Hire someone to do your AI, or outsource it?

Should a home services business hire someone to do AI, or outsource it?

Outsource it, in almost every case at this size. The work is real but it is not full-time, the salary is high relative to a trades business, and the person is hard to hire and unlikely to stay long. Hiring makes sense only in the narrow case below.

This is not the same question as whether to hire a receptionist. That decision is about covering a desk, and it has its own answer. This one is about employing a technical person to put AI to work across the business: the phone, the follow-up, the scheduling, the customer list. Different job, different salary, different math.

The numbers below are all federal statistics or peer-reviewed research, because the internet's usual answers to this question are vendor marketing on both sides.

What does hiring someone to implement AI actually cost?

More than the salary, by roughly 43%. The Bureau of Labor Statistics puts the median salary for a computer and information systems manager at $171,200, and separately finds benefits add about 43 cents on every salary dollar for private employers. That is a real cost near $245,000 a year.

Here is the arithmetic, with the source for each line:

	Amount	Where the number comes from
Median salary	$171,200	Bureau of Labor Statistics, computer and information systems manager, May 2024
Benefits and payroll costs on top	about +43%	BLS Employer Costs for Employee Compensation: benefits are 30.1% of total employer cost, wages 69.9%
Real annual cost	about $245,000	The two lines above, multiplied

One honest caveat on that 43%: it is the private-industry average across all workers, not a figure measured specifically for technical management roles. The exact load varies by employer and by benefits package. It is the right order of magnitude and it is the number BLS actually publishes, but treat it as an estimate rather than a quote.

For most home and pet service businesses, that figure ends the conversation on its own. It is a number that competes with a truck, a crew, or a year of marketing.

Can a small business even hire someone for this?

Often not, and this is the part owners underestimate. In a July 2026 survey of small business owners, 31% had openings for skilled workers they could not fill, and 51% said applicants were not qualified for the roles they were hiring for. The budget is rarely the binding constraint.

This is the constraint owners tend to discover second, after they have already decided they could find the money. The National Federation of Independent Business surveys small business owners monthly. In July 2026, with 887 responses: 36% had job openings they could not fill at all, 31% had openings specifically for skilled workers, and 27% named labour quality or availability as their single most important problem, against a historical average of 12%.

A technical hire in a trades business also faces a quieter problem. The role has no peers, no manager who can evaluate the work, and no career path. That is a hard job to fill well even when the market is soft.

How long would that hire stay?

Plan for a few years, not a career. Median tenure in management occupations in the United States is 5.7 years, and 3.9 years across all workers. Whatever they build has to survive their departure, which means it has to be documented and owned by the business rather than by them.

The practical consequence is not sentimental, it is architectural. If one person builds everything in their own head, their notice period becomes a crisis. Anything built has to be documented, owned by the business, and handed over without them. That requirement does not disappear when you outsource; it just becomes somebody else's contractual obligation instead of your hope.

What if the hire does not work out?

You pay to do it again. Research synthesizing thirty case studies puts the median cost of replacing an employee at about 21% of their annual salary. On this salary that is roughly $36,000, before counting the year of AI work that did not happen while the role sat empty.

That 21% figure comes from the Center for American Progress, synthesizing thirty case studies across eleven academic papers. It is the most defensible number available on this, and it is worth saying it is a meta-analysis of older studies rather than a fresh survey.

It is also worth saying what we deliberately did not use. The two figures usually quoted on this subject, a dramatic six-figure cost for a single bad hire and a rule of thumb pricing replacement at most of a year's salary, have no reachable primary source. The first traces back to an agency executive's remark rather than to a study. The second appears across dozens of vendor blogs citing one another, with no original behind any of them. Both would have made this section more alarming and less true, and neither number is repeated here, because a figure quoted even to reject it is a figure that gets quoted onward.

When is hiring actually the right answer?

When the AI work is what makes your business different, rather than what keeps it running. The established test is core versus context: build in-house what customers choose you for, and buy in what simply has to work. For most home and pet service businesses, call handling is context.

The test is older than AI and it holds up well. Work divides into core, the thing customers actually choose you for, and context, the work that simply has to function. Core work belongs in-house, because it is where being better than the competition pays. Context work is bought in, because being merely excellent at it changes nothing.

For a plumbing company, plumbing is core. For a dog daycare, the care is core. Answering the phone reliably is context: nobody chooses you because your intake process is unusually elegant, but they will leave if it fails.

So the honest case for hiring is narrow and real. If AI is going to be the thing your business is known for, if you are building something your competitors cannot buy, then own it, hire for it, and accept the cost and the risk. If AI is how the phone gets answered and the follow-up gets done, that is context, and context is bought.

What goes wrong when you outsource it?

Governance, mostly, and it is worth naming because it is the real risk on this side. Work gets handed over without anyone defining what success means, who owns the result, or what happens at the end. The failure is almost never technical. It is that nobody wrote down the arrangement.

This deserves more than a sentence, because it is the risk on our side of the argument.

Deloitte's global outsourcing survey found that while 83% of executives were using AI within outsourced services, only 20% had built any strategy for governing that work, and that concrete benefits were limited by exactly that gap: unclear contracting, unclear ownership, unclear definitions of done. The survey covers large global enterprises rather than small businesses, so read it as a warning about the failure mode rather than a statistic about firms your size. The failure mode travels.

Translated to a business with four trucks, the questions to settle before anyone starts:

1. What does success look like, in a number you already track? Calls answered, jobs booked, hours returned. If nobody can name it, nobody can tell later whether it worked.
2. Who owns what gets built? The accounts, the configuration, the data, the phone number. The answer should be you, in writing.
3. What happens if you stop? A good arrangement can be ended without the business breaking. If ending it would break you, that is not outsourcing, that is a dependency.
4. Who do you call when it breaks at 6pm on a Friday? A named answer, not a support form.

An arrangement that answers those four is very hard to get badly wrong. One that answers none of them fails for reasons that have nothing to do with AI.

The short version

Most home and pet service businesses cannot justify the hire, could not fill the role if they tried, and would not need the person full-time if they did. The work is genuinely context rather than core, which is exactly the work that gets bought in.

The reason this decision feels harder than it is: the alternative to hiring usually looks like doing nothing, because nobody in the building has time to evaluate the options. That is the actual choice most businesses are making, and it is the expensive one. Census Bureau data shows firms with fewer than 20 employees use AI at roughly half the rate of firms with 250 or more, and the gap is not conviction.

If you want to talk through what to automate first in your business, that conversation is free and takes fifteen minutes. If you would rather read first, the implementation sequence is here, and an honest account of where AI call answering still fails is here.