KodDeltaGuides

Hire or Automate: How to Decide Between an FTE and a Module

~8 min read

KodDelta builds the software side of this decision and scopes it against one repeating task: a single focused module is $3,000–6,000 and ships in 2–4 weeks, with the first clickable prototype in 2–4 weeks. The licence is perpetual, users are unlimited, and the source code is delivered to the client.

An operations manager comes to you with a headcount request. The reasoning is sound: order volume is up, two people are working late every Thursday, and the backlog does not clear. The request sits on your desk next to a quote for a piece of software that would take the same task off both of them. Both options cost real money in year one, and the finance case for each is built from completely different line items.

Ask what kind of task it is before you cost anything

The comparison is only meaningful for a repeating task with written rules. Sort the work in the proposed role into three piles first:

  • Repetitive and rule-clear. Data entry, matching, routing, chasing, reporting. This pile is the candidate for software.
  • Exception handling. Items that break the standard path and need someone to decide what to do. Software routes these; it does not resolve them.
  • Judgement and relationships. Negotiating with a supplier, calming an angry customer, deciding whether to make a commercial exception. This pile is not automatable and should not be.

If the first pile is more than half the role, keep reading. If the third pile dominates, hire, and stop here. The 5-question automation test gives a more granular way to sort individual tasks.

The comparison that actually matters

Cost is the last column people should look at, not the first. These characteristics differ structurally between a person and a module, and they persist for as long as the task exists.

CharacteristicNew FTESoftware module
Time to productive Recruitment, notice period, then ramp-up on your processes 2–4 weeks to a working focused module; first clickable prototype in 2–4 weeks
Nights, weekends, holidays Contracted hours; cover needed for leave and sickness Runs on a schedule regardless of the calendar
Repeating the same mistake Human error recurs; corrected by training and checking A fixed rule stops recurring — and a wrong rule repeats at scale until corrected
Capacity increase Double the volume generally means overtime or a second hire Double the volume runs through the same module at near-flat cost
Does knowledge stay in the company? Undocumented process knowledge leaves with the person Rules live in code and specification; source code is delivered
Attrition risk Resignation restarts recruitment and ramp-up None for the module; vendor dependency is managed by code ownership
Handling a case nobody anticipated Improvises, asks, escalates, resolves it Routes it to an exception queue and waits for a person
Negotiation and relationships The reason you employ people Out of scope, and should stay out of scope
First-year cost character Recurring run rate that continues every year and rises Mostly one-off capital, plus optional maintenance at 12–25% of build cost

Build the first-year cost yourself — with your figures

We are not going to publish a salary benchmark and pretend it applies to your market, your city and your grade. Fill in these line items from your own payroll and finance systems. The exercise takes about an hour and it settles most of these arguments on its own.

Side A: the hire, first 12 months

  1. 1. Gross salary

    The advertised figure for this grade in your market, for twelve months.

  2. 2. Employer contributions and statutory costs

    Social security, pension, insurance, any payroll taxes, statutory bonus payments. Your finance team already carries this as a percentage uplift on gross — use their number, not a generic one.

  3. 3. Hiring and equipment

    Agency or advertising spend, interview hours from existing staff valued at their loaded cost, laptop, licences, desk.

  4. 4. Ramp-up

    The number of weeks before the person reaches full productivity, multiplied by the shortfall. If someone takes eight weeks to reach full speed, roughly half of that period is unproductive salary.

  5. 5. Management overhead

    The hours per week their manager spends supervising, reviewing and running one-to-ones, valued at the manager's loaded hourly cost. It is small per week and it is not zero over a year.

Year-one hire total = 1 + 2 + 3 + 4 + 5. Note that lines 1, 2 and 5 recur every year afterwards, usually with an inflation uplift.

Side B: the module, first 12 months

  • Build cost. A single focused module is $3,000–6,000 over 2–4 weeks. A system spanning several departments is $8,000–15,000 over 4–8 weeks.
  • Annual maintenance. Optional, 12–25% of build cost. Zero in year one if the work is delivered inside the build period.
  • Your internal time. The hours your own people spend writing the rules down, testing the prototype and running the parallel period. Cost this honestly — it is the line buyers forget.
  • Hosting and licences. Your existing infrastructure in most cases; the licence itself is perpetual with unlimited users.

Year-one module total = build + maintenance + internal time + hosting. Only the maintenance and hosting lines recur, and maintenance is optional.

To convert the hours the task currently consumes into a comparable annual figure, use the cost of manual work before you compare the two totals. If the recurring-versus-one-off shape of the two columns is the crux of the decision, the five-year TCO comparison models it over a longer horizon.

When hiring is the right call

Software vendors are poorly placed to say this, so it needs saying plainly. Hire the person when any of the following is true:

  • The work needs judgement on incomplete information. A rule engine given ambiguous input produces confident nonsense. A person asks a question first.
  • The work is negotiation. Supplier terms, payment plans, commercial concessions. No module should hold that authority.
  • The work is a relationship. Key account management, complaint recovery, anything where the customer needs to know a specific named human is dealing with it.
  • The exception rate is high. If a third of items break the standard path, the exception queue becomes the job, and you have bought software to create a worse version of the same backlog.
  • The process is about to change. Automating a process that is being redesigned next quarter means paying twice. Write the new process down first.
  • The volume is genuinely low. A task performed a handful of times a month rarely repays a build, however irritating it is.

The usual answer is both, in a specific order

Most companies that run this exercise properly end up doing two things. First they automate the rule-clear volume so it runs straight through with an exception queue. Then they hire — often at a more senior grade than originally planned — for the exception handling and judgement work that is left. The hire is smaller, the role is more attractive to a strong candidate, and the run rate is lower than the original request.

The systems we build are shaped for that split: Elevatora SAHA for lift maintenance and field service management, Elevatora İMALAT for lift-component manufacturing tracking, SEMP Group's group-wide operations platform, and PAP for pharmaceutical import and export process tracking. In each, the software carries the repeating volume and named people own the exceptions.

Bring both columns of your first-year calculation and the task description, and we will tell you honestly which pile it belongs in. Request a quote, or read how we scope custom software against a single process.

Frequently asked questions

When is hiring a person clearly the better decision?

When the task requires judgement on incomplete information, negotiation, relationship management, or handling a high proportion of exceptions. Software repeats a rule; it does not read a room, push back on a supplier, or decide what to do about a case nobody anticipated.

How do I compare the cost of an FTE against a software build?

Build both first-year figures from named line items. For the hire: salary, employer contributions, recruitment, equipment, ramp-up time and management overhead. For the software: build cost, optional annual maintenance at 12–25%, and internal time for specification and testing. Use your own payroll figures — no benchmark substitutes for them.

Can software increase capacity without increasing cost?

For a fixed, rule-driven task, yes: processing twice the volume through the same module does not double its cost. A person handling twice the volume needs either overtime or a second person. That difference in the capacity curve is usually the deciding factor for high-volume back-office work.

What happens to process knowledge if the person leaves?

In a manual process, knowledge that was never written down leaves with them, and the replacement rebuilds it. In an automated process, the rules sit in code and in the specification document, so the knowledge stays with the company. KodDelta delivers the source code, so that remains true even if the vendor relationship ends.

Can we do both — hire and automate?

That is the common outcome. Automate the repetitive, rule-clear part of the role so the volume runs straight through, then hire for the exception handling and judgement work that remains. The hire is smaller and the role is more attractive to a good candidate.

Let's talk about what you need.

The 30-minute discovery call is free and carries no commitment.