An assistant that has read everything your company knows
Not an FAQ bot. An assistant that has read every contract, policy, and file your company owns — and cites the source on every answer, so you can check it rather than trust it.
Permission-aware
Slack, Teams, or web
From $7.5k
An AI assistant on company data — often called a RAG system — indexes an organisation's own documents so staff can ask questions in plain language and get answers drawn from those documents, with a citation to the source file on every response. DAvision builds these for Canadian businesses from $7,500 for 500 documents on a single knowledge base up to $28,000 for unlimited sources with department-level access control, retrieval tuning, and Canadian data residency.
The knowledge exists. Nobody can find it.
The answer is in a PDF somebody saved in 2023
Every company has institutional knowledge locked in files nobody can locate under pressure.
Generic AI does not know your business
ChatGPT is confidently wrong about your contract terms, your policies, and your pricing, because it has never seen them.
Staff ask the same senior person everything
One person becomes the search engine, and the organisation stops when they are on holiday.
How it is built
Citation on every answer
Each response links the source document and passage. If it cannot cite, it says it does not know.
Permission-aware retrieval
Each person sees answers only from documents they are already allowed to read. HR files do not surface to the warehouse.
Connects to where files already live
Google Drive, SharePoint, email, or a folder we index. Re-indexing happens automatically as files change.
Available where people work
Web interface, Slack, or Microsoft Teams — not another tab nobody opens.
Retrieval tuned and tested
We build an evaluation set from real questions and measure accuracy rather than assuming it.
Canadian data residency on request
For firms with PIPA, PIPEDA, or contractual obligations that keep data in country.
How it works
Four steps, no surprises. You see a working version before you pay the balance.
Source inventory
Which documents, where they live, who is allowed to see what.
Index and connect
Ingestion, chunking, permission mapping, and connection to Drive, SharePoint, or email.
Evaluate against real questions
We collect the questions your team actually asks and measure how well it answers them.
Roll out and tune
Launch to a pilot group, review what it got wrong, tune retrieval, then widen access.
Packages & pricing
Fixed-price builds in Canadian dollars. Ongoing support is quoted separately from $750 per month.
- Up to 500 documents indexed
- Source citation on every answer
- Web interface or Slack
- Single knowledge base
- 30-day support
- Up to 10,000 documents
- Permission-aware: each person sees only their own
- Automatic re-indexing as files change
- Connects to Drive, SharePoint, or email
- Slack, Teams, and web
- Usage analytics
- 6 months optimization
- Unlimited documents and data sources
- Department-level access control
- Retrieval tuning and accuracy testing
- Canadian data residency on request
- Audit log of every question asked
- Dedicated architect
- 12 months support
Citation is the feature. Everything else is plumbing.
An assistant that answers confidently without showing its source is worse than no assistant, because it is faster to be wrong. The design decision that makes these systems usable in a business is that every answer links the document and passage it came from, and that the system declines to answer when it cannot cite. Staff stop having to decide whether to trust it; they can check it in two seconds.
The second design decision is permission-aware retrieval. A naive implementation indexes everything and answers from everything, which means an HR investigation file can surface in an answer to someone in the warehouse. Proper implementation maps your existing access controls into the index and filters retrieval before generation, so a person can only receive answers from documents they were already allowed to open. We test this explicitly during rollout with deliberately probing questions.
Accuracy is the third, and it is the one most vendors leave unmeasured. We build an evaluation set from questions your team actually asks — collected during the source inventory, not invented — and measure how often the system retrieves the right passage. That number is reported before launch and tracked afterwards. Without it, 'it seems to work' is the only available assessment, and it is not one you can act on.
For firms with PIPA, PIPEDA, or contractual data-handling obligations, hosting location is not a detail. Canadian data residency is available, and we confirm the arrangement in writing before a single document is ingested, because retrofitting a residency requirement after indexing means starting over.
What changes after launch
Concrete, measurable, and checkable against the baseline we take before we start.
Answers with a source attached
Every response cites the file and passage, so verification takes seconds rather than a search.
Senior staff stop being the search engine
The institutional knowledge locked in one person's head becomes available to everyone entitled to it.
Measured accuracy, not assumed
An evaluation set built from your real questions gives you a number before launch and a trend after.
Who this is for
Professional services firms with years of contracts and precedent. Engineering and energy companies with technical documentation and standards. Clinics with policy and compliance material. Any organisation where the answer exists in writing but nobody can find it fast.
We work with dental clinics, construction, energy, real estate, and commercial businesses across Calgary and Canada.
Frequently asked questions
How is this different from ChatGPT?
ChatGPT answers from public training data and has never seen your documents. This indexes your own material and answers from it, with a citation to the source file so you can verify every claim.
How much does it cost?
Up to 500 documents on a single knowledge base starts at $7,500. Up to 10,000 documents with permission-aware access and automatic re-indexing is $15,000. Unlimited sources with department-level access control, accuracy testing, and audit logging starts at $28,000.
Does our data stay in Canada?
It can. Canadian data residency is available at Enterprise tier for organisations with PIPA, PIPEDA, or contractual requirements. We confirm the hosting arrangement in writing before any data is ingested.
Can it accidentally show someone a document they should not see?
Permission-aware retrieval at Pro and Enterprise maps your existing access controls into the index, so retrieval is filtered before an answer is generated. We test this explicitly during rollout.
What happens when documents change?
At Pro and Enterprise the index updates automatically as files change in Drive, SharePoint, or email. At Basic, re-indexing is scheduled.
How do we know the answers are accurate?
We build an evaluation set from questions your team actually asks and measure performance against it before launch. Enterprise includes ongoing retrieval tuning against that set.
Related services
Thirty minutes, no pitch
We map out exactly what you need and what it would cost. If automation is not the answer, we will say so.
