OptimisedI've sat in the meeting where a founder says "we're drowning, we need to hire someone" at least thirty times, and in more than half of those meetings, hiring was the wrong answer. Not because people are bad, but because the thing drowning the team was usually a pile of repetitive work that a £25 a month tool could have eaten for breakfast.
The automation vs. hiring question is one of the most expensive decisions a growing business makes, and most owners decide it on instinct rather than maths. That's how you end up paying £60,000 a year for someone to copy data between systems, or, on the flip side, how you end up with a shelf full of software nobody logs into while your best employee quietly burns out.
I've built automations for service businesses, run hiring processes, and cleaned up the mess when companies got this call wrong in both directions. So this article gives you the actual numbers, the honest failure modes of each path, and a decision process you can run this week.
The Short Answer
Buy software when the work is repetitive, rule-based, high volume, and doesn't require judgment or relationships. Hire when the work needs human context, negotiation, creativity, or someone to own outcomes rather than tasks. And when you're not sure, automate first, because the downside of a bad software purchase is a few hundred pounds and some wasted weekends, while the downside of a bad hire is 30 to 50 percent of that person's annual salary, according to figures compiled by Qureos from SHRM research. SHRM is the Society for Human Resource Management, the main professional body that benchmarks hiring costs in the US.
That's the compressed version. The rest of this article is about why that rule holds, where it breaks, and how to run the numbers for your specific situation instead of trusting a stranger on the internet.
What a Hire Actually Costs, and It's Not the Salary Line
Most owners think of a hire as a salary. That's the smallest lie in business accounting.
Start with just getting the person in the door. According to SHRM's most recent benchmarking data, summarised by Truffle, the average cost per hire in the US is now £4,300 for non-executive roles, up 21 percent from 2022. Executive hires average £28,300. That covers recruiting spend only: job boards, agency fees, recruiter time, background checks. The widely quoted £3,700 figure is out of date, and if anything, costs keep climbing. Obra's analysis notes that Appcast's 2026 Recruitment Marketing Benchmark Report found cost per hire rose another 6 percent in 2025, driven largely by job board price inflation rather than a tight labour market.
Then comes what the person costs once they're sitting at a desk. The Bureau of Labor Statistics reported in June 2026 that total compensation for private industry workers averaged £36.80 per hour worked as of March 2026, with wages making up 69.9 percent and benefits the remaining 30.1 percent. In plain terms, for every pound of salary you agree to, budget roughly 43 more pence for insurance, paid leave, retirement contributions, and legally required benefits like Social Security and unemployment insurance.
Qureos puts the true first-year cost of a US employee at 1.25 to 1.4 times base salary once taxes, benefits, equipment, and onboarding are included. So that £47,500 admin hire is realistically a £65,500 to £85,000 first-year commitment depending on your benefits package and location. Weeman Solutions ran the same maths for UK businesses and found a £35,000 hire commonly turns into a £50,000 to £60,000 annual commitment.
And you don't get full output on day one. Obra's breakdown of recruiting costs points out that indirect costs like recruiter hours, hiring manager time, and the productivity gap of an open role often exceed the visible spend. SHRM's average time to fill is around 44 days, and once someone starts, the common estimate is six to twelve months before a new hire reaches full productivity. Autonoly's comparison framework uses exactly that range, against a typical two to four month payback for a well-targeted automation.
None of this means hiring is bad. It means hiring is expensive, slow, and compounding, so it should be reserved for problems that genuinely need a human. Which brings us to the other side of the ledger.
What Software Actually Costs in 2026
The good news is that workflow automation has become absurdly cheap relative to labour. The three tools I get asked about most are Zapier, Make, and n8n, and they cover the majority of small business automation needs, so let's use real prices.
Zapier is the easiest to learn and the most expensive at scale. As of mid 2026, per Toolradar's detailed pricing analysis, the Free plan gives you 100 tasks a month with two-step automations, and the Professional plan starts at £15.80 a month on annual billing (£23.70 if you pay monthly) for 750 tasks and unlimited multi-step workflows. Zapier bills in US dollars, so the pound figures here are approximate at current rates. A task is one action, not one complete workflow, which is the detail that catches everyone. A five-step automation that runs 200 times a month burns roughly 800 to 1,000 tasks, so real teams usually land somewhere between £40 and £160 a month once volume grows. The Team plan runs around £55 to £82 a month depending on task allocation.
