WORKFLOW COST
How to Calculate the Cost of Manual Work Before You Automate
Measure one recurring workflow for two representative weeks, separate labor from delay and rework, and turn a vague feeling of busyness into a credible next decision.
The short answer
To calculate the cost of manual work, choose one recurring workflow and measure four things:
- the hands-on time spent completing it;
- the time spent correcting, checking, and chasing it;
- the delay it creates for the next person or system;
- the skilled attention it repeatedly pulls away from more valuable work.
Start with direct labor:
Monthly workflow labor cost = cases per month × minutes per case ÷ 60 × loaded hourly cost
Calculate each role separately, then add the results. Keep recovered capacity, avoidable cash cost, faster cash collection, and prevented losses in separate columns. They describe different kinds of value and should lead to different business decisions.
The practical goal is a credible baseline for one workflow. After two representative weeks, you should be able to choose one next move: investigate automation, simplify or stabilize the process first, keep the important work human, or stop because the value is too small.
A small task becomes expensive through multiplication
A manager opens a message, finds a job number, checks two photos, corrects an hour entry, and updates a spreadsheet. It takes 15 minutes. On its own, that sounds harmless.
Now multiply it across people and working days.
Here is an illustrative example, created to show the calculation rather than report a client result:
15 people × 15 minutes per day × 22 working days × $80 loaded hourly cost = $6,600 per month
The multiplication is where the task stops being small. Repetition turns minutes into a standing operating cost.
This is also where bad business cases begin. The $6,600 is labor consumed by the current activity. It does not prove that $6,600 will disappear from payroll after a change. The team may redirect those hours, avoid overtime, delay a hire, respond to customers faster, or simply complete existing work with fewer interruptions. Each outcome has value, but only some create an immediate cash saving.
Actionable take: before dismissing a task because it takes “only a few minutes,” multiply it by its real frequency, every role involved, and the number of times it returns for correction.
Put a clear boundary around the workflow
“Reporting,” “sales admin,” and “payroll work” are too broad to measure well. Several triggers, outcomes, and owners are hiding inside each label.
A useful workflow boundary answers five questions:
| Question | Example answer |
|---|---|
| What starts the work? | A crew marks a job complete |
| What must be true at the end? | One approved job record contains all payroll inputs |
| Who owns that result? | Operations manager |
| What proves completion? | Validated record with approval and timestamp |
| Who or what depends on it? | Payroll preparation |
That boundary keeps the study honest. If a payroll delay begins with incomplete field records, measuring only the final spreadsheet work misses the part creating the cost.
Think of the workflow as a journey through a series of gates. The trip is only complete when the record reaches the last gate with the required evidence. Timing one gate tells you very little if the work spends two days waiting at another.
Actionable take: write the trigger and the end state in one sentence. If the sentence contains several unrelated outcomes or owners, split it and measure one path first.
Run a two-week workflow cost study
Two representative weeks are usually long enough to expose repetition, handoffs, and common exceptions without turning measurement into its own project. Choose a period that reflects normal operating conditions. Avoid a holiday week, an unusual shutdown, or the cleanest week of the year unless that is the condition you want to study.
Record the work as it happens. Memory tends to compress repeated checks and forget short interruptions. A timestamp, observation, system record, simple spreadsheet, or time-capture tool will give you a stronger baseline than a meeting built around recollection.
Capture at least these fields:
| Field | What to record |
|---|---|
| Date and case | A date plus a job, quote, invoice, lead, or request reference |
| Role | The role performing the action, rather than only the person’s name |
| Action | What they checked, copied, entered, corrected, approved, or chased |
| Hands-on minutes | Active time spent on the action |
| Wait or handoff | Where work paused and what it waited for |
| Correction | What was missing or wrong and where the defect entered |
| System or location | CRM, inbox, message, spreadsheet, paper, or field application |
| Completion evidence | The record, status, approval, or output proving success |
| Confidence | Verified fact, estimate, or unknown |
| Exception note | What made this case different from the normal path |
Include normal cases, difficult cases, cancellations, corrections, and work handled outside the official system. Ten unusually clean cases can make a fragile process look efficient.
Actionable take: observe first and explain later. Ask the team to record what they actually did, including workarounds, without asking them to defend why the process exists.
Calculate direct labor by role
Different roles change the economics. A salesperson preparing a quote, an administrator cleaning the record, and an operations manager resolving an exception should not be collapsed into one blended rate.
For each role, use:
Monthly role cost = monthly cases × average minutes that role spends per case ÷ 60 × that role’s loaded hourly cost
Loaded hourly cost should follow the method your finance team already trusts. It may include salary or wages, employer taxes, benefits, and other directly attributable employment costs. Label the rate as an estimate if those inputs are incomplete.
The following example is illustrative:
| Role and action | Volume | Average time | Loaded rate | Monthly labor |
|---|---|---|---|---|
| Field supervisor completes record | 220 cases | 8 min | $54/hr | $1,584 |
| Administrator checks and re-keys data | 220 cases | 6 min | $36/hr | $792 |
| Operations manager resolves rework | 40 cases | 10 min | $72/hr | $480 |
| Direct labor total | $2,856 |
The manager row uses 40 rework cases rather than all 220 cases. Applying exception time to every case would inflate the result.
