Why professional-services firms are over-hiring in 2026, and the five-question audit that stops it happening to yours.
If you lead a law firm, accountancy, recruitment agency, or MSP, you probably have a job spec open in a tab right now. A paralegal, a junior accountant, a resourcer, a first-line engineer. Same role you posted eighteen months ago. Same role you will probably post again next year.
This document is about why most of those hires are a mistake, what the mistake is actually costing you, and the one-page audit you can run before your next post-the-role click.
I run a solo AI consultancy called lessClick. I build systems for mid-sized professional-services firms in the UK and US. I wrote this after two hundred first conversations with firm owners had the same pattern in them, and I got tired of saying it one at a time.
This is the short version. The full book, Don't Hire, goes deeper. If the argument here lands, put your email at the bottom and I will send it when it drops.
You wrote £30,000 on the job spec. That is the visible number. Here is what you are actually signing up for, per the CIPD, HMRC, and lessClick's own modelling from real firm engagements:
| Line item | Annual cost |
|---|---|
| Gross salary | £30,000 |
| Employer's NI (15% from April 2025) | £3,750 |
| Pension (3% auto-enrolment) | £900 |
| Benefits (medical, life, perks) | £1,500 |
| Workspace (desk, software, IT) | £4,000 |
| Management time (year one supervision) | £14,000 |
| Onboarding & training | £2,500 |
| Recruitment | £5,000 |
| Lost productivity months 1-6 | £9,000 |
| Fully-loaded first-year cost | £70,650 |
The visible number is forty-two percent of the real number.
And this is before the cost of a bad hire — the US Department of Labor puts it at a minimum of thirty percent of first-year earnings, and CIPD 2024 data shows forty-one percent of UK new hires leave within the first twelve weeks. Two in five, before probation.
The firms I work with, when they do this calculation honestly, are shocked for thirty seconds and angry for fifteen minutes. The question that follows the anger is the one that matters.
Your partners are not lazy. Your finance director is not ignorant. The reason firms default to hiring is that the hiring decision is the legible one.
When a partner meeting discusses how to clear a probate backlog, "we need to hire a paralegal" is a sentence every partner understands. It has a cost, a timeline, a role description, and a recognised body of professional experience behind it. "We need to build a system" is, for most senior partners, not a recognised move.
Four structural reasons the hiring reflex is so hard to interrupt:
None of this is a character flaw. All of it is a structural incentive. And the entire ecosystem — regulators, professional bodies, benchmarking, accounting conventions — rewards the hiring decision even when the economics have stopped supporting it.
I have pulled time-tracking data from dozens of junior roles across legal, accounting, recruitment, and MSP firms. The pattern is remarkably consistent. In a 37.5-hour working week for a typical junior paralegal:
Pattern work means: intake questionnaires, beneficiary letters, bank and IFA chases, IHT400 drafting, status updates, template correspondence. Work that follows a template, uses a fixed set of inputs, produces a predictable output.
Judgement work means: reading a bereaved client's distress on the phone, making a call on a non-standard IHT scenario, flagging an unusual clause for partner review. Work that needs a human.
This ratio holds across every sector I have worked with. Junior accountants: 65-85% pattern. Recruitment resourcers: 60-75% pattern. Tier-one MSP engineers: 65-85% pattern.
Your junior hire is eighty percent pattern work. You are paying qualified-adjacent rates for unqualified-adjacent work, while the twenty percent that actually needs qualification is getting crushed underneath the pattern workload.
A well-built AI system, handling the pattern-work portion of a junior role with a senior human reviewing every draft before it goes out, costs approximately:
Fully-loaded first-year cost of the alternative: £12,000-£25,000.
Compared to the £70,650 hire. One third of the cost. Every pattern task done. A senior reviewing every output before it ships. No recruitment cycle. No onboarding lag. No retention problem. No bad-hire risk. No institutional memory walking out of the door.
Allen & Overy deployed Harvey AI across 4,000+ lawyers in 43 jurisdictions. Published outcome: 2-3 hours/week saved per lawyer, 30% reduction in contract review time, 7-hour average saving on complex document analysis.
Ironclad AI: first-pass contract redlining cut from 40 minutes to 2 minutes. Orangetheory Fitness used it to standardise 1,000+ agreements at 80% faster-than-manual speed.
