A Joust Tool. Free. No email required.
Operations Tax Calculator
Size the recurring labor cost AI could replace, workflow by workflow. Same formula we open every Joust audit with.
What is the operations tax?
Most companies size AI by what they pay vendors. The bigger number sits one layer deeper, in the recurring labor cost embedded in workflows that AI could partially replace. We call this the operations tax. It does not appear on a vendor invoice. It hides inside salaries, shared services, and senior individual-contributor time. Nobody cuts a check for it. So nobody measures it.
This calculator gives you a defensible first-pass estimate. It uses the same three-variable formula we open every Joust AI Operations Roadmap with: a workflow's annual hours, the loaded cost of those hours, and the share of those hours AI can do today with a human in the loop. Multiply, and you have the upper bound of what is recoverable per year.
It is rough by design. The goal is to size whether a workflow is worth a deeper look, not to produce a board-ready ROI. If the rough math says six figures, the deeper math (with implementation cost, tooling, and workflow redesign subtracted) usually holds up. If the rough math says less, the workflow is probably not your first one.
- Pick a workflow you suspect carries hidden cost. Start with one. Pattern-rich, high-volume work is the best first candidate.
- Plug in the three numbers. Hover the question marks for help on each one. Numbers stay in your browser.
- Add more workflows to see the cumulative tax across a function or business unit. Try the industry presets below to see what typical mid-market scenarios look like.
Run your numbers
| Workflow | Annual hours | Loaded $/hour | AI-replaceable % | Annual tax |
|---|
What each variable actually means
The methodology behind the formula
We use this three-variable formula because it isolates the variable that matters most (volume of replaceable hours) from the variable that gets the most attention but matters least (model accuracy). A few notes:
On annual hours. The right input is your own time-tracking or staffing data. Vendor productivity benchmarks are usually built from a different population and a different workflow. The hours that matter are senior-IC and senior-rep hours. AI replaces the cognitive work, not the data entry. Counting at the wrong rate produces the wrong number, both ways.
On loaded hourly cost. The fully-loaded rate (salary + benefits + overhead) is the rate the business actually pays for the work. Base wage understates the real cost by 30-40%. Common loaded ranges: $50/hr for early-career analyst work, $75/hr for mid-level operations, $90-110/hr for senior knowledge work. If you have payroll data, use it. If not, your CFO does.
On AI-replaceable %. This is the variable people argue about most and measure least. We anchor it against published productivity research (see references below) and against the patterns we see in audits. For high-volume, pattern-rich workflows (ticket routing, exception handling, document classification), 30-50% is typical. For novel or high-judgment work, 5-15%. We deliberately exclude hypothetical future capability. If a pilot cannot run tomorrow with current models, the number is zero.
On the order of magnitude. The output is intended to be accurate to one significant figure. If the tool says $416,000, treat it as "around four hundred thousand." The deeper modeling we do in the roadmap engagement refines the number; the rough cut is for triage.
How your numbers compare to typical roadmap findings
Aggregated annual operations-tax ranges per workflow, drawn from Joust mid-market roadmap findings. Use these as sanity checks, not benchmarks.
Ranges are per workflow, not per company. Most audits surface 3-5 workflows in this band, so total addressable tax typically multiplies the per-workflow number.
The trap (and how to avoid it)
- Implementation cost (one-time, amortize over three years). Includes audit, integration, model selection, change management.
- Recurring AI tooling cost (annual). Model API spend, vector store, observability, governance overhead.
- Workflow redesign cost (often the largest line, and often missed). The work to actually change how the team operates around the new tool.
If the math still works after those subtractions, run the pilot. If it does not, the workflow is not the right first one. The number above is the ceiling, not the wallet.
We run this calculation across three to seven workflows in five weeks. Fixed scope. Senior-led. The deliverable is a model your CFO can read and a roadmap your COO can run. Not a slide deck. Not a maturity assessment.
References and where this approach comes from
Joust internal sources
- Six Figures Hiding (Joust, 2026), the long-form companion to this calculator. joustagency.com/blog/six-figures-hiding/
- Joust AI Operations Roadmap findings, aggregated patterns across mid-market engagements (2024-2026).
External research informing the inputs
- Brynjolfsson, Li, Raymond. Generative AI at Work. NBER Working Paper 31161 (2023). Measured a 14% productivity increase in customer service tasks with AI assistance, with the largest gains for newer, less-experienced workers. Anchors the AI-replaceable % range for high-volume support workflows. nber.org/papers/w31161
- McKinsey & Company. The state of AI annual reports (2023, 2024). Documents cross-functional adoption, ROI realization patterns, and the persistent gap between AI tooling spend and value capture. mckinsey.com/capabilities/quantumblack/our-insights
- MIT Sloan Management Review and Boston Consulting Group. AI and Business Strategy research series. Surfaces the workflow-integration gap as the dominant reason mid-market AI investments fail to capture value. sloanreview.mit.edu
- GitHub. Quantifying the impact of GitHub Copilot (2022, 2023). Measured ~55% completion-time improvement on a controlled software task. Anchors the high end of pattern-rich knowledge work productivity gains. github.blog/research
What we deliberately exclude
- Vendor productivity claims unless backed by independent measurement.
- Studies that aggregate hours saved across knowledge work generally; they hide the variance that matters at the workflow level.
- "AI will replace X% of jobs" macro projections; they do not help operators size next-quarter investment decisions.