What one manual process costs, in one formula
Annual cost of a manual process = volume x (minutes / 60) x loaded hourly rate. Payback in months = build cost / (net annual saving / 12). Loaded rate is salary plus benefits plus overhead, never base pay. When payback lands under 12 months and the freed hours move to revenue work, automate. Above 18 months, or with nowhere to redeploy the time, the case is weak.
Why a small task is the most expensive sentence in your business
"It is just a small task" is how most manual processes survive for years without anyone counting the cost. No one lies. No one notices. The task persists because stopping to measure it feels slower than doing it. The formula in this article takes five minutes to run and usually surfaces a number that changes the conversation.
The volume trap: small minutes, real money
Volume is the multiplier that turns a small task into a real cost. A 4-minute job done 30 times a day adds up to 243 hours a year. At a fully-loaded rate of £30 an hour, that is over £7,000 before a single error is counted. The individual instance looks fine. The annualized total does not.
Asana research found that the average knowledge worker loses about 209 hours a year to duplicative work alone. That is more than five full working weeks, spent re-entering work that already happened once. An industry survey of US knowledge workers found over 40% spending at least a quarter of their week on manual, repetitive tasks (2017 data, directionally consistent with more recent numbers). The volume trap is not a cost-centre problem. It is a calendar problem.
The hidden tax of swivel-chair work
Swivel-chair work is pivoting between two systems to copy data from one into the other. It looks trivial. It almost always follows the same rule-based steps. And it carries a second hidden cost beyond time: every re-keying task is a context switch that takes minutes of mental recovery on top of the task itself.
A 2025 industry survey of 500 US professionals found that manual data entry costs companies roughly $28,500 per employee per year, with knowledge workers spending 9 or more hours a week on re-keying alone. That is not a rounding error. It is a salary-sized cost hiding inside business as usual.
The formula: price one process in five inputs
Annual cost = volume per year x (minutes per run / 60) x loaded hourly rate. Payback in months = total build cost / (net annual saving / 12). Five inputs feed the formula: volume, minutes, loaded rate, capture rate (what fraction of volume the automation handles), and build plus recurring cost. All five are knowable before you spend anything on a build.
The five inputs, defined
| Input | What it means | Where to find it |
|---|---|---|
| Volume per year | How many times the task runs annually | Count monthly runs, multiply by 12 |
| Minutes per run | Average time to complete one instance, start to finish | Time three real runs, take the mean |
| Loaded hourly rate | Salary plus benefits plus overhead, never base pay | Ask finance for total staff cost per head |
| Capture rate | Fraction of volume the automation handles without human intervention | Estimate conservatively: 70 to 90% for structured inputs |
| Build and recurring cost | One-time build fee plus annual tooling and maintenance | Vendor quote or internal estimate; include all licence fees |
Worked example: re-keying email orders into the system
The numbers below are chosen to demonstrate the formula method, not to represent a real client or an industry benchmark. Round, readable inputs are used intentionally so you can follow the arithmetic. Replace them with your own volume, rate, and cost figures to get your actual number.
| Input | Illustrative value | Working note |
|---|---|---|
| Volume per year | 5,200 orders | Say 20 orders a day, 260 working days |
| Minutes per run | 4 minutes | From email receipt to system entry, measured |
| Loaded hourly rate | £35/hr | Salary plus benefits plus overhead, not base pay |
| Annual manual cost | £12,133 | 5,200 x (4 / 60) x £35 |
| Capture rate | 80% | Well-structured email format; 20% handled manually |
| Net annual saving | £9,707 | £12,133 x 80% |
| Build cost (one-time) | £6,000 | Build fee plus first-year tooling |
| Payback period | 7.4 months | £6,000 / (£9,707 / 12) |
Real DigX client inputs, payback figures, and outcome data from specific engagements are held by the founders. The formula above is the one we use in every diagnosis. We will not publish client data without explicit consent.
For a complete before/after including the real process, the build, and the outcome, see the FIZI build.
From cost to decision: payback and first-year ROI
Payback under 12 months, with the freed hours going to revenue work: automate. Payback between 12 and 18 months: name exactly what the freed person will do instead, then decide. Payback above 18 months, or with no plan for the time: the case is weak. Fix the process first, then revisit.
