"What does it really cost?" — the most sensible question, and the worst answered in most tech pitches.
You hear ranges from $5,000 to $500,000 for "automation" or "AI" with no context. Leaders either delay accessible gains or sign projects that blow the budget. Here's how I frame expectations with Quebec SMBs — without magic promises.
At a glance
- Cost depends on scope, not the buzzword "AI" or "automation"
- A useful pilot often costs less than a fraction of the manual work it replaces
- Budget in three blocks: discovery, pilot, scale — not all at once
- ROI on a good case is measured in weeks or months, not abstract years
- The most expensive hidden cost: doing nothing (lost hours, errors, turnover)
Three budget blocks (recommended order)
| Phase | Typical duration | Indicative range (SMB) | What you get |
|---|---|---|---|
| Discovery | 1–2 days | $3,000 – $8,000 | Mapping, prioritization, business case for one first case |
| Pilot | 4–8 weeks | $8,000 – $25,000 | Working flow, before/after metrics, trained team |
| Scale | Variable | Scope-dependent | Rollout to other teams or processes |
Indicative ranges for consultant support + modest tooling — not a global enterprise license.
Skipping discovery to "save money" often costs more: wrong tool, wrong process, frustrated team (when automation fails).
What actually moves the price
Lower end:
- Process already repetitive and somewhat documented
- Accessible data (not ten disconnected systems)
- Motivated pilot team
- Limited scope — one meeting type, one report, one flow
Higher end:
- Complex integrations between legacy systems
- Sensitive data requiring legal review and stronger security (data and AI)
- Broad organizational change without executive sponsorship
- Expectation of automating "the whole department" at once
A simple calculation any leader can run
Before discussing price with a vendor, estimate cost of the status quo:
Hours per week × number of people × 52 weeks × loaded hourly cost
Example: 3 people × 2 h/week of re-keying and formatting × $50/h loaded ≈ $15,600/year — for one irritant.
A $15,000 pilot that recovers half that time can pay back in under a year, often faster when you include avoided errors.
Not a universal formula — an order of magnitude to decide if the conversation is worth having.
What quotes should clarify (insist on this)
- Exact pilot scope (inclusions and exclusions)
- Who does what (your team vs consultant vs integrator)
- Recurring costs (licenses, hosting, maintenance)
- Measurable success criteria before start
- Next step if the pilot succeeds — without automatic lock-in
Be wary of "AI packages" with no target process or "we'll figure it out as we go."
ROI: three metrics are often enough
- Time — hours before/after on the target process
- Quality — errors, omissions, rework
- Turnaround — delivery speed (e.g. notes J+2 → same day)
Intangibles (morale, retention) matter but don't replace numbers when convincing a board or a cautious partner.
Budget mistakes I see often
- Comparing the pilot to an employee's annual salary ("are we hiring a robot?") — wrong frame; compare to mechanical work removed
- Forgetting recurring licenses — small lines that add up
- IT budget only, with no team time for validation and adoption
- Expecting 500% ROI in 30 days — possible on a micro-case, not on a vague transformation
Related on this site
- Spot automation gains without a massive project
- Too small for automation?
- Start automation without disrupting everything
"What does it cost?" deserves an answer tied to your friction point, not a generic range. If you want an honest estimate for a first case — discovery, pilot, order of magnitude — let's talk before you sign anything.