This is the final part of a five part series on the practical AI and web stack for a small or medium business. The earlier parts covered the tools: where AI runs, where your site lives, how you take money, and how to put AI on your own documents. This part is about people, which is the most expensive decision of all and the one most often made backwards.
The reflex, when AI feels important, is to hire an "AI person." For most small businesses that is the wrong first move, and an expensive one. There are three real options, and they fit very different situations.
What the three paths actually cost
Hire a specialist. The market for serious AI talent is high. A machine learning engineer's median total compensation is around $272,000, AI engineer base pay runs roughly $145,000 to $310,000, and a security clearance adds another 10 to 20 percent on cleared work. Those numbers are real and they are climbing. They are also almost never what a small business actually needs. You hire at that level to build differentiated AI as your product, not to put a chatbot on your website.
Upskill someone you already have. The off-the-shelf AI tools a business actually uses cost about $18 to $30 per person per month: Microsoft 365 Copilot at roughly $18 to $21 for the business tier, ChatGPT Team around $20 to $25 a seat, Claude and Gemini in the same range. Give those tools to a capable employee who already understands your business, and pair them with a few free courses, and you have most of what a small business needs for a tiny fraction of a hire. The person who already knows your customers and your processes is usually a better bet than a specialist who knows neither.
Outsource a defined project. For a specific build with a clear finish line, a custom integration, a one-time data cleanup, a website with a booking system, a contractor or small agency is the right tool. You pay for the project, not a salary, and you are done when it is done. The risk to manage is the handoff: make sure you own the result and that it is documented, or you have traded a salary for a dependency.
How to choose
Run your situation through three questions.
- Is AI your product, or a tool you use? If you are building AI capabilities that customers pay you for, and you will keep building them, that is a hire. If AI just makes your existing business run better, it is not. A property-management company using AI to draft notices and answer tenant questions is using a tool. A startup whose product is an AI model is building one. Only the second hires a specialist.
- Is the need ongoing or one-time? Ongoing, everyday use points to upskilling someone internal, because the capability needs to live inside the business. A one-time build points to outsourcing, because you do not need the skill on payroll after it ships.
- Do you have someone curious and capable already? Almost every small business does. The bookkeeper, the office manager, the owner's kid home from college. Give that person the tools, an afternoon a week, and permission to experiment, and you will be surprised how far it goes, and at how much cheaper it is than any of the alternatives.
For the overwhelming majority of small businesses, the answer is the same: upskill someone internal and hand them off-the-shelf tools, then outsource the occasional defined project. The $200,000 hire is for the rare business where AI is the product, and for medium businesses that have grown a steady, differentiated, in-house AI need.
When a medium business should hire
There is a real line where hiring becomes right, and it is worth naming so you know when you have crossed it. Hire a specialist when AI work is recurring, when it is specific to your business in a way no off-the-shelf tool covers, and when the cost of getting it wrong is high enough to justify someone whose whole job is getting it right. A regional services firm running AI across many locations, with real data and real compliance exposure, has crossed that line. A single shop adding a chatbot has not. If you are genuinely at that point, the companion piece to this series, what $200k AI jobs actually ask for, is the map of the skills to hire for or to build toward yourself.
The path I would actually take
If I were advising a small business owner today, the order would be simple. Buy a few seats of a mainstream AI tool this week and give them to your most curious employee. Point that person at one painful, repetitive task and let them prove it saves time, the same start-small approach from part one of this series. Outsource the one or two builds that genuinely need a developer, and insist on owning and documenting the result. Revisit the idea of a hire only when the in-house need has clearly outgrown the tools, which for most businesses is later than they think, and for many is never.
That is the whole philosophy behind the Digital Empire book The $100 Network: build the capability inside the business you already have rather than buying it at a premium you do not need. The cheapest, most durable AI capability a small business can have is a person who already understands the business and now has good tools.
The honest summary
Most small businesses should not hire an AI engineer. They should give good tools to someone who already knows the business, outsource the occasional defined build, and keep the result documented and owned. Hire a specialist only when AI is your product or your in-house need has genuinely outgrown the tools. Get that order right and you spend hundreds a year where others spend hundreds of thousands, and you end up with capability that actually lives inside your business. That closes this series: cheap intelligence, portable infrastructure, money you control, knowledge you own, and people you have grown.
The series
- Previous: Part 4, RAG and agents over your own documents
- Part 5 (this post): Hire vs upskill vs outsource
- Start over: Part 1, local AI vs cloud APIs
Related reading
- What $200k AI jobs actually ask for, the skills to hire for or build toward
- The $50 a month AI stack for small business, the tools an upskilled employee would use
- AI employees and small-business stacks for 2026, where the off-the-shelf tools are heading
- Claude for small business, a walkthrough, a concrete first project
Fact-check notes and sources
Salaries and tool prices change; treat these as approximate mid-2026 figures and confirm before relying on them.
- AI and ML compensation: Levels.fyi machine learning engineer (median total comp ~$272,000); KORE1 AI engineer guide (~$145K to $310K base). Cleared premium (~10 to 20%): ClearanceJobs 2026 compensation report coverage.
- Off-the-shelf AI seat pricing: Microsoft 365 Copilot pricing (~$18 to $21 business tier); ChatGPT Team (~$20 to $25 a seat) via OpenAI; Google Gemini for Workspace; Claude pricing.
This post is informational and not hiring, financial, or career advice. Salaries and prices are current as of mid-2026 and change; verify before relying on them. No affiliation with the products or job boards mentioned is implied.
