How to Choose an AI Consultant Who Actually Delivers Results

AI consulting is the practice of assessing a business's operations, data, and workflows to identify where AI automation will deliver measurable returns, then implementing and optimizing those systems. The quality gap in this market is extreme: NTT DATA found 70-85% of generative AI deployments fail to meet ROI expectations, largely because neither the consultant nor the client defined success in measurable terms before starting.

Gartner forecasts worldwide AI spending will total $2.52 trillion in 2026, up 44% year over year (Gartner, January 2026), a figure Gartner revised upward in May 2026 to $2.59 trillion and 47% growth (Gartner, May 2026). That growth rate tells you two things: demand is real, and a lot of new entrants are flooding the space with varying degrees of competence.

The uncomfortable reality is that most AI consulting engagements do not produce the results they promise. NTT DATA's 2024 research found that between 70 and 85 percent of generative AI deployment efforts are failing to meet their desired ROI (NTT DATA). Gartner's own tracking shows the trajectory worsening, not improving: by the end of 2025, at least 50 percent of generative AI projects were abandoned after proof of concept (Gartner), up from their earlier prediction of 30 percent (Gartner, July 2024). This is not an argument against AI consulting. It is an argument for choosing carefully. The difference between the consultants who deliver and the ones who do not is identifiable before you sign anything — if you know what to look for.

what Toronto small businesses pay for AI consulting__

The Quality Problem in AI Consulting

The barrier to entry in AI consulting is effectively zero. Anyone who has used ChatGPT can call themselves an AI consultant. The result is a market where technical sophistication varies wildly, and most buyers lack the expertise to distinguish between a consultant who understands machine learning infrastructure and one who builds automations on top of consumer tools.

This matters because the failure mode is rarely dramatic. You do not lose everything at once. Instead, you spend three months on a project that technically works — it produces outputs, it runs on schedule — but the business impact is negligible. A good consultant identifies this before starting. A mediocre one starts building anyway.

Gartner found that 63 percent of organizations either do not have or are unsure if they have the right data management practices for AI, based on a Q3 2024 survey of 248 data management leaders (Gartner). Without proper data foundations, defining clear business value becomes nearly impossible, which is why so many engagements stall. The pattern is consistent: the consultant sells the technology, the client buys the promise, and neither side defines what success actually looks like in measurable terms before the engagement begins.

Five Questions That Separate Good Consultants From Bad Ones

1. "Walk me through a project that failed."

Every competent consultant has failed projects. The question is whether they can articulate why — and more importantly, what they learned. A consultant who claims a perfect track record is either lying or has not done enough work to encounter real complexity.

The answer reveals their diagnostic thinking. Do they blame the client's data? That is a red flag — data quality is something they should have assessed upfront. Do they blame scope creep? That suggests weak project management. The best answers identify a specific assumption that turned out to be wrong and explain how they changed their process to catch similar assumptions earlier.

2. "What will you measure and when will you measure it?"

This is the single most important question. The reason 70 to 85 percent of AI deployments fail to meet ROI expectations is not that the technology does not work. It is that nobody defined what "working" means in advance.

A good consultant arrives with a measurement framework before they arrive with a technical architecture. They should be able to tell you: what the baseline metric is today, what the target is in 30, 60, and 90 days, and what happens if the target is not met at each checkpoint. If the answer is vague — "you'll see improvements in efficiency" or "it depends on the data" — that is a consultant who plans to deliver activity, not outcomes.

Thryv's 2025 survey found that 66 percent of small businesses using AI save $500 to $2,000 per month (Thryv). A return you can prove rests on specific metrics that changed. Not feelings. Not anecdotes. Numbers with dates.

3. "Who on your team will do the actual implementation work?"

AI consulting firms frequently sell with their senior talent and deliver with their junior staff. This is not inherently wrong — but you should know about it before you commit. Ask specifically: who will be in the room during discovery? Who will build the system? Who will be your point of contact when something breaks at 4 PM on a Friday?

The answer tells you whether you are buying access to expertise or buying a brand name attached to a relatively inexperienced team. For small businesses, this distinction is critical. A senior practitioner who has implemented AI for businesses your size will identify pitfalls in week one that a junior team will discover in month three — after you have already spent the budget.

4. "How does your work continue after the engagement ends?"

The most expensive AI consulting outcome is not a failed project. It is a successful project that you cannot maintain. If the consultant leaves and the system breaks, you have a dependency, not a capability.

Ask about knowledge transfer. How will your team learn to operate, monitor, and troubleshoot the system? What documentation will you receive? What happens if the underlying platform changes its API or pricing? A consultant who builds something only they can maintain has created a recurring revenue stream for themselves, not a business asset for you.

