AI automation and consulting: frequently asked questions

Dassjay Labs Corp is a Canadian AI consultancy and product studio. We design and build business process automation, agentic AI systems, RAG and document AI, private LLM deployments and AI-powered web and mobile products, and we run our own products, The Filing Fool and Akurion.ai.

We are based in Kitchener, Ontario, in Waterloo Region, and work with clients across Canada, including Toronto, the GTA, Ottawa, Vancouver, Calgary and Montreal, as well as the US and internationally. Most work is delivered remotely, with on-site workshops in Ontario when useful.

With a short call about the process or product you have in mind. We usually recommend starting with our AI Readiness Assessment: two weeks, a fixed $2,500 CAD plus applicable taxes, ending in a scored, costed plan. The fee is fully credited toward a build that starts within 90 days, and you see working software on your own data before committing to a larger rollout.

We sign NDAs, use the minimum data needed, and can work entirely inside your cloud tenancy or on-premises with open-weight models, so sensitive data never goes to a third-party AI service.

Both. Smaller teams often start with one automation that pays back quickly. Larger organizations usually start with a strategy and prioritization phase across departments.

Yes. Most of our work connects AI to systems you already use: Microsoft 365, Google Workspace, Salesforce, HubSpot, ERPs, accounting systems, ticketing tools, databases and custom APIs.

Yes. Our Managed AI Care Plan is a monthly plan that covers monitoring, evaluation re-runs as models change, fixes and ongoing improvements. Or we hand over fully to your team with documentation and training.

RPA mimics clicks and keystrokes on fixed screens, so it works for stable, rule-based steps. Business process automation covers the end-to-end process. With AI in the mix, it can also handle unstructured inputs like emails and PDFs, and cases that need judgment, with human review where needed.

Start with high-volume, repetitive work that has clear inputs and a measurable outcome, such as invoice processing, request triage or report preparation. We score candidates on volume, error cost, data availability and risk, so the first project pays for itself and builds confidence.

Yes. Many engagements start with a single process and a short prototype sprint. The approach scales from a 10-person firm to an enterprise department.

Yes. We can deploy into your cloud account (including Canadian regions) or on your own hardware with open-weight models, so sensitive data stays under your control.

Yes. We're based in Kitchener and automate processes for businesses across Waterloo Region, Guelph and Southwestern Ontario, with on-site discovery when it helps and remote delivery for the build.

Agentic AI refers to systems where a language model does more than answer a prompt. It plans steps, calls tools and APIs, checks results and keeps going until a goal is met, within limits you set.

Only with the right controls. We give each agent the minimum permissions it needs, require human approval for irreversible or high-value actions, log every tool call, and test against a scenario suite before and after every change.

Whichever fits the job and your data rules: frontier models from Anthropic, OpenAI and Google, or open-weight models such as Qwen and Llama running privately. Agents are built so the model can be swapped without a rewrite.

The Model Context Protocol is an open standard for connecting AI applications to tools and data. If you want several AI tools or agents to use the same internal systems, an MCP server is usually the cleanest way to expose them once, securely.

Yes. Dassjay Labs builds AI agents and agentic workflows from Kitchener, Ontario, for companies across Waterloo Region and the rest of Canada.

Retrieval-augmented generation (RAG) is a pattern where the system first retrieves relevant passages from your documents, then has a language model write an answer grounded in those passages, ideally with citations back to the source.

For answering questions from documents that change, RAG is almost always the better starting point. It is cheaper, it stays current and it can cite sources. Fine-tuning is useful for teaching a model a format, style or narrow skill, and the two can be combined.

Graph RAG adds a knowledge graph of entities and relationships extracted from your documents. It helps with questions that require connecting facts across several files, which plain similarity search tends to miss.

Yes. We connect to SharePoint, OneDrive, Google Drive, Confluence, file shares, databases and email archives, and keep the index in sync as documents change.

Yes. We build cited document search over your files and can host it in a Canadian cloud region or on your own hardware so sensitive documents never leave your control.

A focused prototype on real data usually takes a few weeks. A production release depends on integrations, security review and scope. We agree acceptance criteria up front so progress is measurable.

