Workflow economics · 9 July 2026 · 8 min read

What is Copy/Paste Tax?

It is the recurring cost of moving, rekeying, checking, formatting, and repairing information by hand because a workflow does not flow. The individual actions look small. Their annual burden often does not.

The calculator produces a planning estimate, not a savings guarantee. Your inputs stay in your browser.

Start with one workflow
Use observed effort
Separate waste from judgement
Test before rollout

A useful definition

The work between the systems.

Copy/Paste Tax is not literally limited to keyboard shortcuts. It includes the human glue required when information, rules, and responsibility are scattered across inboxes, documents, portals, spreadsheets, and business applications.

01

Finding

Searching for the latest version, chasing a missing attachment, locating the right record, or asking which source is authoritative.

02

Moving

Copying fields between tools, renaming files, reformatting tables, rebuilding standard reports, or translating one system’s labels into another’s.

03

Repairing

Reconciling mismatches, removing duplicates, correcting transcription errors, recreating lost context, and checking that the transfer was complete.

Do not count the wrong thing

Friction is not the same as valuable judgement.

A financial analyst questioning an unusual variance, an operations lead deciding how to handle an exception, or a reviewer protecting a high-impact decision may be doing essential work. The tax is the avoidable handling around that judgement—not the judgement itself.

Likely workflow tax

  • Entering the same fact into more than one system
  • Rebuilding a recurring document from stable sources
  • Checking routine transfers for predictable mistakes
  • Waiting because ownership or status is invisible

Likely human value

  • Interpreting a genuinely ambiguous situation
  • Approving an action with material consequences
  • Challenging weak evidence or an unusual exception
  • Taking responsibility for the final decision

A defensible estimate

Measure representative periods before annualising them.

Choose one recurring workflow with a clear start and finish. Observe normal and peak periods or ask the people doing the work to keep a short activity log. Separate normal handling, waiting, corrections, and genuine decision time.

The simple burden model

Annual handling cost = people × average handling hours per person per week × working weeks × loaded hourly cost.

Add identifiable rework or external fees only when they are not already included. State every assumption. A range is usually more honest than a single precise-looking number.

Keep four figures beside the euro estimate

  • Cases processed per week
  • Median hands-on minutes per case
  • Share of cases needing correction or escalation
  • Elapsed time from request to completion

Worked example

Use a range, not theatre.

Suppose three people each spend between two and three hours a week assembling the same operational report. At 46 working weeks and a loaded rate supplied by the organisation, calculate both ends of the range. Then record rework separately. This is a recovery hypothesis: it shows the size of the opportunity, not what an implementation will save.

Baseline rule Observe first Use representative cases, document assumptions, and keep uncertainty visible.

Where to look first

Follow the handoffs.

The best candidates are rarely described as “copying.” Listen for phrases such as “I export this first,” “then I clean it up,” “we check every row,” “only one person knows which version,” or “I spend Friday assembling the update.” These sentences reveal boundaries where systems and ownership stop lining up.

  • Recurring reporting: gathering figures, normalising labels, writing the same commentary, and distributing versions.
  • Document intake: collecting attachments, extracting fields, checking completeness, and routing exceptions.
  • Inbox operations: triaging requests, retrieving context, drafting routine responses, and updating status.
  • Assessment work: comparing submissions with criteria, assembling evidence, and preparing a reviewer’s view.
  • Data reconciliation: matching records, identifying differences, resolving duplicates, and documenting corrections.
  • Knowledge retrieval: searching across approved sources and rebuilding an answer with traceable evidence.

Choose the smallest fix

A visible tax does not automatically justify an AI agent.

Remove unnecessary approvals. Agree one source of truth. Improve a form. Add a validation rule. Use an existing product feature. Connect two systems with deterministic automation. AI becomes interesting when the remaining work depends on language, messy documents, retrieval, or bounded interpretation—and when its output can be evaluated and reviewed.

Simplify

Delete the step, clarify ownership, or standardise the input before automating anything.

Automate

Use rules and integrations for stable, testable transformations and transfers.

Assist

Use AI for bounded variable work, with evidence, exceptions, and a human decision boundary.

From estimate to evidence

What makes the business case credible?

A baseline alone is not enough. Define the target before building, run a small end-to-end slice on representative cases, and compare the new workflow with the old one. Measure handling time, output quality, exception rate, waiting time, review effort, and operating cost. Record failure cases too.

A promising candidate

It repeats often, consumes meaningful skilled time, has representative data, has one accountable owner, and allows important outputs to remain under human review.

A reason to pause

The process is rare, the burden is trivial, ownership is missing, success cannot be observed, data cannot be accessed responsibly, or a simpler change solves the problem.

Turn friction into a testable decision

Calculate. Check. Prove.

Estimate one workflow’s annual burden privately, check whether the evidence foundations exist, then use a ten-working-day sprint only if the workflow earns further investment.