This is the pattern that Gartner's research points to when it flags over 40 percent of agentic AI projects facing cancellation by end of 2027 due to escalating costs and unclear business value (Gartner). Many of those escalating costs come from ongoing consultant dependence that was not planned for.

5. "Can I talk to a client in my industry who you worked with more than six months ago?"

Recent references are easy to find. The real test is what happened after the consultant left. A client who is still benefiting from the implementation six months later is a fundamentally different reference than one who is still in the honeymoon phase.

Industry relevance matters here. McKinsey's State of AI research found that organizations where AI generates the highest revenue impact are more than three times as likely to use specialized, domain-specific AI implementations rather than generic solutions (McKinsey). That premium exists because specialized consultants have already encountered the edge cases in your industry. They know which data sources are unreliable, which integrations break under load, and which processes look like automation candidates but actually require human judgment. A generalist discovers those things on your dime.

Red Flags That Should End the Conversation

There are patterns that consistently predict consulting failures. If you encounter any of these, end the evaluation:

The consultant leads with technology rather than business problems. A conversation that starts with "we'll build you a custom model" or "we use the latest GPT-4 architecture" before understanding your operations is a technology-first approach. The research is clear on the outcome: most AI projects fail because they solve the wrong problem, not because they use the wrong technology. We documented the root causes in our analysis of why most AI projects fail before they start.

The proposal does not include a discovery or assessment phase. Any consultant who quotes a fixed price and timeline before understanding your data, processes, and team capabilities is either working from a template or guessing. Both are unacceptable when you are spending real money.

Consultant B asks what process costs you the most time and money. They propose a two-week paid assessment ($2,500) before quoting any implementation work. They define a specific metric (hours saved on lead follow-up) and a measurement timeline (30/60/90 days). Their reference is a firm your size in your industry.

The consultant cannot explain the pricing model clearly. AI consulting for small businesses ranges from $2,000 for a basic readiness assessment to $150,000 for full custom implementation, with the shape of the engagement mattering more than the headline figure. Job Bank Canada reports AI consultant wages between $30 and $69.74 per hour (Job Bank Canada) — a wide enough band that the same 100-hour engagement can price very differently depending on who is doing the work. If the pricing is opaque, value-based without defined deliverables, or requires a long-term commitment before proving value — that is a structure designed to protect the consultant, not the client. We break down the real cost landscape in our AI enablement pricing guide.

The consultant dismisses your need for measurement. Any resistance to defining success metrics upfront is a signal that the consultant is not confident in their ability to deliver measurable results. Our framework for measuring AI ROI for small business outlines what good measurement looks like.

What a Good Engagement Actually Looks Like

A well-structured AI consulting engagement for a small business follows a predictable sequence:

Phase 1 is assessment. The consultant audits your current processes, data assets, team capabilities, and technology stack. This typically takes one to two weeks and produces a prioritized list of AI opportunities ranked by feasibility and business impact. You should receive a written deliverable — not just a slide deck.

Phase 2 is a focused pilot. One process, one clear success metric, one timeline. The pilot proves the approach works in your specific environment before you commit to a larger engagement. If your business is ready for this step, our readiness checklist walks through the specific prerequisites.

Phase 3 is implementation with measurement. The system is built, deployed, and monitored against the metrics defined in Phase 1. Weekly or biweekly check-ins ensure the project stays aligned with business objectives, not just technical milestones.

Phase 4 is knowledge transfer and handoff. Your team learns to operate the system. Documentation is complete. The consultant becomes available for support rather than required for operation.

If a consultant proposes skipping Phase 1 or combining Phase 1 with Phase 3, you are being sold an implementation before the problem is understood. That is the single most common cause of the failures the research documents.

The Decision Framework

Here is the practical filter. Before engaging any AI consultant, confirm:

They can name a specific metric they will improve and by how much. Not "efficiency" or "productivity" — a number attached to a business outcome.

They have worked with businesses your size. Enterprise consultants often underestimate the constraints of small businesses — limited data, smaller teams, tighter budgets. If their case studies are all Fortune 500, your engagement will be their learning experience.

They charge for assessment separately from implementation. This structure aligns incentives. An honest assessment might conclude that you are not ready for AI — and a consultant who profits from that honesty is one you can trust.

They can explain their approach in plain language. AI is technically complex but the business logic behind it is not. If you cannot understand what they are proposing and why, the problem is their communication, not your technical knowledge.

talk to a DeployLabs consultant__.