Make, formerly Integromat, is the value play. Its published pricing lists a Core plan at £9.50 a month for 10,000 credits and a Pro plan at £16.60, per Omid Saffari's August 2026 verification of the live pricing pages. Make also bills in dollars or euros rather than pounds, so those figures are approximate too. Make's own comparison obviously favours Make, but independent testing backs the cost gap: Toolradar modelled identical workloads and found Zapier consistently 4 to 15 times more expensive than Make for the same volume of work. The tradeoff is a steeper learning curve, maybe three days of fiddling before the visual scenario builder clicks.
n8n is the technical option: open source, self-hostable, with a cloud plan starting around €24 a month, billed in euros (roughly £20 at current rates), that charges per workflow execution rather than per step. Self-hosted, it's dramatically cheaper than anything else at high volume, but you're now running your own server, and if the words Docker and Postgres mean nothing to you, skip it.
Beyond general automation platforms, there's category-specific software: accounts receivable automation, appointment scheduling, AI phone answering, invoice processing. These typically run £40 to £320 a month for small businesses. US Tech Automations, an automation agency, frames the decision as whether your high-friction workflows involve enough volume to justify a £120 to £320 monthly platform, and for most businesses handling 20 or more manual interactions a week, they argue the answer is yes. They're a vendor, so weigh that, but the Goldman Sachs 10,000 Small Businesses survey data they cite is harder to dismiss: 71 percent of small businesses that deployed workflow automation reported positive ROI within twelve months, with a median payback of 7.4 months. ROI, return on investment, just means the value you got back exceeded what you paid.
So the raw comparison is stark. A meaningful automation stack costs £25 to £320 a month, or roughly £300 to £3,800 a year. A single junior hire costs £60,000 or more in year one. Even if the software only replaces a third of a role, the maths favours software by an order of magnitude. Which is exactly why the next section matters, because if the maths is that lopsided, why do so many software purchases fail?
The Shelfware Problem Nobody Mentions in the Sales Call
Here's the part vendors won't tell you, and it's the strongest argument the pro-hiring side has.
Businesses are terrible at actually using the software they buy. Zylo's research on shelfware, which is the industry term for software that sits on the shelf unused, found the average large organization wastes £15.6 million a year on unused licences. Their 2025 SaaS Management Index, cited by Marlborough Street Partners, found enterprises use only 47 percent of the seats they pay for. Vertice's 2025 SaaS Wastage Report, summarised by Last Rev, found 21 percent of applications are complete shelfware with zero meaningful usage, and another 45 percent are underutilised. Flexera estimates 25 to 30 percent of IT budgets are simply wasted on redundant tools and idle licences. SaaS, if the acronym is new to you, means software as a service, the subscription model nearly all modern business tools use.
Those are enterprise numbers, but I've watched the small business version play out repeatedly. An owner buys a CRM (customer relationship management software, the database that tracks leads and clients), pays for the annual plan to get the discount, spends one enthusiastic Saturday setting it up, and by March nobody has logged in since January. The subscription auto-renews. Nobody notices.
The failure is almost never the tool. It's that software automates a process, and if you don't have a defined process, you've just bought a very organised empty box. A human hire will muddle through an undefined process, ask questions, and figure it out. Software will do exactly nothing until you tell it exactly what to do.
So here's my honest rule: if you cannot write down the process you want to automate as a numbered list of steps with clear if-then logic, you are not ready to buy software for it. Fix the process first. This is also why Medius, writing about invoice processing, argues the right question isn't hire or automate, but how much of the work is still manual and why. Sometimes the answer is that the work itself is structured badly, and neither a hire nor a tool fixes that.
Where Software Wins Every Time
With that caveat on the table, there are categories of work where buying software instead of hiring is nearly always correct, and I'd push back hard on anyone hiring for these in 2026.
Data entry and re-keying. If anyone on your team copies information from one system into another, that's a solved problem. Zapier or Make will sync your form submissions, invoices, and orders between systems for pocket change.
Follow-up sequences. Lead follow-up, appointment reminders, payment reminders, review requests. These are the highest ROI automations in small business because they directly generate revenue. Abivo's analysis of accounts receivable found automated reminder software with clear payment links measurably increases on-time payment, and the pattern holds across industries. Nobody should be manually typing "just checking in on that invoice" in 2026.
Scheduling. Calendly, Cal.com, and their competitors cost £0 to £16 a month and eliminate the email tennis that used to eat an admin's mornings.
Reporting. If someone spends Friday afternoons building the same spreadsheet, connect the data sources once and let the report build itself.