No. Chat is one interface. Many of the most useful AI features are invisible: classification, extraction, routing, drafting and summarization inside a workflow your team already uses.

Yes. We can lead the build, pair with your team, or set up the architecture and evaluation framework and hand it over.

Yes. We help start-ups add AI features, build MVPs and set up evaluation and deployment, either leading the build or working with the in-house team.

For many document, extraction, classification and summarization tasks, yes, especially when paired with good retrieval. For the hardest reasoning tasks, frontier hosted models still lead. We benchmark on your own tasks so the decision is based on evidence.

Yes, with the right design: a clear purpose, consent where required, data minimization, safeguards, retention limits and transparency. Running models privately removes the question of a third-party processor entirely. This is general information, not legal advice.

It depends on model size and volume. Smaller quantized models run on a single workstation or Apple Silicon Mac. Larger models and many users need one or more data-centre GPUs. We size it to your workload.

Yes. Smaller teams can run capable open-weight models on a single workstation or Apple Silicon Mac in the office, with no data leaving the building.

It depends on scope. Strategy engagements are fixed-scope and fixed-fee, and build work is scoped in phases with clear acceptance criteria. Get in touch with a short description of your goals and we'll suggest a starting point.

Not if it's sized right. For smaller teams this can be a short working session that identifies two or three automations worth doing, followed by a prototype.

Some companies use programs such as NRC IRAP or SR&ED tax credits for eligible innovation work. Eligibility depends on your situation, so check with the program or a qualified adviser.

Yes. We run AI readiness and use-case workshops on-site in Kitchener, Waterloo, Cambridge and Guelph, and remotely anywhere in Canada.

AI can reduce false positives by making better judgments on aliases, transliterations and context, and it can surface indirect exposure through ownership chains. Final decisions should stay with trained staff, and every match should be explainable.

Yes. Pipelines can combine Canadian sources, such as the Consolidated Canadian Autonomous Sanctions List and provincial business registries, with US, UK, EU and UN sources.

Both. Walkthroughs use real footage from your product, usually on a demo account with sample data. Animated explainers are fully illustrated and need no screen footage.

We use licensed preset AI voices by default. With documented consent, we can also produce narration in your own voice.

Mostly TypeScript, React, Next.js and React Native on the front end, Python and Node.js on the back end, Postgres, Firebase and BigQuery for data, and GCP, AWS, Azure or Vercel for hosting. We adapt to your existing stack when it makes sense.

Yes. We start with a code and infrastructure review, fix what's urgent, then plan new features, including AI features, on top.

A written report and a findings session. The report covers the processes reviewed, a scored shortlist of automation opportunities, feasibility notes from a test on your sample data, privacy and risk recommendations, a prioritized roadmap and a scoped, priced proposal for the first build. It is yours to keep and act on, with us or with another team.

If you go ahead with a build with Dassjay Labs that starts within 90 days of receiving the assessment report, the full $2,500 is deducted from the cost of that build. The fee is in Canadian dollars, plus applicable taxes.

A kickoff conversation, walk-throughs with the people who run each process under review, access to a small set of sample documents or records, and a findings session at the end. We schedule around your team's work so day-to-day operations are not disrupted.

Then the report says so and explains why, for example too little volume, data that is not yet accessible or a process that needs to be standardized first. You also get the steps that would make it a good fit later. A clear no is a useful result and costs far less than a failed project.

Because the parts around them keep changing. Model providers update and retire models, your documents and processes evolve, and new users ask questions nobody tested. Each can lower quality without causing an obvious failure. Regular monitoring and evaluation catch these changes early, while they are still small fixes.

Monitoring of quality, usage and cost, evaluation re-runs before any model or prompt change, model version updates, bug fixes, small improvements agreed with you, security and access reviews, and a monthly report with a short review call.

As a monthly plan sized to the system: how many workflows it runs, how much it is used and how much change you expect. We quote it at launch, or as part of the AI readiness assessment if you are planning a new build.

Yes. Every build ends with documentation, runbooks and training. Some clients take over fully, some keep the care plan, and some start with the plan and move to their own team once they are comfortable.