For a Canadian small business, AI consulting can mean anything from a two-hour workflow review to a production system touching customer data. Those are not comparable purchases. The practical question is: what evidence, implementation and ongoing accountability are included?
Planning ranges, not a rate card
The following are AndChill planning estimates in Canadian dollars, not a market survey or quotation:
- Focused advisory session: $250-$750. One workflow, preparation, a decision memo and next steps.
- Workflow/AI opportunity assessment: $1,500-$5,000. Interviews, process map, baseline, ranked opportunities and pilot recommendation.
- Bounded pilot: $5,000-$20,000. One workflow, integrations, testing, human review, training and measurement.
- Multi-workflow implementation: $20,000-$75,000+. Multiple systems, data cleanup, access controls, change management and support.
- Ongoing optimization: $1,000-$6,000+ per month. Monitoring, prompt or workflow changes, vendor changes, incident handling and reporting.
Complex integrations, regulated data, custom software and round-the-clock support can move well beyond these ranges. Sales tax may be additional.
Why hourly comparisons mislead
Canada's Job Bank reports national wages for information-systems business analysts at $30.67 low, $45.13 median and $62.50 high per hour, based on its cited reference period. A consultant's billable rate must also cover non-billable discovery, sales, insurance, software, continuing education and business overhead. More importantly, hourly cost says nothing about whether the consultant can frame the right problem or safely implement it.
Ask for a statement of work showing deliverables, assumptions, exclusions, acceptance criteria, data handling, ownership, ongoing fees and the responsible person on each side.
A break-even example
An illustrative GTA service company is considering an $11,000 pilot to triage inbound requests and draft internal job summaries. Expected ongoing cost is $350 per month. The workflow currently consumes 42 hours per month at a loaded cost of $46 per hour. The team expects to recover 60% of that time.
| Monthly capacity value | $1,159 |
| Less ongoing cost | ($350) |
| Net monthly benefit | $809 |
| Simple first-year net | ($1,292) |
| Simple payback | 13.6 months |
The project does not clear a twelve-month payback on labour capacity alone. It may still be worthwhile if it measurably reduces response time or errors, but those benefits need evidence. The honest answer may be to narrow the pilot, negotiate scope or choose another workflow.
Choose the right commercial model
Fixed-fee discovery
Best when the goal and deliverables are clear but the solution is not. It protects the client's ability to stop.
Fixed-fee pilot with change control
Best for a bounded workflow with written inputs, outputs and acceptance tests. Uncontrolled “small extras” are the main threat.
Time and materials
Useful when legacy-system uncertainty is high. Require weekly spend visibility, decision logs and a cap before re-approval.
Retainer
Appropriate after something important is operating, not as a substitute for defined work. Specify response times, included hours and monitoring.
Performance-linked fees
Potentially attractive but easy to misunderstand. Define the baseline, attribution, time period, data source and what happens when volumes change.
Questions to ask before signing
- What business metric will change?
- What is explicitly out of scope?
- Who owns the workflow, documentation and created assets?
- Where will customer and employee data go?
- What human approvals remain?
- How will accuracy and exceptions be tested?
- What recurring vendor and support costs remain?
- Can we export our data and operate without this consultant?
- What is the stop condition?
Start with real operating examples
Review named, sourced AI cases before deciding what kind of engagement you need.
Ask what you will own after the project
Two proposals with the same price can have very different economics. Clarify ownership of prompts, workflow logic, documentation, accounts, data connections and training materials. Ask who monitors failures, what a change request costs and whether another provider can maintain the system. A cheap build that only its creator can understand is a subscription wearing a project costume.
Compare intervention sizes
| Manual process cost per year | $31,200 |
| Process repair costing $2,500; expected reduction | 25% |
| Ordinary automation costing $8,000; expected reduction | 55% |
| AI workflow costing $18,000; expected reduction | 70% |
The AI option saves the most gross dollars, but the ordinary automation may have the faster payback and lower operating risk. The consultant's job is to recommend the smallest intervention that clears the business case-even when it is not AI.
Three common engagement shapes
| Workflow diagnosis: baseline, process map and opportunity model | $2,500-$6,000 |
| Bounded pilot: working workflow, controls, documentation and training | $8,000-$25,000 |
| Multi-system implementation: integrations, migration, governance and support | $25,000+ |
These are AndChill planning ranges, not market quotes. The useful comparison is the decision each engagement buys. Discovery should end with a defensible yes, no or not-yet. A pilot should prove one measurable workflow. A larger implementation should leave the business with maintainable infrastructure rather than a permanent mystery.
Sources and methodology
Price one useful decision first.
Bring one workflow and the outcome you want. We will help you determine whether it deserves a pilot before discussing a larger build.
Book a free call →Ranges are illustrative planning estimates as of August 31, 2026, not quotes or a statistical survey. Actual fees, taxes and scope vary.