For irregular workflows, calculate a reasonable range. A low, expected, and high case is more useful than a precise-looking average built from weak evidence.
Actionable take: keep the source beside every material number. A calculation is only as credible as its volume, time, and cost inputs.
Add the costs the labor formula misses
Direct labor is the visible layer. Many workflows become valuable to fix because of what waits, breaks, or requires attention around that labor.
Correction and rework
Count the time spent finding missing information, returning work, correcting records, checking the correction, and rebuilding context. Also record where the defect first entered. Fixing the collection point may remove more cost than making the correction step faster.
Delay
Measure elapsed time separately from hands-on time. A quote may require only 20 minutes of active work while sitting in three queues for two days. An invoice may be quick to prepare once the job evidence finally arrives.
Connect the delay to a real downstream event:
- quote sent;
- invoice issued;
- payroll prepared;
- next job scheduled;
- customer updated;
- approval completed.
Avoid assigning a dramatic financial value unless the link is supported. “Invoices wait an average of 2.3 days” is a strong verified baseline. “Every delayed day costs us $10,000” needs separate evidence.
Leadership attention
Short interruptions are easy to overlook because they disappear into the day. Record when an owner or manager has to locate a record, answer a routine exception, approve an unclear case, or ask for a status update.
You can value the direct time at the role’s loaded rate. Treat wider opportunity-cost claims with care. The useful first question is where that attention repeatedly goes and whether the process genuinely needs that level of judgment.
Dependency
Record what happens when a particular employee is unavailable. If work pauses because one person owns the access, customer history, exception rules, or personal spreadsheet, the workflow carries key-person exposure even when its labor cost appears modest.
Actionable take: for every recurring workflow, ask one downstream question: “What cannot continue while this is incomplete?” The answer often reveals the strongest reason to improve it.
Separate recovered capacity from cash savings
Suppose a change removes 50 hours of repeat work each month. What has the business actually gained?
| Value type | What it means | Evidence that makes it credible |
|---|---|---|
| Recovered capacity | The same team has hours available for other work | Before-and-after time study plus a named plan for the released time |
| Avoidable cash cost | A real expense can be reduced or avoided | Overtime removed, contractor hours ended, fee eliminated, or planned hire changed |
| Faster cash movement | A revenue or billing step happens sooner | Shorter quote, approval, or invoice cycle tied to a measured baseline |
| Prevented loss | Errors, penalties, write-offs, or missed work become less likely | Historical incidence and an agreed method for valuing avoided events |
| Reduced dependency | Routine work continues without one specific person | A second qualified person or controlled system completes representative cases |
Recovered capacity is often the first and most defensible result. It becomes financially useful when the business can name what those hours will now support: more jobs, faster responses, fewer overtime hours, a delayed hire, or more time for skilled work.
This distinction protects the decision. Calling every recovered hour “cash saved” makes an automation proposal look stronger in a spreadsheet and weaker under scrutiny.
Actionable take: finish the sentence “If we recover these hours, we will…” If nobody owns the answer, record capacity as capacity and leave the cash claim out.
Label facts, estimates, and unknowns
A baseline can contain estimates. It simply needs to show where confidence is strong and where evidence still has to be collected.
| Claim | Classification | Current source | Next check |
|---|---|---|---|
| 220 records are completed each month | Verified fact | System export | Recheck after the study |
| Each record takes about 12 minutes | Estimate | Manager interview | Observe representative cases |
| 15% of records require correction | Unknown | No reliable source | Sample and classify 50 cases |
| Approval rules are the same at every site | Assumption | Current process map | Compare sites and ask the process owner |
| Two systems disagree on completion status | Conflict | CRM and finance records | Define the approved source of truth |
The labels make disagreement useful. Instead of debating whose memory is correct, the team can decide which observation, record, or owner decision will close the gap.
Actionable take: do not wait for perfect information. Continue with labeled estimates when the decision is reversible; collect stronger evidence before a large, costly, or high-risk commitment.
Simplify the path before valuing automation
The study will often expose reports nobody uses, duplicate entry created by an old system, approvals that never change an outcome, and checks added after a one-off incident years ago.
For every action, ask whether it can change at least one of these:
- the operating result;
- the evidence proving completion;
- a necessary control or approval;
- the next handoff.
If it changes none of them, question why it exists. Removing an unnecessary step produces immediate value and reduces what a future system has to maintain.
This is the operational equivalent of clearing a blocked lane before building a faster vehicle. Speeding up every car will not solve a bottleneck created by an avoidable checkpoint.
Actionable take: mark each step keep, change, remove, or investigate before anyone marks it “automate.”
Turn the baseline into one decision
The measurement has done its job when it changes the next decision.
- Investigate automation: the cost or constraint is material, the path repeats, and the important rules appear stable.
- Simplify or stabilize first: avoidable work, weak evidence, changing rules, or unclear ownership prevents a responsible design.
- Keep human: judgment, negotiation, empathy, or material approval creates most of the value.
- Stop: the workflow is too rare, inexpensive, low-consequence, or disruptive to justify further work.
Write down the owner, the evidence behind the decision, the remaining unknowns, and the date of the next review. If you proceed, keep the same baseline. It becomes the before measurement used to prove whether the change worked.
Denis Rafailov
Denis has spent 14 years building business software and operational automation. AITOC uses this material to help operators make better workflow decisions before anyone recommends a build.