AI Employee vs Hiring a Part-Timer: How to Choose

The published part-time numbers, and who publishes them

The most quoted figure for part-time support is 20 hours a week at $520 to $1,200 a month, with 40 hours a week running $1,040 to $2,400. Those bands are agency-published and Philippines-based, so they describe one specific slice of a very wide global market.

Treat them the way you would treat any list price. The sources are staffing agencies whose business is placing assistants, which makes those pages the top of a sales funnel rather than neutral market research. Useful for setting a budget band, not for planning a decision around a specific number.

You will notice this article does not put our price next to that figure. That is deliberate and it is company policy. A monthly rate for a person's hours and a metered software subscription behave completely differently as your volume changes, so a table with both in it flatters the software and tells you nothing durable.

The genuinely useful research on this decision is not about rates at all. It is about what happens in the months after someone starts, and it points at something most founders underestimate.

What the onboarding research actually says

It takes 6 to 7 months for a new hire to feel settled in a role, and 86% of new hires decide whether they will stay long term within the first 6 months. Those two figures overlap almost exactly, which means the retention decision gets made during the period when the person is still finding their footing.

The early-departure numbers make the point sharper. 22% of workers left within the first 90 days, and 60% of those cited a lack of training or disorganised training. That is not a statement about the people who left. It is a statement about how well the companies prepared for them.

The cost of getting it wrong is documented too: most HR managers put the cost of a failed new hire at up to $25,000, while CHROs put it closer to $50,000. Again, the biggest lever in those numbers is the quality of the onboarding, which is entirely on the employer's side of the table.

So if you are leaning toward a part-time hire, the research says the same thing it has always said: the hire works when you have prepared for it. Written processes, a real first-week plan, and someone available to answer questions are what turn month one into month seven.

The half of this comparison that does not exist

Here is where most articles on this topic quietly cheat. Having quoted the human ramp figures, they produce a matching number for software setup time, usually something suspiciously round like fifteen minutes or one afternoon. No credible published dataset exists behind any of those numbers.

We are not going to make one up. The comparison is one-sided, and saying so is more useful than filling the gap with a figure you would be right to distrust. What can be said honestly is that the setup work is different in kind rather than merely faster.

With software, the effort goes into writing things down: what the job is, what the definition of done is, what must never happen. With a person, much of that transfer happens through conversation, questions, and watching. Whether one of those takes longer than the other in your business is something you would have to measure yourself, and nobody has published the answer. It is the honest shape of setup at Sistava as well: you brief an AI Employee in plain English on the job, the definition of done, and the lines it must never cross, and the writing is most of the work.

That honesty cuts both ways, and it should. If a vendor cannot cite a source for a setup-time claim, treat it the way you would treat an unsourced revenue promise. What you can reasonably expect is a faster feedback loop rather than a proven shorter total: a job that was badly specified tends to show that within days when it is written down, whereas a badly specified role handed to a person often takes weeks of goodwill before anyone says it out loud. That is a real difference, and it is still not a number.

Which one fits which job

The decision is much easier when you stop asking which option is better and start asking what the job is made of. Run your candidate work down this list and the answer usually falls out on its own.

Look at that table honestly and most roles are mixed. That is why the framing of one versus the other tends to mislead: what usually happens in practice is that the written, repeatable, high-volume rows move to software and the judgment rows become the reason a person's time is worth more. If you take that route, the pattern that holds up is a single AI Employee from Sistava scoped to the repeatable rows only, with the judgment rows staying with the person, and the boundary written down where both sides can see it.

What a person does that automation does not

The published assistant task list is heavier on judgment than its job titles suggest. Inbox management is really email prioritisation. Calendar scheduling is really travel coordination when the flight is cancelled. Alongside those sit customer follow-ups, meeting preparation, operations coordination, and writing an SOP for how this particular company works.

There is a measured reason software struggles with exactly that half. The OSWorld 2.0 benchmark of authentic professional tasks names three failure patterns, and the third is that agents struggle to maintain hidden state across steps, repeating work or executing from a plan that went stale several steps earlier. Holding the thread of a week-long piece of work and noticing when the premise changed is the thing a good assistant does without being asked.

That is a technical statement about how software behaves over long horizons, not a comment about anyone's colleagues. Two more categories close the boundary: anything physical, and anything needing a legally accountable person to sign it. Neither closes with a better setup.

A decision process you can finish this week

Five steps to a decision you can defend

1. Write the role out as tasks, not as a title — Twelve to twenty lines, each one a thing that actually happens. Titles hide the mix. Task lists expose it, and this list is useful whichever way you decide.
2. Mark each line written or judgment — Written means you could hand it to a new starter with five bullets. Judgment means it needs someone weighing what matters. Be strict, because the temptation is to call everything judgment.
3. Count the two piles — A role that is mostly written lines with high volume is a strong software case. A role that is mostly judgment lines is a hiring case, and no amount of setup changes that.
4. Prepare the onboarding either way — The research is blunt that missing or disorganised training drives early departures. Writing the process down helps a new hire settle and is the same artifact software needs, so it is never wasted work.
5. Set a review date and a number — Four weeks, one metric written down before you start. Hours returned, backlog cleared, response time. Without a baseline you will be arguing from impressions later.

Step four is the one people skip and the one that pays regardless of outcome. Every business that writes down how its work actually happens gets better at delegating it, and that is true whether the next pair of hands is a person, software, or both. If writing it up is the part you keep postponing, the free SOP generator among the free tools at Sistava will get a first draft out of you in a few minutes, and it costs nothing whichever way you decide.

One caution on the mixed case. If you split a role, be explicit about which half went where, because a half-delegated job with no clear boundary is the version that frustrates everyone involved and produces the worst results from both sides.

If your task list comes back mostly judgment, the right answer is the hire, and the most valuable thing this article can do is tell you so. A written task list plus a real first-week plan is exactly what the onboarding research says turns month one into month seven, and you will have both by the time you finish step four.

The framing of one against the other is popular because it makes for a clean table, and it is usually the wrong question. Roles are mixtures, and the mixture is what decides, which is why fifteen minutes with a task list beats an hour with any comparison chart including this one.

Write the tasks down, mark them honestly, and prepare the onboarding regardless of which way you go. That document is the thing that makes a new hire settle faster and the thing software needs in order to be useful at all, which makes it the only work in this entire decision that cannot be wasted.