Frequently asked questions about our AI software

It depends on the project type. For document extraction, we need between 200 and 500 labelled examples per document type. "Labelled" means someone has marked which fields contain which values on each sample. If you have the raw documents but no labels, we can do the labelling for you at an additional cost.

Chat assistants are different. They rely on your existing knowledge base: product manuals, FAQ pages, policy documents, support ticket transcripts. We index and embed that text. The more complete your docs, the fewer gaps the bot will have.

Predictive models need historical records. Twelve months of data is a workable minimum for most forecasting tasks, though 24 months gives us more seasonal signal.

Yes. About a third of our clients are in Ontario and British Columbia, and we have completed projects for companies in the United States and France. All project management happens through video calls, a shared Notion board, and a staging environment you can access from anywhere.

That is fine. Book a discovery call and we will evaluate whether AI actually solves your problem better than a simpler approach. Sometimes a well-designed database query or a rule-based script does the job faster and cheaper. We will tell you if that is the case. We have talked clients out of AI projects when the data was too sparse or the task too straightforward to justify the investment.

Starter projects usually go live in two to three weeks. Professional-tier projects take three to five weeks. Enterprise engagements with multiple integrations and custom deployment requirements can run six to ten weeks. These timelines assume you provide the training data within the first week. Delays in data delivery push the schedule back proportionally.

The project fee covers development, deployment, and 90 days of monitoring. After that period, you can either manage the model yourself (we provide documentation) or subscribe to our extended support plan at $800 per month. Cloud hosting costs are separate and depend on usage volume; a typical Starter deployment runs about $60–$120 per month on AWS.

Yes. The initial 50% payment covers work already completed up to the cancellation date. You receive all code and models produced so far. We do not charge the remaining 50%. We have only had two cancellations in four years, both because the client company was acquired mid-project.

We offer a 15% discount on project fees for registered Canadian non-profit organizations. Send us your CRA registration number during the discovery phase and we will apply the discount to your quote.

We track model accuracy, latency, and error rates through an automated dashboard. If accuracy drops below the threshold we agreed on during scoping, we investigate and retrain within five business days. Bug fixes and configuration adjustments are included at no extra charge during this window. The Starter tier includes one retrain; Professional includes two.