Applied AI North

Six phases. Every AI build, every time.

How an AI Consulting build runs, by week. The calendar is published before the contract — no “TBD” phases, no ten-week discovery, no surprise change orders.

Governance is built into each phase, not handed over at the end: impact screening up front, a person approving the AI’s work during the build, and an off switch plus security testing before go-live. Workshops and websites run on shorter rhythms — see below.

A typical AI Consulting build, shown here as six weeks. An AI assistant takes 4–8 weeks and a workflow automation 3–5; shorter builds run the same six phases on a tighter clock.

  1. Assess: week 1
  2. Shape: week 2
  3. Build: weeks 2 to 4, mostly week 3
  4. Harden: week 4
  5. Pilot: week 5
  6. Adopt: week 6
Kickoff MondayCalendar set in writingHandoff & 30-day support

  1. Assess

    Phase 1 · Week 1

    On site with the team. Watch the work. Decide what is worth building.

    We do not start in a workshop. We start at a desk, on a couch, on a job site, watching the actual person do the actual task. Some ideas don’t survive this stage, and that’s the point. The rest come out with a sharper scope than the client walked in with.

    Assess includes an Algorithmic Impact Assessment (AIA) screening on the government’s framework (the Directive on Automated Decision-Making, impact Levels I–IV). Before any code, we classify what each use case touches and where human oversight is non-negotiable — risk scoped at the on-ramp, not found in production. If you start with an AI Roadmap, its two weeks cover this phase: each opportunity scored, with rough costs and payback, the main risks, and a note on any funding program worth checking. The roadmap doesn’t replace a program’s own paperwork; BDC’s LIFT financing, for example, requires BDC’s own Digital Plan.

    What you give us

    • A kickoff call with leadership
    • A half-day with one team, for shadowing
    • Sample documents or call recordings (anonymized fine)

    What you get back

    • A scored opportunity list, with a risk rating for each
    • A written plan (Do now, Do next, Not yet, Don’t), with rough costs and payback, noting any funding programs worth checking
    • An honest go / no-go recommendation

    Maps to: Algorithmic Impact Assessment (AIA) · the Voluntary Code of Conduct (Accountability)

  2. Shape

    Phase 2 · Week 2

    Write the tests before the system. Define what right looks like.

    We pull 100 to 500 real historical examples and label them. This is the unglamorous week that decides whether the project succeeds. Without a test set, you do not have a project, you have a vibe.

    Shape is where we set how far the system may act on its own. Following the federal Guide on the Use of Agentic AI, we keep it assistive: read-only by default. The AI reads, extracts, and drafts, but is blocked from writing to systems or sending external messages without a person’s sign-off. Every step that touches a person or a record gets a human approval step — a design decision, not a later patch.

    If you can’t describe success on a spreadsheet, you can’t build for it. Week 2 is the spreadsheet week.

    Maps to: Guide on the Use of Agentic AI (Bounded Autonomy, Levels 1–2) · the Voluntary Code of Conduct (Human Oversight)

  3. Build

    Phase 3 · Week 3

    First working pass. Crude but end-to-end.

    By the end of week 3 you have a working system on a test site. It is not polished. It is end-to-end, and we can run it against the tests and see exactly where it fails. That failure list is the build plan for week 4.

    The limits from Shape are now wired in. The human approval step becomes a real checkpoint — a Slack approval, a dashboard sign-off — between the AI’s output and anything leaving the system. Protection against hidden instructions ships here too: outside text (documents, emails, web pages) is treated as data, not instructions — the model reads it but cannot obey it. The AI proposes; your team executes.

    What ships in this phase

    • The system itself, wired to your tools — read-only by default
    • The human approval step on every write or external action
    • A test-results dashboard, updated nightly
    • A short Loom walking through it

    Maps to: Guide on the Use of Agentic AI (prompt-injection defense) · the Voluntary Code of Conduct (Human Oversight, Safety)

  4. Harden

    Phase 4 · Week 4

    Tool calls, retries, cost limits, monitoring.