First-line customer questions. Not full customer service, but the "what are your hours" and "where's my order" tier. Modern AI chat and phone tools handle these well, and the SBE Council's March 2026 Tech Use Survey found 82 percent of small business employers have already invested in AI and automation tools, so your competitors likely already run this play.
The common thread: rule-based, repetitive, high volume, low judgment. McKinsey's automation research, cited by Stealth Agents, estimates 57 percent of US work hours are already automatable with existing technology, and adopters see productivity gains of 20 to 30 percent in the first year. I find blanket percentages like that less useful than looking at your own task list, but the direction is right. Most businesses are sitting on automatable work right now.
Where a Human Wins and Software Embarrasses You
Now the other side, because I've also seen automation deployed where it absolutely should not be, and the results range from awkward to genuinely damaging.
Anything involving negotiation or nuance. A payment plan conversation with a struggling long-term client is not a workflow. Abivo makes this point well in the accounts receivable context: people handle disputes, read tone, and adapt when a situation falls outside the pattern. Software escalates or fails.
Relationship ownership. Clients don't build loyalty to your automation. If your business depends on someone knowing the client's history, preferences, and unspoken concerns, that's a hire.
Exception-heavy work. If more than roughly a quarter of cases need special handling, automation becomes an exception-routing machine that frustrates everyone. Automate the clean 75 percent, staff the messy 25.
Work that defines your quality. If you run a design studio, automating the design is automating away the reason clients pay you. Automate the invoicing around it instead.
Growth roles. Software executes; it doesn't spot opportunities, coach juniors, or notice that a whole service line is underpriced. Aplos AI's writeup on this exact decision frames it cleanly: automation removes the need for people doing work software does better, not the need for good people. Their case study of a service business owner is instructive. She automated her admin, then hired fourteen months later when growth demanded someone for client relationships, and because the admin was already automated, the new hire spent zero time on envelope-stuffing work.
There's also a category I'd call false automation candidates: tasks that look repetitive but carry hidden judgment. Payroll approval. Final proofreading of client deliverables. Anything where a rare error is catastrophic rather than annoying. Keep a human in those loops even when a tool exists.
Five Questions I Ask Before Spending Either Budget
When a client asks me whether to buy software instead of adding staff, I walk them through five questions. Answer them honestly and the decision usually makes itself.
Is the work rule-based or judgment-based? Write out the task. If you can express it as "when X happens, do Y, unless Z," it's automation territory. If your description keeps including the phrase "it depends," it's a human's job. Autonoly formalises a version of this with their SCALE framework, scoring scope, cost, availability, learning needs, and error impact, which is more structure than most small businesses need, but the instinct is sound.
Is the volume growing? This is Medius's core argument, and I agree with it: hiring spreads manual work across more people, but it doesn't remove the bottleneck. If invoice volume doubles, the hired team is drowning again in eighteen months. Automation capacity scales at nearly zero marginal cost. If the workload is growing and rule-based, automation compounds; hiring just delays.
What does an error cost? Automation makes fewer errors than humans on repetitive work, but when it fails, it fails identically at scale until someone notices. If an error costs a few minutes of cleanup, automate freely. If an error costs a client or a compliance violation, either keep a human review step or keep the human.
Is this a task problem or a capacity problem? Owners often say "I need to hire" when they mean "I'm doing eleven hours a week of admin." That's a task problem, and tasks can be automated one at a time. A capacity problem is different: you need someone to own a function, make decisions inside it, and be accountable for outcomes. Software cannot own anything. If you need an owner, hire.
Have you actually timed the work? Not estimated. Timed. Track two weeks of the task in a spreadsheet. Most people are shocked in one direction or the other. The task that "eats my whole day" turns out to be four hours a week, or the quick daily check turns out to be nine. You cannot compare a £60,000 hire to a £40 tool without knowing how many hours are actually at stake.
Run the Maths on One Real Workflow
Let me make this concrete with a composite example drawn from businesses I've worked with, a twelve-person home services company.
The pain: leads come in from the website, Google, and phone. Someone has to log each lead, send a first response, schedule an estimate, send reminders, and chase the quote afterwards. The office manager spends about 12 hours a week on this, and slow follow-up is visibly losing jobs to faster competitors.
Option one, hire a part-time coordinator at £17.40 an hour for 20 hours a week. That's roughly £18,100 a year in wages, and applying the BLS benefits reality, call it £23,000 to £25,500 all-in, plus a few thousand to recruit, plus ramp time. Real first-year cost: about £28,000. Capacity added: 20 hours a week, during business hours only.
