AI or a New Hire? How to Decide Before You Recruit

List the new role's tasks and the hours each takes. If AI and simple automation can absorb most of the routine hours and what's left fits into your current team's week, trial that for 60 to 90 days before recruiting. If the role is mostly customer contact, physical work or judgement, or 15 or more hours a week remain, hire.

Often the honest answer is both: a smaller hire, with AI doing the drafting. The worked example below, an independent bookshop weighing a 20-hour role, ends up there. You'll also get a task-sorting method, a 12-month cost comparison to fill in, and a way to run a trial with a clear trigger for recruiting, so the decision doesn't drift for six months.

Follow me on Instagram@sagnikteaches

Step 1: break the role into tasks and hours

Write down what the new person would actually do each week, with a rough number of hours per task. Not the job title: the tasks. If you can't list them, you're not yet ready to decide either way, because neither a hire nor an AI tool can do "help out".

Connect on LinkedInSagnik Bhattacharya

The first version of the list is usually a job title in disguise. An illustrative plumbing and heating firm's owner starts with "office help: phones, admin, keeping on top of things". Pushed for detail, over a week of noting what he actually does after 6pm, it becomes: booking boiler services from the reminder list (4 hours), chasing unpaid invoices (2), turning engineers' job notes into invoices (3), answering the phone while engineers are out (6), ordering parts (2). Only now can anyone see that six of the seventeen hours are the phone, which no chat assistant will answer on his behalf, and that invoicing from job notes is a drafting job AI does well.

Subscribe on YouTube@codingliquids

Say an independent bookshop is planning a 20-hour-a-week assistant. The owner's list, illustrative:

Step 2: sort every task into one of three groups

• AI drafts, a person checks. Writing tasks with a pattern: replies, descriptions, posts, summaries. The time drops but doesn't disappear, because someone must read and correct the output.
• Rules-based automation, no AI needed. "When X happens, send Y." The special-order "your book has arrived" email is this: a fixed message triggered by a status change. AI versus rule-based automation helps you tell which is which, and rules are cheaper and more reliable where they fit.
• Human only. Physical presence, relationships, judgement, anything where a mistake costs trust. The Saturday till is here, and so is the conversation with a regular who wants a recommendation.

Then estimate the hours after the change. Be conservative; first estimates of time saved are nearly always too high. How much time AI can really save each week gives realistic ranges by task type.

AI and automation take roughly eight hours out of twenty. Twelve still need a person, three of them standing at a till.

Why do stock listings only drop from four hours to two, rather than to half an hour? Because of the checking. Here is the kind of prompt the bookshop might use, and an illustrative first draft back:

PROMPT
You write short website descriptions for an independent
bookshop. Using only the publisher's blurb below, write
60-80 words in a warm, plain voice, British spelling.
Don't invent awards, reviews or page counts. Add "Ask us
about signed copies" only if the details say it's signed.
[publisher's blurb pasted here]

ILLUSTRATIVE OUTPUT
A luminous, award-winning debut, this unforgettable novel
follows a lighthouse keeper's daughter as she uncovers her
family's secrets on a storm-battered island. Perfect for
book clubs, it's a story about memory, loss and the pull
of home. Pick up your copy today, and ask us about signed
copies!

Three fixes before it goes live: "award-winning" is invented, despite the instruction; "luminous" and "unforgettable" are blurb clichés a bookseller wouldn't write; and the signed-copies line appears though the book isn't signed. That's about two minutes a listing to read and correct, and at 60 new titles a week it adds up to the two hours in the table.

A different role can produce a very different result. Suppose a small executive-coaching firm plans a 15-hour admin assistant: session follow-up summaries (5 hours), invoicing and payment chasers (3), scheduling back-and-forth (3), the newsletter and posts (2), and phones and post (2). Recorded sessions summarised by AI and checked by the coach take the first down to about 1.5 hours, as long as clients have agreed to recording. Chasers and scheduling are rules work for the accounts package and a booking link, half an hour each. Posts halve. Phones stay human. That leaves about 5.5 hours, under the 8-hour line, and the two coaches can absorb it. Here AI first is the right call.

Step 3: cost both routes over 12 months

Use a full year, because the costs land differently: a hire is steady, while the AI route is front-loaded with setup. The figures below are illustrative. Employer costs (payroll taxes, pension or benefits, holiday cover) vary a lot by country and contract, so ask your accountant for your percentage; the example uses 20%.

The hybrid saves about $7,200 in cash in year one and more in later years, once setup is paid for. It costs the owner more time up front, and it leaves a 12-hour role that must include checking AI drafts. Whether that's the better deal depends on whether the owner has those 60 hours in the next two months. If not, the cheaper option on paper will stall.