Unilever: screened 250,000 applications to hire 800 candidates. £1M+/year saved, 50,000 hours of recruiter time eliminated.
Clio: AI adoption among legal professionals jumped from 19% in 2023 to 79% in 2024. Clio estimates AI could add $27,000/year of billable capacity per lawyer.
Thomson Reuters (2025 survey of 2,200 professionals): AI saves 5 hours/week per professional, worth $19,000 per head annually. US legal industry: $20B in annual savings by 2029.
PwC (2025 Global AI Jobs Barometer, based on 1 billion job ads): AI-exposed industries have grown revenue per employee at 3x the rate of less-exposed industries since 2018.
Big Four receipts: KPMG cut UK graduate intake by 29%. PwC targeting a one-third cut to entry-level hiring over three years. HBR (October 2025): "AI is dismantling the traditional hiring model of professional services firms, which relied on large classes of junior associates to supply a handful of future partners."
The largest professional-services firms in the world have already made the decision. Mid-market firms copying the old hiring model in 2026 are copying a model the Big Four have already abandoned.
Before you post the next role, run it through this.
Not the job title. The problem. Write it as a sentence. "We cannot keep up with probate client intake because the partners are doing the drafting."
Use the table above. For a UK £30K junior, you are committing to £65K-£75K in year one. If you do not write this number down, you will keep making the decision on the salary line.
Ask the person already in that seat to keep a one-week time log, broken into 30-minute blocks. Score each block as pattern, mixed, or judgement. Sixty percent pattern or higher and the hire is the wrong default.
For most single-function automations in professional services, £12K-£25K in year one. Model costs + senior QA time + build + maintenance.
A bad hire costs 30% of first-year earnings minimum (US DOL) plus tail effects. A bad build costs you the build fee and you keep hiring as planned. The downsides are asymmetric: the build has a ceiling, the hire does not.
You will run into objections. The five most common, and the one-sentence response to each:
"Our work is too bespoke to automate."
The work has bespoke components. It also has pattern components. Automating the pattern components does not touch the bespoke ones.
"Clients want a human on the other end of the phone."
Agreed. The human is still there. They are doing the conversation, not the drafting.
"Our regulator will not allow AI in the workflow."
Your regulator has guidance on AI. It does not prohibit it. The guidance almost always requires a human in the loop, which is how every build we ship works.
"The risk of a hallucination is too high."
The risk of an unreviewed junior error is also high. The difference is that the system's errors are logged and reviewable.
"We tried AI. It hallucinates. It is not ready."
What you tried was probably a general-purpose chatbot on an unbounded task. A well-scoped system on a bounded task with a human gate is a different product.
Three moves.
Move 1: Run the audit on the role you currently have open. Thirty minutes. Even if the decision is to post the role, you will have made it on the right basis.
Move 2: Pick one pattern-heavy task across your firm and build the first system against it. Thirty-day project. Build cost £3K-£8K. Payback in month four or five.
Move 3: If you want help running the audit, I do a £5,000 Operational Diagnostic. Two weeks. Interviews, process review, ranked bottleneck list, and three first-build recommendations with scope and pricing. Designed to be useful whether or not you build anything with me afterwards.
Two weeks. Interviews with partners and senior ops people. A ranked bottleneck list. Three scoped first-build recommendations with pricing. Non-refundable. Commits in one working day.
See the full offer · £5,000 Or email godfrey@lessclick.co.uk with a line about your firm.Don't Hire is the long version of this argument. Sector-by-sector time audits for legal, accounting, recruitment, and MSP. The thirty-day build playbook. The human-in-the-loop quality control standard. The ninety-day firm-wide conversion. The honest chapter on what happens to the people you do not hire. And the full 94-source research dossier with primary citations — McKinsey, Goldman, Thomson Reuters, PwC, Deloitte, Clio, LexisNexis, CIPD, SHRM, HBR.
Drop your email. I will send it when it drops.
No list-trawling. No weekly newsletter. One email when the book is ready. You can reply to unsubscribe.
Citations available on request. All research findings drawn from published primary sources 2023-2026. Written by Godfrey Tundube, founder of lessClick. lessclick.co.uk