Free · 3-5 days · No obligation
AI Opportunity Audit
We map your operations, find the highest-ROI automations, and hand you a ranked plan with payback math. Yours to keep, whoever builds it.
No obligation. No follow-up sequence.
AI knowledge brain
The answers your firm already wrote, one question away
Hybrid retrieval over your documents, SOPs, precedents and transcripts. Every answer cited, every gap admitted. 5x ROI in 30 days. Or we work for free.
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
The problems this solves
01
The answer exists and nobody can find it.
The firm has answered this exact question before. It sits in a memo, a matter file, an old email thread. Redoing the work is faster than finding it.
02
Generic AI tools guess.
You tried a chatbot on the documents. It answered fluently, cited nothing, and was wrong in a way that mattered. Trust never recovered.
03
Your senior people are the search engine.
New hires interrupt the partner because asking beats searching. The most expensive person in the building fields the same question every week.
04
Knowledge walks out the door.
When someone leaves, the reasoning behind their decisions leaves with them. The documents stay. The map to them does not.
05
Nobody knows what the corpus covers.
You suspect there are gaps. Until coverage is measured against the questions your team asks, a gap only shows up when a client hits it.
What we build
01 · Collection
Documents, SOPs, precedents and call transcripts gathered from the systems they live in today. Nothing gets retyped or moved by hand.
02 · Indexing
Hybrid retrieval built over the whole corpus, so a question finds the right passage whether it matches the wording or only the meaning.
03 · Answering
Every answer arrives with its source attached. You read the passage it came from, not a paraphrase you have to take on faith.
04 · Refusal and routing
When the corpus does not hold the answer, the brain says so and routes the question to a person. The gap gets logged, not papered over.
05 · Coverage measurement
The brain is tested against your team's questions. Gaps become an ingestion list instead of a surprise.
When the corpus is silent, the brain says so. A logged gap costs you a question. A confident guess costs you a client.
What's included
- Your corpus collected, cleaned and indexed in your own isolated instance
- Hybrid retrieval tuned on your real documents, not a demo set
- A citation on every answer, linked back to the source passage
- Grounded refusal wired in from day one
- A coverage report built from the questions your team really asks
- Ask from chat or the brain surface, wherever your team already works
- Training, monitoring and tuning after launch. Cancel anytime
Where this lands hardest
AI for accounting firms→
Prior treatments and firm policy, found before the client calls back.
AI for law firms→
Precedent and matter history surfaced with citations a lawyer can check.
AI for financial services→
Policy and product answers grounded in the documents compliance approved.
AI for healthcare clinics→
Protocols and admin policy answered from your own manuals, or routed to a person.
Weeks, not quarters.
How the full engagement works01 Discovery
WEEK 0
02 Prototype
WEEKS 1-3
03 Deploy & train
WEEKS 4-8
04 Run & improve
WEEK 9+
How we build it
Step 01
Collect before you index.
The first weeks go to gathering the corpus from the systems it lives in and agreeing what belongs. If people know a document exists and the brain has not read it, they stop asking.
Step 02
Prove retrieval before generation.
Retrieval is measured on its own against your team's questions before any answer is generated. If the right passage does not come back, no model downstream can save the answer.
Step 03
Ship the refusal with the answers.
Grounded refusal is wired in from the first day, not added after an embarrassment. The brain earns trust by admitting what it does not know, and every refusal becomes a line on the ingestion list.
What we will not automate
Guessing. The brain will not fill a gap in your corpus with what a model remembers about your industry. An unanswered question goes to a person and onto the ingestion list, never into a confident paragraph.
How it is built
See the full stackRetrieval
Your documents, SOPs, precedents and transcripts are indexed for meaning and for exact wording, fused into one ranking. The passage behind every answer is retrieved before a word is generated, and retrieval quality is measured on its own.
- pgvector
- Full-text BM25
- Reciprocal rank fusion
Agents and orchestration
Ingestion, indexing and gap routing run as scheduled jobs, every run logged. When a question cannot be answered, the gap is filed and a person is notified rather than the question disappearing.
- agent-worker
- Scheduled ingestion
- Gap routing
Evaluation
The brain is graded with your people judging the answers. Coverage is measured against your team's questions and reported with the gaps listed.
- Eval graders
- Coverage sets
- quality-worker
Models
Frontier models write only from the passages retrieval returned. Embedding models handle meaning search. When retrieval comes back empty, no model fills the silence.
- Claude via OpenRouter
- OpenAI embeddings
- Grounded generation
Data boundary
One corpus in your own isolated instance, scoped at the database layer. A person retrieves only what they are allowed to read, and nothing you index trains anyone's model.
- Supabase row-level security
- Per-user scoping
- No-training API terms
Named tools in this build: Claude · Supabase + pgvector · Eval graders · Slack
What that means in practice
Retrieval and RAG
Connecting AI to your actual documents so it answers from your knowledge, accurately and with citations, instead of making things up.
Where we stop. A dedicated vector database is justified by scale, not by default. Most of RAG quality is won or lost in chunking and indexing strategy, not in the model choice, and we have walked clients back from RAG to plain search when that was the honest answer.
Evaluation and observability
How we prove the AI actually works: measured, monitored and regression-tested like real software, not vibes.
Where we stop. There is no engagement where we skip this. The honest variable is depth: a document pipeline gets faithfulness and extraction suites, an outbound agent gets human review sampling, a classifier gets a held-out test set. We size the harness to the risk, never to zero.
Security, privacy and governance
Keeping your data yours, and your AI safe to put in front of customers.
Where we stop. We do not claim certifications we do not hold, and we will not ship a customer-facing agent without a human gate and an eval suite. If a vendor cannot offer no-training terms, it does not get into the stack.
Questions? Straight answers.
Also built here
Three ways in
Three ways in. No long discovery.
Paid · Fixed scope
First Build
One high-ROI system, built on your real data and deployed in your stack, with your team trained to run it. Fixed scope. Quoted after the audit. Covered by the 5x ROI guarantee.
Scope my first build
Paid up front. Cancel anytime after.
Free · 15 min
Intro Call
Fifteen minutes with James, not a sales rep. Bring your worst bottleneck, leave with a straight answer.
Book 15 minutes
No pitch deck.
Every first build is covered: 5x ROI in 30 days. Or we work for free. Read the full terms
Start free. Know your number in five days.
A 3 to 5 day audit of your operations, ending in a plan with the ROI math attached. No obligation.