One of the most common questions we hear from Vancouver business owners is some version of: "I've talked to a few AI vendors and gotten wildly different quotes. How do I know what's reasonable?" The answer requires understanding the different pricing models in play — because comparing a project-based quote to a retainer to a SaaS subscription is like comparing apples to helicopters.
The Four Main AI Pricing Models
1. Project-Based (Fixed Scope)
You pay a flat fee for a defined deliverable: a chatbot, an automation workflow, a data pipeline. The quote covers discovery, build, testing, and initial deployment. Ongoing support is either not included or priced separately.
Project pricing works well when the scope is well-defined and the AI use case is relatively contained. It gives you budget certainty and a clear end date.
The risk: scope creep. AI projects often expand once you see what's possible. A fixed-scope contract can create friction between you and your vendor if requirements change mid-stream. Good vendors handle this with a structured change order process; less experienced ones either absorb the scope creep (and cut corners) or bill surprise change orders (and damage the relationship).
Typical range for a mid-market AI project in Vancouver: CAD $15,000–$80,000 depending on complexity.
2. Time and Materials (T&M)
You pay for hours worked at a stated daily or hourly rate. The total cost depends on actual effort.
T&M suits exploratory or research-heavy projects where the right solution is not clear upfront. It protects the vendor from scope risk and gives the client flexibility to pivot. It protects the client from a vendor who underscopes to win a fixed-price bid and then delivers something incomplete.
The risk: open-ended cost. Without careful milestone reviews and budget controls, T&M projects can run significantly over initial estimates. Experienced buyers of professional services know to negotiate a "not to exceed" cap or to require weekly time reporting.
Typical rates in Vancouver for senior AI engineers: CAD $180–$300/hour. For a dedicated team, expect $25,000–$50,000/month in burn rate.
3. Monthly Retainer
You pay a fixed monthly fee for a defined scope of ongoing services — typically a combination of model monitoring, optimization, support, and new feature development.
Retainers make sense once a system is deployed and you want the vendor to stay engaged. The best retainer relationships are proactive: your vendor is monitoring performance, flagging degradation, and suggesting improvements before you notice problems.
Typical retainer range: CAD $3,000–$15,000/month depending on system complexity and SLA commitments.
4. Outcome-Based or Revenue Share
The vendor takes a portion of the measurable value generated — a percentage of revenue increase, cost savings, or efficiency gains — instead of (or in addition to) a fixed fee. This aligns incentives perfectly in theory.
In practice, outcome-based pricing is rare and requires rigorous baseline measurement, agreed-upon attribution methodology, and mutual trust. It works best when the outcome is unambiguous and measurable: e-commerce conversion lift, number of leads converted, compliance processing hours saved.
We have seen outcome-based arrangements in the range of 10–20% of documented savings, typically capped at a multiple of what a fixed-price engagement would have cost.
What Drives Cost Up
Custom vs. configured: A bespoke ML model trained on your proprietary data costs more than configuring an existing LLM with your business context. Both are valid — the question is whether bespoke performance justifies the cost difference for your use case.
Integration complexity: Connecting an AI system to a modern REST API costs a fraction of integrating with a legacy mainframe or proprietary data warehouse. Expect a significant premium if your core systems are old or undocumented.
Compliance requirements: Healthcare, financial services, and legal AI carry compliance overhead — privacy impact assessments, security reviews, audit logging, access controls. This is legitimate cost, not padding.
Data preparation: If your data is messy (and it usually is), data remediation adds to the project cost. Vendors who don't include this in their scoping are setting you up for a painful mid-project conversation.
Red Flags in AI Proposals
Vague deliverables: "AI chatbot implementation" tells you nothing. A proper proposal specifies the number of intents trained, integration points, testing methodology, and performance benchmarks.
No measurement plan: If a vendor can't tell you how you'll know whether the AI is working, they don't have a serious approach to delivery.
Zero discovery phase: Any vendor who quotes a complex AI project without a discovery phase is either underscoping to win the bid or doesn't know what they don't know yet. Discovery — typically 1–2 weeks — is how you define the actual scope and surface the risks.
Suspiciously low quotes: The economics of AI consulting are not a mystery. If a quote seems too good to be true, ask what's not included. Common omissions: training data preparation, post-launch support, integration testing, documentation.
What You Should Expect for Your Budget
Under $10,000: Can typically buy a configured chatbot using existing tools (Intercom, Drift, custom GPT-4 integration), light automation using Zapier/Make, or a focused discovery and strategy engagement. Do not expect custom model development at this price point.
$10,000–$30,000: Entry-level custom AI project. A well-scoped chatbot with CRM integration, a document processing workflow, or a lead scoring system. Tight scope, limited iterations.
$30,000–$80,000: Mid-complexity custom AI project. Multi-system integrations, custom model fine-tuning, robust monitoring and dashboarding. This is the most common range for meaningful business AI projects in Vancouver.
$80,000+: Enterprise AI — complex data pipelines, multiple models, compliance-grade architecture, large-scale deployment. This is where ROI analysis becomes critical before signing.
Understanding these ranges lets you evaluate proposals intelligently and have a more productive conversation with any AI vendor about where their quote comes from and what you are getting for the investment.