AI development questions, answered plainly

We are an applied AI and software engineering studio in Vancouver, BC. We build AI agents, AI assistants on company knowledge, document processing and AI features in products, plus the software around them: SaaS platforms, websites, iOS and Android apps, integrations and cloud infrastructure. We have been building software since 2018, for more than 75 clients.

Mostly mid-size and enterprise Canadian companies, roughly 200 to 5,000 people, in sectors such as financial services, insurance, healthcare, manufacturing, retail, logistics and the public sector. The common thread is high-volume work with documents, systems and approvals.

Book a free 30-minute call with a senior engineer. We talk through the process you want to improve and tell you honestly whether it is a good first project. If it is, the next step is a two-week discovery phase with a fixed scope and price.

A short conversation with an AI on our website. You describe a business problem and it drafts a possible solution, the systems involved and a realistic timeline. It is a starting point for discussion, not a quote or professional advice.

Yes, that is what discovery is for. We interview the people who do the work, measure how the process runs today, and rank three to five candidate use cases by value, data readiness and risk. You leave with a clear first project, or a clear reason to wait.

Yes. We are based in Vancouver, BC and work with clients across Canada. Most of the work happens remotely, with on-site sessions for discovery workshops when they help.

It depends on the process, the systems involved and how much the AI is allowed to do. We do not publish rates. After the first call you get a fixed price for discovery and an estimate for the whole path. Each later phase gets a firm, fixed price at the decision point before it.

Because it puts the risk of estimating in the right place. You know what each phase costs before it starts, and if we underestimate the effort, that is our cost. Pricing one phase at a time also means each price is based on evidence, not guesswork.

Cloud hosting and AI model usage. Those are billed directly to your own cloud and model accounts, so you see the real costs and keep control. We estimate them per transaction during the prototype, before you commit to the build.

Yes. Every phase ends with a go or no-go decision that is yours to make. If you stop, you keep everything delivered so far: reports, code and documentation.

Yes, we regularly sign mutual NDAs before discovery. For the engagement itself, we can work from your master services agreement or provide ours, and we are used to procurement and legal review.

In your own cloud account (AWS, Azure or Google Cloud), in a Canadian region by default, or on your own infrastructure. If a model you need is not available in a Canadian region, we tell you before any data moves and document the decision.

No. We use private model endpoints under provider terms that exclude training on your data, or open-weight models hosted in your environment. Your data does not improve anyone else’s model.

We design for them from discovery. Every project includes a privacy impact assessment covering data flows, purposes, where data is stored, providers, risks and retention. We also design notices and human review for any decision made solely by automated processing. Your privacy officer and counsel make the final calls.

Yes. We answer your questionnaire line by line with evidence about the actual system: architecture and data flow diagrams, threat model, access matrix, test results and our list of subprocessors.

Only the engineers assigned to your project, through accounts you issue, with the narrowest permissions that work. You can see and revoke that access at any time, and every access is logged.

The one that performs best on your cases at an acceptable cost. We evaluate models from Anthropic, OpenAI, Google, Cohere and Mistral, as well as open-weight models you can host yourself. We have no partnership that pushes us toward one vendor.

We plan for it. Every system has measured accuracy, confidence checks that send uncertain cases to a person, and approval steps for anything consequential. Errors that reach review become new test cases, so the same mistake is caught automatically next time.

We build an evaluation set from your real cases with answers checked by your experts, and measure the system against the success metrics agreed in discovery. The same tests run again on every change and every model update.

Yes, within limits you set. An agent can update records, route documents or draft transactions. You choose per action whether it suggests, acts with approval or acts alone below a threshold, and every step is logged and reversible.

We test it against your evaluation set. If it is better or cheaper on your cases, we can switch, because the system is designed with a thin model layer rather than built around one provider.

No. AI is our main focus, but we have built SaaS platforms, websites, iOS and Android apps, e-commerce, integrations and custom business software since 2018. Most AI systems need that software around them to be useful.

Yes. We build mobile apps for customers and for field teams, including offline use, sign-in through your identity provider and connections to your back-office systems.

Yes. Most projects connect to systems you already run, such as an ERP, CRM, document store or an in-house application. We can also take over an existing codebase after a short technical review.

Yes. We design and build fast, accessible, bilingual websites with search engine optimization designed in, and connect them to your CRM and marketing tools.

Senior engineers. The people you meet in discovery are the people who build the system and run the weekly demos.

Every week. You get a demo of working software on real data, a written update with risks and decisions needed, and a shared Slack or Microsoft Teams channel with the engineers.

Typically an executive sponsor for about 30 minutes a week plus decision meetings, a product owner for 2 to 4 hours a week, a few hours from subject-matter experts, and IT and security contacts at key points.

Our documentation, deliverables and user interfaces can be produced in French, and we test AI systems in both official languages. Tell us about your language needs in the first call so we plan the project around them.

Mostly remotely, which keeps costs down. We come on site for discovery workshops or launch training when being in the room makes a real difference.

You do. The code, infrastructure definitions, evaluation suite and documentation live in your repositories and accounts from the first commit.

Yes. Under a monthly agreement we monitor the system, fix issues within agreed response times, re-run evaluations on every model update, and send a short results report each month.

Yes. We hand over in stages, with pairing sessions, runbook walkthroughs and a period where we stay on call as backup until your team is confident.

Each month you get a short report on the numbers agreed in discovery, such as volume handled, accuracy, time saved and running costs, compared with the baseline, plus what we plan to improve next.