Automate or hire?

Currency

Most AI business cases are built on a vendor’s calculator, and those calculators answer a question nobody is asking. They assume day-one performance, no implementation cost, no rework and no one left to supervise. Put those back and a good many cases invert.

The number that decides it is the one nobody quotes: the containment rate the system has to clear before automating beats simply hiring. If the pilot is running below that line, the answer is hire — however good the demo was. Set the case below and watch where the line falls.

The call · 36 months · realistic

Automate.

To beat hiring, the system has to close 42% of in-scope work without a person touching it. This case assumes 65%: it pays back in month 17 and is worth €178k in today’s money over 36 months.

NPV by stance

Conservative: −€133k
Realistic: +€178k
Aggressive: +€463k

The stances disagree on the sign: the estimate, not the arithmetic, is what needs work.

Hurdle

42%

containment needed to beat hiring; this case assumes 65%

NPV at 12%

+€178k

automate versus hire, in today’s money

Payback

Month 17

when cumulative cost falls below hiring, before discounting

Hire path

13 FTE

by month 36, up from 8 as volume grows

Cumulative cost · months from go-live

Automate
Hire

People cost €4.13 a unit at today’s volume. Over 36 months hiring costs €1.9m and automating €1.7m, implementation included.

Stance

Each stance sets containment, ramp, rework, oversight and implementation together.

The work today

People doing this work8 FTE

Fully loaded cost per person€62,000

Salary plus employer taxes, benefits and tooling.

Units of work a year120,000

Invoices, tickets, reconciliations — whatever the unit is.

Volume growth+15%

Headcount scales with this; software mostly does not. That is the whole argument.

The system

Automatable share70%

How much of the volume is in scope at all.

Containment rate65%

Of the work in scope, what the system closes with no human touch. The mark on the track is the hurdle: 42%.

Ramp to steady state6 mo

Months from go-live until containment reaches the rate above.

Implementation€120,000

Build, integration, change management. The line most calculators omit.

Second-order assumptionsSeven more

Platform cost a year€90,000

Licences, hosting, model and maintenance: fixed whatever the volume.

Cost per automated unit€0.35

Inference or per-transaction fees. People cost €4.13 a unit today.

Rework rate10%

Automated work that comes back for a person to fix.

Rework cost multiplier1.5×

Fixing it later costs more than doing it right first time.

Permanent oversight0.30 FTE

Someone still owns quality, exceptions and the model.

Discount rate12%

Horizon36 months

The hurdle · NPV by containment rate

23 points above the hurdle

Every other input held. Where the line crosses zero, automating and hiring cost the same in today’s money. The hatched band clears the hurdle by too little to call.

Where this case is weak

None flagged

Nothing structural on these settings. That says the arithmetic holds — not that the estimates behind it are right.

Notes on the model

What it shows

The hurdle: the share of in-scope work the system must close untouched for automating and hiring to cost the same in today’s money. It is the number to take into a vendor conversation, and to test in a pilot before signing.
Whether the case survives a sceptical stance. If conservative and aggressive disagree on the sign, the number is not the answer yet — the estimate is.
Where the saving comes from. Headcount grows in whole people as volume rises; software mostly does not. That step is usually where the case is won.

How it works

Headcount is lumpy, software is not. The hire path buys whole people as volume grows. The automate path pays a platform fee, a cost per automated unit, and people for whatever the system does not close.
Benefit ramps, cost does not. Implementation is spent before go-live and containment climbs in a straight line to steady state. Assuming steady state from day one is the most common way these cases overstate themselves.
Someone still owns it. Oversight is permanent headcount, not a transition cost, and automated work that comes back carries a rework multiplier — because fixing it later costs more.
Both paths are discounted monthly at the chosen rate, with implementation spent at go-live. Payback is read off undiscounted cumulative cost; the hurdle is solved by bisection on containment, everything else held.

What it leaves out

Option value, morale, hiring lead times and the strategic cost of not building the capability. They are real — and they are the arguments people reach for when the arithmetic does not work, so they belong in the discussion, not buried inside the number.
The people left on the automate path are costed per unit of work, as if they came in fractions. In practice they are hired whole too, which flatters automation slightly.
Tax, financing and the accounting treatment of the build. The comparison is cash, before tax, and every default is illustrative rather than a benchmark.

Next experiment · 01

The close, compressed

Month-end as a critical path. Switch on automation task by task and watch the close collapse.

Compress the close

Using this on a real decision?

Talk it through

The defaults are illustrative. The inputs that decide the answer are the ones you only know from inside the business.

Get in touch