AI consultants: what they actually do, and when to hire one

An AI consultant is someone who helps a business figure out where AI actually fits, scopes the highest-value problem to solve, and either builds the solution or hands it to a build team.

If you've spent any time looking for help with AI for your business, you've almost certainly seen the title "AI consultant" thrown around. What you've probably struggled to find is a plain-English explanation of what one of those people actually does day-to-day, what they cost, and whether your business — at the size and stage it is now — actually needs one.

That's what this post is for.

We'll cover what an AI consultant does in practice (the unglamorous version, not the LinkedIn version), how the engagement structure typically works, what it costs, and the three questions that tell you in two minutes whether your business is at a stage where hiring one would be worth it. We'll also cover when you genuinely don't need one — because honest answers to that question are rarer than they should be in this market.

The role gets used to mean a dozen different things — strategic advisor, prompt engineer, change-management facilitator, model fine-tuner, agent-framework integrator, content automation specialist. For a small or mid-sized business, most of those distinctions don't matter.

What matters is the underlying pattern. An AI consultant for an SMB does three things, in order:

1. Identifies which 2 to 3 tasks in the business are taking 10 hours that should take 2. Most owners can name one within a minute of being asked. The good consultants ask depth questions until the candidate list is sharp, then narrow it further.
2. Designs the connections between existing tools to close that gap. The work happens around the systems your team already uses — your CRM, your accounting tool, your email, your project management — not by replacing them. (We've covered the four building blocks of this work in our plain-English guide to how business automation works.)
3. Deploys and measures. The first measurable result typically lands inside 3 to 5 weeks. From there, the consultant either continues with the next loop or wraps up depending on what the business needs.

Concrete example: one of our recent engagements was with a B2B industrial company that participated in dozens of external events every year — tradeshows, sponsorships, donations, speaker engagements. None of it was tracked centrally. The fix wasn't a chatbot or an AI agent. The fix was automating the data flow between the company's existing tools so the events team gained visibility into the full year for the first time. Result: 6 to 8 hours per week recovered, $4,000 to $10,000 per month in operational savings.

That's what "AI consulting" looks like in practice for an SMB more often than not. Not flashy chatbots. Not generative AI for content. The unglamorous work of making the systems you already have visible and connected. For teams already exploring AI tools internally, vibe coding workflows are one common adoption path we'll often help an SMB pilot alongside the core automation.

These roles get conflated a lot. The differences matter when you're deciding who to hire:

• An AI consultant focuses on outcomes — hours saved, money recovered, time back. The deliverable is a working integration plus a measurement framework. The relationship is typically per-project with a defined endpoint, not an ongoing retainer.
• A software agency ships features. They build what you spec. Outcomes are your responsibility, not theirs. Useful when you know exactly what you want built; less useful when the harder question is which problem to solve.
• A chatbot or AI tool vendor sells a specific product. Their incentive is to fit your problem to their product, not the other way around. Useful when you've already identified that you need their specific tool; risky when you're still scoping.
• Your IT person keeps the existing systems running. That's a different job — important, but different. Asking them to drive an AI initiative on top of their existing workload usually fails for capacity reasons, not skill reasons.

Most engagements that go badly trace back to one of these mismatches: a business hires a software agency thinking they're hiring a consultant, or a chatbot vendor thinking they're getting strategy. Worth being deliberate about which role you actually need.

Here's where the market splits. Enterprise AI consulting — the McKinsey/Accenture/BCG tier — typically runs $500K to several million per engagement. That's not the market most SMBs need.

For SMB-focused AI consulting, the patterns we see (and bill at):

• Project minimum: around $3,500 USD for a focused single-loop automation.
• Typical engagement size: $3,500 to $25,000 depending on how many systems are involved and how complex the logic is.
• Pricing model: per-project, not hourly. The number is fixed before work begins.
• Free initial walkthrough: 30 minutes, by video. Used to qualify fit on both sides before any commitment.
• Median time from first conversation to signed contract: around 10 days.

The economics work because most SMB automation projects recover $4,000 to $10,000 per month in operational costs once they're running. Across our work, the engagement pays for itself within the first quarter, often the first month.