Option two, build the automation. New lead triggers an instant text and email response, gets logged in the CRM, receives a scheduling link, and enters a reminder sequence; unanswered quotes get a three-touch follow-up automatically. On Make, this runs comfortably inside a £16.60 a month Pro plan. Add a scheduling tool at £12 a month and an AI phone answering service at around £80 a month for the calls that come in after hours. Total: roughly £109 a month, or about £1,300 a year, plus perhaps £1,600 to £2,400 if you pay someone to build and test it properly rather than doing it yourself. Call it £3,700 in year one, and about £1,300 a year after that.
The automation recovers maybe 9 of the office manager's 12 weekly hours and responds to every lead within a minute, around the clock. It does not handle the angry customer, the complicated estimate, or the supplier negotiation. It's not supposed to. Those stay with the human who now has nine extra hours for them.
That gap, £28,000 versus £3,700 for overlapping (not identical) coverage, is why the first-year ROI figures floating around the automation industry sound inflated but often aren't. Numbers like 200 to 400 percent first-year returns with two to four month breakeven show up repeatedly across benchmark roundups, and while I'd treat any vendor's specific claim sceptically, the direction matches what I've seen when automation targets a real bottleneck. The failures come from automating things nobody actually needed, which loops back to shelfware.
My Recommended Sequence
If you're staring at an overwhelmed team right now, here's the order of operations I'd follow, and it's deliberately boring.
First, list every recurring task your team does and time it for two weeks. This costs nothing and kills most bad decisions before they're made.
Second, sort the list into three buckets: rule-based and repetitive, judgment and relationship work, and hybrid. Be ruthless. "Sending the proposal" is hybrid: assembling it is automatable, pricing it is judgment.
Third, automate the top two or three rule-based tasks by hours consumed. Start on cheap monthly plans, not annual contracts, no matter what discount the vendor dangles. You want the freedom to be wrong for £16 instead of £190.
Fourth, measure for 60 to 90 days. Did the hours actually come back? Did anything break? Is anyone routing around the automation because it annoys them? That last one is the silent killer; an automation people bypass is shelfware with extra steps.
Fifth, and only now, look at hiring. What's left on the list after automation is judgment work, relationship work, and ownership. If those remaining hours justify a role, hire for that role, and you'll write a much better job description because the position no longer includes the robotic stuff. The hybrid model, software on the repetitive layer and humans on the judgment layer, is where nearly every efficient small business I know has landed, and the sources across this research say the same thing from different angles.
One warning on buying help: if you bring in an automation agency, ask them what they would not automate in your business. Anyone who says "everything can be automated" is selling you shelfware. The good ones will name your judgment-heavy work and tell you to keep humans on it.
An Honest Reality Check
Three things the enthusiastic version of this article would leave out.
Automation has maintenance costs. APIs change, tools update, workflows silently break. Budget a few hours a month for care and feeding, or pay someone £80 to £240 a month to monitor it. An unmonitored automation that's been failing quietly for three weeks can do real damage.
The ROI distribution is lopsided. PwC's 2026 AI performance research, cited in Automaton Agency's analysis, found a small share of companies capture most of the economic gains from AI and automation, largely because they redesigned their processes around the tools instead of bolting tools onto old processes. Buying software is the easy part. Changing how work flows is the actual project.
And sometimes hiring first is right even for automatable work. If your process is chaotic and undocumented, a smart temporary hire who works the process manually for six months will surface every exception and edge case, handing you a documented process that's then trivially automatable. Medius makes a version of this point for invoice teams: hiring can buy time while you fix the underlying structure. Expensive discovery, but cheaper than automating chaos.
The decision to buy software instead of adding staff isn't really a technology decision or even a cost decision. It's a clarity decision. Businesses that understand their own processes automate the right things, hire the right people, and pull ahead. Businesses that don't will buy either the wrong software or the wrong hire, and the invoice for that mistake arrives either way.
What to Do This Week
Pick your single most annoying recurring task. Time it for five working days, honestly, in a note on your phone. Multiply the weekly hours by 50, then by the loaded hourly cost of whoever does it (salary plus about 40 percent). That's the annual cost of the task. Now go price one tool that could handle it, using the free tier if one exists.
If the tool costs less than a tenth of the task, and it usually will, build the automation this weekend on a monthly plan. If the task turns out to be judgment work in disguise, congratulations, you've just written the first line of a job description instead. Either way, you've replaced instinct with numbers, and that's the whole game.