There's a conversation we have regularly with operations managers, founders, and finance directors. It usually starts with something like: "We've got a process that takes the team about a day each week — could software help with that?" And when we dig in, "a day a week" turns out to mean three or four people, each spending four to six hours, on something a well-built system could do in seconds.
The costs were always there. They just weren't visible because nobody added them up.
This post is about how to do that calculation properly — and how to think about automation investment in a way that leads to decisions you won't regret.
The Hidden Cost of Manual Processes
Manual processes have a peculiar accounting problem: their cost is distributed across many small moments rather than sitting on a single line in the P&L. Nobody writes an invoice for "four hours of copy-pasting between systems." It just gets absorbed into the working week, normalised as the way things are done.
The categories that consistently add up to more than people expect:
Repetitive data work
Copying data between systems. Re-entering the same information in different formats. Reconciling spreadsheets. Generating reports by hand. These tasks feel trivial because each instance takes ten or twenty minutes — but they happen dozens or hundreds of times per week, and they require your most expensive resource (people) to do work that produces no value beyond its own completion.
Manual approvals and routing
Document flows that travel by email. Purchase orders that need three signatures via forwarded threads. Onboarding checklists managed in someone's inbox. Each individual approval is fast — but the aggregate latency, the dropped balls, the time spent chasing, and the cost of delays caused by the process (a hire delayed two weeks, a supplier invoice held up for thirty days) is material.
Bespoke customer communication
Individually written status updates, booking confirmations, follow-ups, and reports that are 90% identical every time but require a human to write them. Sales teams are particularly susceptible to this. When a senior salesperson is spending four hours a week writing boilerplate proposals, that's four hours not spent on actual selling.
Error correction
The cost of fixing mistakes made in manual processes is usually several times the cost of the original task. A data entry error discovered three steps downstream. A miscommunication from a handoff that wasn't documented. A report calculated with a stale formula. The error itself takes minutes. The investigation, correction, re-communication, and relationship repair takes hours.
How to Calculate What a Process Is Actually Costing You
Here's the framework we use when scoping an automation project. It's deliberately simple — the goal is a directional number, not a precise accounting exercise.
Annual process cost = (Hours per instance × Instances per year × Fully-loaded hourly rate) + (Error rate × Cost per error × Instances per year)
Work through it with a real example. Say you have an end-of-month reporting process that involves pulling data from three systems, reformatting it in Excel, and sending it to six stakeholders. A team member spends four hours on it. It happens twelve times a year. That person's fully-loaded cost (salary + employer NI + benefits) is around £45,000 — roughly £22/hour.
Time cost: 4 hours × 12 × £22 = £1,056/year
Modest. Probably not worth automating on its own. But now add: the report is wrong about twice a year. Each error takes half a day across two people to investigate and reissue, and once caused a delayed board decision worth an estimated £15,000 in opportunity cost.
Error cost: 2 × (4 hours × £22 + £15,000) = £30,176/year
Now the process is costing over £31,000 a year. Automating it starts looking very different.
Most businesses have five to fifteen processes in this range. Finding them is the first job.
Where Automation Delivers the Fastest Returns
Not everything is worth automating, and the payback period varies enormously. Based on projects we've delivered, here's roughly what to expect across common categories:
| Process type | Automation approach | Typical saving |
|---|---|---|
| Data entry & reconciliation | API integrations, ETL pipelines | 60–90% of time cost |
| Report generation | Scheduled data pipelines, dashboards | 80–95% of time cost |
| Document processing | OCR, structured extraction, AI parsing | 50–80% of time cost |
| Approval workflows | Custom workflow engine or integration | 40–70% of elapsed time, significant error reduction |
| Customer communications | Templated automation, LLM-assisted drafting | 30–60% of time cost |
| Compliance & audit logging | Automated record-keeping, event sourcing | Often eliminates audit preparation time entirely |
The Processes Most Worth Targeting First
When we help a business identify automation opportunities, we score candidates across four dimensions. The best starting points score high on all four:
High frequency
A process that happens daily is a better automation target than one that happens quarterly, even if the quarterly one takes longer. Frequency drives ROI, and it also means you'll discover edge cases and improve the automation faster.
High consistency
Automation works best when the process follows predictable rules. If the logic is "always do X when condition Y is met," it automates cleanly. If the answer is "it depends, and Sarah knows when to do what," there's a design challenge — though good automation should codify those rules, not ignore them.
Measurable output
You need to be able to tell whether the automation is working correctly. A process with a clear, verifiable output (a report, a payment, a dispatched notification) is easier to automate safely than one where the "right" answer is qualitative.
High cost of errors
Where mistakes are expensive — financial calculations, compliance records, customer-facing communications — automation often pays back not through speed but through accuracy. A system that's right 99.9% of the time versus a human process that's right 97% of the time is a transformative improvement at scale.
What Automation Actually Costs to Build
The honest answer is: it varies enormously, and anyone who quotes you a price without understanding your systems and processes is guessing. But here's a rough framework.
Simple integrations — connecting two systems that already have APIs, with straightforward logic — typically take days to weeks and cost in the low thousands. If you're connecting Salesforce to your accounting system to eliminate manual invoice creation, that's not a complex engineering problem.
Custom workflow automation — building a purpose-designed system to manage approvals, routing, and notifications — typically takes weeks to months and costs in the £10,000–£50,000 range depending on complexity. This is where you're replacing a meaningful chunk of a person's job, not just saving them twenty minutes.
Complex data pipelines and processing systems — automating document intake, multi-system reconciliation, or high-volume transaction processing — scale in proportion to the complexity of the rules and the number of edge cases. Projects in this category typically run £30,000–£150,000+, but they're also replacing processes that cost more than that annually.
The key question is always: what is the process currently costing you, and what would a year's saving be worth? If the answer to the second question is larger than the build cost, the automation is worth doing — the only question is payback period and risk tolerance.
What Gets in the Way
Automation projects fail or underdeliver for predictable reasons. The most common:
- Automating a broken process. If the manual process is chaotic — inconsistently applied, full of exceptions, dependent on individual judgment calls that aren't documented — automating it encodes the chaos. The right order is to clean up the process first, then automate it.
- Underestimating the edge cases. The core 80% of a process usually automates cleanly. The remaining 20% — the unusual scenarios, the exceptions, the error states — takes 80% of the engineering time. Scope with this in mind.
- No ownership of the automated system. Automation needs maintaining. APIs change, data formats drift, business rules evolve. Someone needs to own the automated system and have the access and knowledge to update it. If that person isn't identified before the project, the automation will degrade.
- Building instead of integrating. Before building a custom solution, check whether an existing tool already solves the problem well. The answer isn't always no-code or SaaS — sometimes custom is clearly the right call — but the analysis should happen explicitly.
A Practical Starting Point
If you're not sure where to begin, here's a straightforward audit exercise. Ask everyone in your team to track, for one week, every time they do something they've done in the same way more than five times before. Collect the list. For each item, estimate: how long does it take, how often does it happen, and what goes wrong when it goes wrong?
Most businesses that do this exercise surface three to five processes that obviously shouldn't be manual. Those are your starting points. The savings from fixing even one of them often fund the automation work for the rest.
F5 Dev builds custom automation systems for businesses that have outgrown manual processes — from simple API integrations to complex multi-system workflow engines. If you've got a process you suspect is costing more than you think, we're happy to take a look.