Be cautious of consultants who quote hourly without a ceiling. Hourly billing creates an incentive to drag out timelines. Per-project pricing creates an incentive to ship.

Most engagements follow the same four phases — for a deeper walkthrough, see how a typical engagement actually runs:

1. Discovery. Mapping the current process. This is where most of the value lives. Surface answers won't work; the consultant should ask depth questions, look at the actual systems, and pull real numbers. Expect 2 to 4 hours of focused conversation across one or two sessions.
2. Scope. Together you pick the highest-leverage friction point. One or two, not ten. The goal is concentrated, measurable wins — not a roadmap covering every possible automation.
3. Build. Bridges and listeners go live. Your team keeps using the same tools. The consultant handles the integration work; you handle reviewing outputs and giving feedback.
4. Measurement. Recording what changed. The first measurable result typically lands inside 3 to 5 weeks. From there, the engagement either continues with new loops or wraps up.

About 90% of focused SMB automation engagements reach production. The 10% that don't almost always trace back to one pattern, which we'll cover in the "when you don't need one" section below.

Three questions, asked of yourself, that tell you in two minutes whether you're at a stage where hiring one would pay off:

1. Which task in your business is taking 10 hours that should take 2? Most owners can name one within a minute. That's usually the highest-leverage place to start.
2. Where is information being re-entered between two tools? The friction points where one person re-types data the company already has — those are bridges waiting to be built.
3. What does someone on the team do every Monday that they always wish they didn't have to? Recurring weekly work is the easiest thing to automate, because the trigger is already implicit.

If you can answer all three with specifics, you're probably ready to act on this. If you can answer two, you're in the right ballpark — a free walkthrough is worth your time. If you can answer none of them, the next step isn't hiring an AI consultant — it's a different conversation about what's actually slowing the business down.

Three patterns where AI consulting is the wrong investment for an SMB:

You have a CIO, an established IT team, or procurement-driven purchasing. You'll get more leverage from an enterprise consultancy with the staffing to match your scale. The owner-to-owner pattern that fits SMB consulting won't translate to your decision-making structure.

Your processes are already unified and your systems already talk to each other. You're past the building-block stage. A focused automation isn't where the next dollar of investment goes — it goes to model-level work, custom AI features, or scaling existing systems.

You don't have time to invest in the engagement itself. AI consulting requires the client to invest in the engagement, not just buy it. The 10% of projects that don't reach production almost always trace back to the same pattern: the client couldn't carve out time to provide context and respond to questions during discovery and build. That's the most honest "when not to hire" framing — it points the finger at engagement quality, not your intelligence or business. If you can't carve out 4 to 6 hours across the first month for the discovery and review work, the project will produce a fraction of what it could.

A good consultant flags this in the sales conversation rather than charging for a thin result.

If you've decided you want to engage one, these are the questions worth asking — for both sides — during the walkthrough:

1. What's your typical engagement structure? "Per-project, fixed price" beats "hourly, open-ended" almost every time for an SMB. Hourly billing creates the wrong incentives.
2. How do you measure results? A consultant who can't articulate the measurement framework before the project starts probably can't deliver one after it ends.
3. What happens when a project stalls? Every honest consultant has had projects stall. Listen for whether they take responsibility for engagement quality on their side, or whether they only blame clients. Discovery is a two-way responsiveness test — when projects fail, both sides usually contributed.
4. What kind of businesses do you typically work with? "Anyone" is a yellow flag. Specialization usually beats breadth at this stage of the market.
5. What does discovery look like? A consultant who's willing to do 30 minutes of unpaid discovery before quoting is signaling that they care about fit. One who quotes immediately based on a project description without depth questions is selling a template.

• McKinsey & Company, "A future that works: Automation, employment, and productivity", 2017.
• Zapier, "Business Automation Statistics".
• Stanford HAI, "AI Index Report" — annual benchmark on AI adoption rates and how fast the consulting profession is evolving.

If you're considering AI consulting and want concrete examples of what specific projects look like, two posts go deeper than this one: how to automate manual data entry for a small business and three automation quick wins to start with. For the broader framework that informs every engagement, how business automation works is the longest piece. And if you're ready to talk specifics, our AI consulting services page lays out what an engagement looks like end-to-end.