Notice what the pure AI option would look like: $1,740 and a lot of owner time, with twelve hours a week of work still to be done by somebody. Unless the existing team genuinely has twelve spare hours, "AI instead of a hire" here really means "AI plus overtime".

When AI wins, when a hire wins, when you need both

The 8 and 15 hour lines are rules of thumb, not laws. Adjust them for how stretched your current team really is.

One situation overrides the table: nobody on the team can judge the AI's output. Take an illustrative three-person design studio weighing AI against a part-time bookkeeper. An assistant can categorise bank transactions and draft month-end summaries quickly, and on a test month it looks tidy. It also books a client's 50% deposit for a project not yet delivered as income for that month, which flatters the month's figures and would mislead the owners about cash they've already half spent. Neither designer spotted it; the studio's accountant did, at year end. Where checking needs expertise you don't have, the "AI drafts, a person checks" group doesn't exist for you. Hire the expertise (here, a few hours a month of a bookkeeper) and let that person use AI to go faster.

For contrast, picture a bakery planning to hire a second baker for the 4am shift. Run the same sort and almost every hour lands in the human-only group: mixing, shaping, baking, cleaning down. AI might help the owner write the job advert, and that's all. Hire. The method matters most for mixed roles like the bookshop's, where writing and admin sit alongside work that needs a person in the room.

Run a 90-day trial with a trigger to recruit

If the sums point to AI first, don't let "trying AI" become an open-ended reason to put off a hire you need. Set the trial up with an end date and a trigger.

1. Weeks 1 to 2: set up. Build the drafting and automation for the top two or three tasks. Measure the current hours for a week before switching anything on.
2. Weeks 3 to 8: run and measure. Track hours spent on the role's tasks each week, how long customers wait for replies, and how many AI drafts needed heavy correction.
3. Weeks 9 to 12: decide. Compare against the trigger you wrote down on day one.

A trigger might read: "If, from week six, the role's tasks still take more than 12 hours a week, or customer emails regularly wait more than two working days, or anyone is working evenings to keep up, we advertise." Write it down before you start. Once you've paid for the setup, it's tempting to move the goalposts.

The bookshop used the trial differently: to test whether a 12-hour role would really be enough before advertising it, with the owner and the Saturday part-timer covering the tasks. Its sheet at week 8, illustrative:

The hours settle at around 12 once the prompts improve, so the 12-hour role holds. The one breach, in the event week, is useful too: the advert now says "12 hours, with extra hours in author-event weeks", which is cheaper than planning every week around the busiest one.

Trials also end the other way. An illustrative electrical contractor with two vans tried AI for quote follow-ups, invoicing from job sheets and booking confirmations instead of hiring a 16-hour administrator. The trigger was written down on day one: more than 10 hours a week on the role's tasks from week six, or quotes waiting more than two days. At week six the tasks still took 14 hours, because the job sheets were handwritten photos that needed retyping before AI could use them, and two quotes had waited four days during a busy week. So they advertised. The trial wasn't wasted: the new administrator started with tested prompts for follow-ups and invoices, and the job advert asked for 16 hours rather than the 25 first pencilled in.

Where the AI-first route usually goes wrong

When "AI instead of a hire" disappoints, it tends to be for one of these reasons. Watch for them during the trial.

• Counting the drafting time saved but not the checking time. If the email replies take one hour to draft instead of five but another hour and a half to check and correct, the real saving is two and a half hours, not four.
• Forgetting the busy season. A bookshop's December can bring several times the usual volume of orders and queries. AI copes with extra drafting; it doesn't staff a queue at the till. Run the sums for your busiest month, not an average one.
• The owner becomes the bottleneck. If every draft waits for the owner's approval, replies now depend on one person's diary. Give a second person the authority to approve routine drafts.
• Nobody notices when it stops. An automation that silently fails for a week creates a backlog a new hire would have spotted on day one. Check a sample every week and set an alert if the tool offers one.
• The team quietly absorbs the rest. Twelve leftover hours spread across three busy people doesn't look like much on paper and feels like a lot on a Friday.

What a person brings that AI doesn't

The cost table can't show these, so weigh them deliberately.

• Cover. When you're ill or on holiday, AI can't open the shop.
• Relationships. Regular customers come back for people. In a bookshop, a recommendation from someone who has read the book is part of what's being sold.
• Noticing. A person spots the shelf that's emptying, the customer who seems upset, the supplier invoice that looks wrong.
• Capacity to grow. A good hire brings ideas and takes on new work. AI takes on what you give it.
• The checking load. Every AI draft needs a reader. If that reader is always you, the saving is partly an illusion.

