version

Losing hours to work an AI system could handle?

Tell me about the process. I'll tell you straight whether it's worth automating — no pitch, no obligation.

Book a free audit call