Why instant ROI is a red flag
Vendors who promise instant ROI are usually using base pay instead of loaded rate, ignoring integration and maintenance costs, or assuming 100% capture on a process that rarely hits 80%. Deloitte measured average automation payback at 16 months in 2020, rising to 22 months by 2021/22. The same research found that over half of companies never calculate the cost reduction from automation at all. If a case study skips the formula, ask why.
Hard savings vs soft savings
Hard savings reduce a real cash outflow: fewer overtime hours billed, a contractor role removed, a software licence cancelled. Soft savings free up time that may or may not reach revenue. A 2025 study by LSE and Protiviti found that AI users self-report saving roughly 7.5 hours a week, worth approximately £14,000 per employee per year. The same study notes that value stays soft unless the freed hours are actively redeployed to revenue work.
Build your ROI case on hard savings. Treat soft savings as upside, not as an input. If you cannot name a specific thing the freed person will do instead, the business case will not survive scrutiny.
When NOT to automate
Do not automate a process you have not stabilized. Automation runs the broken process faster. If inputs are unpredictable, genuine judgment is needed at every step, or the task changes monthly, the build cost will be spent rebuilding. Diagnose and simplify first. Build second.
The ceiling on automation is also worth stating plainly. McKinsey estimates that generative AI could potentially automate activities absorbing 60 to 70% of employee time, but potential is not captured value. They put the half-of-activities milestone around 2045. The realistic near-term win is automating roughly one-third of activities across about 60% of roles. That is why you price one process at a time rather than a department.
- The process changes frequently: the automation cost becomes a maintenance bill.
- Inputs are unstructured or highly variable: capture rate will be too low to justify the build.
- Every case requires genuine human judgment: a rule-based engine cannot substitute.
- Payback exceeds 18 months and there is no plan for where the freed hours go.
- The process is not yet stable: automate the final form, not a work-in-progress.
A 15-minute self-diagnosis you can run today
You do not need a consultant to run the formula. Pick one task, time it over three real runs, count the monthly volume, and ask finance for the loaded rate. The whole exercise takes about 15 minutes and produces an actionable number.
- Pick one task that repeats at least 10 times a week and follows the same steps each time.
- Time three real runs. Average them. That is your minutes per run.
- Count monthly occurrences. Multiply by 12 for annual volume.
- Ask finance for the fully-loaded hourly cost: salary plus benefits plus overhead, per person.
- Multiply: volume x (minutes / 60) x loaded rate. That is your annual manual cost.
- Get a rough build estimate. Divide: build cost / (annual saving / 12). That is payback in months.
- Under 12 months payback, with hours going to revenue work: automate. Otherwise: diagnose the process first.
If you want to run the same logic across several processes at once, the waste calculator does it in about two minutes with guided inputs.
Frequently asked
Q.What is the loaded hourly rate and why not use base salary?
The loaded rate is total employment cost divided by annual working hours: salary plus employer taxes, benefits, pension contributions, and a share of office overhead. Using base salary understates the true cost by 25 to 50% in most businesses and makes automation look more attractive than it is.
Q.How do I estimate the capture rate?
Capture rate is the fraction of manual volume an automation handles without human intervention. Start with how structured your inputs are. A well-formatted email order form might reach 85 to 90%. A freeform request that varies every time might reach 40%. Use a conservative estimate in the initial case, then refine after a pilot.
Q.Should recurring tooling costs go into the payback calculation?
Yes. The payback formula uses total build cost plus first-year tooling against net annual saving. For years two and beyond, compare ongoing tooling cost against the annual saving. Some platforms bill per task or per execution, which can erode the saving at volume. That is a separate build-vs-buy question.
Q.What counts as redeploying freed hours to revenue work?
If an automation saves a salesperson 4 hours a week of data entry and they spend those 4 hours on calls, the saving converts to real revenue potential. If the same hours go back to admin elsewhere, the saving stays soft. Specify what the freed person will actually do before counting the hours as ROI.
Q.When does it make sense to bring in an external diagnosis?
Run the self-diagnosis when you have a clear, isolated process and an owner ready to act. Bring in an outside diagnosis when the process crosses departments, when people disagree on how it actually runs, or when the cost of the wrong sequence is larger than the cost of a proper audit. See the audit page for what that engagement looks like.