If you hire, hire for the AI-assisted version of the job

Whatever you decide, the role you advertise shouldn't look like the role would have looked five years ago. Write it around outcomes and include the AI tools as part of the job, so the person you hire expects to draft with AI and check its work. Writing a job advert for an AI-savvy admin assistant has wording you can adapt, and redesigning job roles once AI handles routine tasks helps if existing roles need to change too.

Tell your current team what you're weighing and why before the trial starts. People who hear that AI is replacing a planned hire tend to wonder whether they're next. A few sentences at the weekly meeting are enough, for example: "We were going to hire a 20-hour assistant. Before we do, we're spending twelve weeks testing whether AI can take the email drafting and listings, so we might hire for fewer hours. Nobody's job is changing because of this. If the extra work lands on you and it's too much, tell me, because that's one of the things we're measuring." The last sentence matters most: it makes workload a measure of the trial, not a complaint. A straight answer early, and a look at whether AI will replace your employees, heads off most of that worry.

Questions owners ask at this decision

Is it fair to my existing staff to use AI instead of hiring?

It can be, if the extra work doesn't simply land on them. The fairness test is whether their week gets more manageable or less. Ask them before you decide, include them in the trial, and treat their feedback on workload as one of the measures. If they're absorbing hours on top of full jobs, the trial has answered the question: hire.

Can AI cover for someone on long-term leave?

Partly. It can take on drafting and admin tasks the absent person did at a desk, which reduces how much a temporary cover role needs to do. It can't cover the person's presence, relationships or judgement, and someone still has to check its output. Most small businesses end up with shorter-hours temporary cover plus AI rather than AI alone.

What if I hire and AI later makes the role smaller?

Plan for it when you write the role. Describe it around outcomes (customers answered within a day, stock listed within a week) rather than a fixed list of tasks, and include using AI tools as part of the job. Then time saved turns into capacity for work you couldn't do before, instead of a role that shrinks awkwardly.

Further reads

• How to Test Job Candidates' AI Skills in an Interview — If you hire, check candidates can actually work with AI.
• How to Hire Your First Customer Service Person Alongside AI — Hiring a person to work alongside AI from day one.
• AI Receptionist vs Front-Desk Hire: The Real Cost for a Salon — The same decision worked through for a front-desk role.
• How to Calculate AI ROI for Your Business (Worked Example) — Put a return figure on the AI route with a worked example.
• How to Talk to Staff Who Fear AI Will Take Their Job — How to discuss the decision with a worried team.
• Why AI Saves Time but Not Money, and How to Capture the Gain — Why hours saved don't always show up as money saved.
• Do I Need AI in My Business? A Decision Guide for Owners — A decision guide for owners: the three situations where AI is worth it, when waiting is sensible, a ten-question score and a 30-day test.
• What AI Costs a Small Travel Agency and What It Saves — The monthly AI bill for a one-to-three adviser travel agency, set against a task-by-task ledger of the hours it can give back.
• How to Build a Restaurant Staff Rota With AI Demand Forecasts — Turn a covers forecast into an hourly staffing plan and a fair rota, with a worked Saturday that shows the labour-cost difference in dollars and hours.
• Hire an AI Specialist or Upskill Your Agency Team? — When an agency should train the team it has, when a specialist hire pays for itself, and the hybrid most agencies under twenty people end up with.
• How to Write Job Adverts With ChatGPT Without Biased Language — A four-step method for bias-free job adverts with ChatGPT, with a requirements sheet, drafting and audit prompts, sample outputs and a full before and after.
• AI Receptionist vs Front Desk Hire: A Dental Practice Comparison — What a dental front desk really does, where an AI receptionist beats a hire and where it can't, and a worked decision for a practice replacing a leaver.
• AI Lead Generation vs Hiring a Salesperson: A Cost Comparison — Itemised costs of an AI prospecting stack and a salesperson, a caterer's year-one comparison per meeting and per client, and when each wins.
• AI Assistant vs Virtual Assistant: Which Should You Pay For? — A side-by-side comparison, a food truck owner's week sorted task by task, real costs of each and when a small business needs both.
• AI Marketing Tools vs Hiring a Marketer: A Cost Comparison — Tools cost far less than a salary, but only replace the drafting. A task audit, three set-ups priced over a year, and signs you need a person.
• AI Recruiting Tools for Small Businesses: What's Worth Paying For — Nine kinds of AI recruiting tool, what each costs, which a small business should pay for, and three to skip, with examples from real hiring tasks.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.