Key Takeaway: For almost every small business, a full-time in-house AI hire is the wrong first move — and the reason is the job, not just the salary. The salary alone is steep: the U.S. Bureau of Labor Statistics puts the median data-scientist wage at $112,590 (May 2024), and an employee's true cost runs 1.25–1.4× base once you add payroll taxes, benefits, and overhead — call it $140,000–$158,000 a year all-in at the median. A first consulting project runs $2,000–$10,000. But the deeper problem is that most small-business AI work is front-loaded: you build a couple of automations, then the work drops to light maintenance. You can't keep a full-time hire busy, and you can't hire a fraction of a person. The honest sequence for a 5-to-50-person business is a project to build, then a light retainer or in-house upkeep to maintain — and an in-house hire only once AI becomes continuous, core work.
This is the build-vs-buy question owners ask once they're past "should we even use AI." Here's the straight version.
At some point the thought occurs to a lot of owners: instead of paying a consultant per project, why not just hire someone who does AI? Put them on payroll, keep the knowledge in-house, stop renting outside help.
It's a reasonable instinct. For most small businesses it's also a mistake — not because in-house talent is bad, but because a full-time role is the wrong-sized container for the amount of AI work a small business actually has. Let's put real numbers on both sides and then talk about when the instinct is right.
Should a Small Business Hire an In-House AI Person or a Consultant?
Start with the question underneath the question: do you have a full-time job's worth of AI work, indefinitely?
Most small businesses don't, and it's not close. The AI work in a 10-person law firm or a 30-person service company is a burst, not a stream: map the workflows, automate the two or three that eat the most hours, test them, hand them off. That's weeks of concentrated work, then a long tail of small tweaks. A full-time hire is built for a continuous stream of new work — which is exactly what you don't have after the first projects ship.
So for most owners the real comparison isn't "consultant vs. employee." It's "a project that ends vs. a salary that doesn't." That reframing usually settles it on its own, but the dollar figures make it concrete.
What Does a Full-Time In-House AI Hire Actually Cost?
More than the job posting suggests, in two layers.
Layer one — base salary. The U.S. Bureau of Labor Statistics reports a median annual wage of $112,590 for data scientists (May 2024), with the bottom 10% under $63,650 and the top 10% over $194,410. A software developer — the more common profile for someone building business automations — runs higher, at a $133,080 median. Postings for "AI engineer" specifically tend to advertise even more; national salary guides routinely put specialized AI roles well into the $150,000–$200,000 range. We'll use the conservative BLS data-scientist median to make the point, because even the low anchor is decisive.
Layer two — everything on top of base. A salary is not the cost of an employee. The widely cited rule of thumb (originating with MIT's Joseph Hadzima) is that an employee's true cost runs 1.25 to 1.4 times base salary once you add the employer's share of payroll taxes (7.65% FICA), health and retirement benefits, workers' comp, equipment, and software. Applied to the BLS median:
- At the median ($112,590): roughly $140,000–$158,000 in fully loaded year-one cost.
- At the low end ($63,650, a true junior): still around $80,000–$89,000 loaded.
And that's before the costs that don't show up in any salary table: recruiter fees (often 20–25% of first-year base for a hard-to-fill technical role), the weeks or months of ramp before the hire ships anything, and the risk that the one specialist you hired leaves and takes the institutional knowledge with them.
What Does an AI Consultant Cost by Comparison?
A fraction of one year of that, for the part most small businesses actually need built.
A first engagement for a small business typically runs $2,000–$10,000 for a single-workflow automation — audit, build, parallel testing, training, and a window of support. Our full pricing breakdown has the model-by-model detail: hourly ($100–$250/hour for independents), project-based, or a light maintenance retainer ($500–$2,000/month). For ongoing upkeep after the build, a retainer at the low end of that range covers far more than its cost in saved time.
Put the two side by side honestly: one year of a single in-house hire, fully loaded, is roughly 14 to 80 first projects. Almost no small business has 14 distinct automation projects queued up, let alone 80. You'd be buying a full-time capacity to do work that doesn't exist yet — and paying for the empty calendar in between.
The comparison isn't really about which is "cheaper" in the abstract, though. It's about matching the spend to the shape of the work. A project is the right size for front-loaded work. A salary is the right size for continuous work. The question is which one you have.
When Does an In-House AI Hire Actually Make Sense? The Full-Time Threshold
Sometimes hiring is the right call — and when it is, the per-hour economics flip in the employee's favor and you build durable institutional knowledge instead of renting it. Here's the honest test for whether you're there. An in-house AI hire makes sense when you can answer yes to all three:
- Volume. Is there genuinely enough AI and automation work to fill a full-time role for the foreseeable future — not a one-time build, but a continuous backlog? A role you can't keep busy is a role you can't justify.
- Continuity. Is the work ongoing build-and-iterate, rather than a setup that drops to occasional maintenance once it ships? Most small-business automation is front-loaded; a few businesses (high transaction volume, constantly changing processes) genuinely aren't.
- Differentiation. Is AI becoming part of what you sell or a core operational moat — not just back-office plumbing? Automating invoicing is a project. Building an AI feature into your product is a hire.
Three yeses means you have an actual full-time job to offer, and hiring is sound. One or two means you have a project followed by maintenance — which is a consultant, then a light retainer or a capable existing employee, not a new salary. The mistake isn't hiring; it's hiring before the work is continuous enough to keep the hire busy.
What About a Part-Time Hire, Contractor, or "AI Champion" on Staff?
These middle options are often the real answer, and they're worth naming:
- A capable existing employee as your "AI champion." Frequently the best move after the first projects ship. Someone already on your team — operations-minded, comfortable with software — owns the automations a consultant built, handles small tweaks, and flags the next opportunity. No new salary; the consultant trains them during handoff. This is what starting with ChatGPT and a consultant tends to grow into.
- A fractional or contract specialist. If you want outside expertise on a recurring basis without a full-time commitment, a retainer or contractor gives you senior skill for a few hours a month — the maintenance tier, not a hire.
- A part-time hire. Rarely the right fit for genuinely specialized AI work — the talent is hard to attract part-time, and you're back to the "not enough continuous work" problem at a smaller scale.
The throughline: match the commitment to the work. Front-loaded build work wants a project; ongoing light upkeep wants a retainer or an existing employee; only continuous, differentiating work wants a full-time hire.
How Should You Actually Sequence This?
For almost every small business, the same order:
- Build with a project. Automate your single biggest time-sink first — the one workflow that eats the most hours — and prove the ROI before spending more. Our free ROI calculator runs that math on your own numbers.
- Maintain cheaply. A light retainer or a trained in-house "champion" keeps it running. This is where the bulk of the post-build year actually lives — and it's a fraction of a salary.
- Revisit the threshold annually. If the work has genuinely grown into continuous, differentiating AI development that clears all three Full-Time Threshold tests, then hire — with real evidence that the role will stay busy, not a hunch.
If you're weighing build-vs-buy for your business right now and want a straight answer on which parts are a project, which are a retainer, and whether you're anywhere near needing a hire, book a free assessment — and if you'd rather start by understanding what a good local partner should cost and how to vet one, see our guide to finding an AI consultant in Omaha and the cost breakdown for setting up AI at a small law firm.