    Most agency projects quietly stop at the happy path. Week 4 is the part nobody photographs: rate limits, fallback prompts, handling partial failures, daily cost ceilings, and alerts when the system starts to drift. It’s what makes the system survive a real Tuesday afternoon.

    It’s also where recovery gets built and tested: audit logs the AI can’t alter, written so a person can read them, plus an off switch that cuts the AI’s access at once and puts the work back in your team’s hands. Before go-live we try to break it on purpose — hidden instructions, data leaks, unsafe outputs — and keep the evidence. Same architecture as the four layers of defence: recovery and limits built into the structure, not left as settings.

    Maps to: Guide on the Use of Agentic AI (Recoverability, automation drift) · the Voluntary Code of Conduct (Validity & Robustness, Safety)

  5. Pilot

    Phase 5 · Week 5

    Live with 2 or 3 real people. Measure friction.

    We pick 2 or 3 actual end users (not managers) and put the system in front of them for a week. We watch what they do, where they go around it, where they distrust it. The thing we ship in week 6 is shaped by what week 5 tells us, not by what looked good on the demo.

  6. Adopt

    Phase 6 · Week 6 + 30 days

    Handoff, training, 30 days of running support.

    We write a runbook your team can actually read. We train a named owner. We sit in your Slack for 30 days, on the house, watching the metrics and patching the small things that surface in real use. Then we get out of the way. The point of the engagement is that you do not need us afterwards.

The other two run shorter.

Same rules — calendar first, your real work, yours to keep — on a tighter clock. See AI Literacy & Enablement and Web Development for formats and prices.

Team Workshop

Half or full day
  1. Prep. A 20-minute call with your team lead; your team sends a few real tasks (anonymized is fine).
  2. Session. 3 or 6 hours for up to 15 people, hands-on, on your team’s actual work — not toy prompts.
  3. Take-home. A short field guide and 3 workflow templates.
  4. 30-day follow-up. One live check-in about a month later to fix what didn’t stick. After that, office hours or the Enablement Program if you want more.

Starter Website

3–4 weeks, up to 5 pages
  1. Week 1. Strategy call, brand audit, sitemap, copy outline.
  2. Week 2. First full pass — design, copy and working pages. Review.
  3. Week 3. Second pass, polish, mobile, edge cases, copy edits.
  4. Weeks 3–4. Deployment, analytics, handoff, content training.

How we operate, regardless of the project.

The rules that have not changed since the practice opened — now read alongside Canada’s federal AI frameworks, because good process and good governance turned out to be the same thing.

  • Tests before code.

    If we can’t measure right, we don’t build right. Every AI build opens with a labelled set of real examples.

  • Ship to one user first.

    No rollout without a single named pilot user who can break it for us.

  • Default to your stack.

    We use what your team already runs. We’ll add one new tool at most, and only if the payoff is obvious.

  • Canadian data residency where it matters.

    We sign data processing agreements and use enterprise endpoints. Where the engagement allows, systems are architected for Canadian data residency — deployable on Canadian cloud regions (e.g. AWS Canada Central) and Canadian foundation models such as Cohere — so proprietary data and IP can stay onshore, engineered to align with PIPEDA and Quebec Law 25. We show the model your data; we never train on it.

  • People keep the judgment. You own it.

    Systems augment the people who own the decision; they don’t replace judgment. We build bounded automations, assistants and websites, train your team, and hand over the keys. Every project ends with a handoff doc — so you could run it without us, or swap us for another vendor, on a Monday. Digital infrastructure, not vendor dependency.

  • Drift checks after handoff.

    Governance doesn’t end at handoff. Drift checks are part of the 30 days of support after every build: we watch for the slow divergence between what a system was scoped to do and what it has started doing, and tightening the limits before drift becomes an incident. After that, any further check is quoted as its own small fixed-price piece of work.

A calendar before a contract.

We write the calendar before we sign. If the dates don’t work for either of us, that conversation happens before any money moves.

Book a free intro call See pricing