What is the ROI and payback period on automating a process?
Value is hours saved multiplied by fully loaded hourly cost. Cost is the one-off build plus the monthly retainer. Payback is the build cost divided by monthly net saving. Observed 2026 build bands are $5,000 to $15,000 for a single workflow, $15,000 to $50,000 for multi-workflow, with retainers of $3,000 to $20,000 a month.
What goes in, and what comes out
In: People doing the work, hours per person per week, and how much of it is automatable
In: Fully loaded hourly cost of those people
In: One-off build cost and monthly retainer
In: Ramp period before the saving is fully realised
In: Error-rework hours per month, if you want them counted
Out: Monthly and annual hours saved
Out: Monthly net saving after retainer
Out: Payback period in months, including ramp
Out: Year one and three-year net position
Out: A shareable link that reproduces the model
The method, stated
This is a labour-substitution model. It converts hours saved into money at the loaded hourly cost you supply, subtracts the retainer, and divides the build cost by what is left to get a payback period. It models the inputs you entered; it is not a forecast and it does not know your business.
The formulas, in evaluation order
weekly_hours_saved = people x hours_per_person_per_week x (automatable_share / 100)
monthly_hours_saved = weekly_hours_saved x 4.33
gross_monthly_value = (monthly_hours_saved + rework_hours_saved) x loaded_hourly_cost
payback_months = build_cost / net_monthly, plus the ramp months you set
year_one_net = (net_monthly x 12 x ramp_factor) - build_cost
three_year_net = (net_monthly x 36) - build_cost - (0 if retainer already deducted monthly)
Every assumption this model bakes in
Loaded hourly cost means salary plus payroll taxes, benefits, paid time off, software seats and supervision. If you type in a bare wage the model will flatter the project by a wide margin.
Hours saved are treated as money saved. That is only true if the hours are redeployed to something valuable or the headcount genuinely changes. If neither happens, the saving is capacity, not cash, and the tool labels it that way.
A ramp period defaults to two months at partial benefit, because no automation delivers its full saving in week one. Set it to zero to see the optimistic case.
The retainer is deducted every month from month one, including during ramp. Build cost is charged in full at month zero.
Error-rework hours are optional and default to zero. Include them only if you can name the errors and roughly how long they take to fix.
No revenue upside is modelled. Faster turnaround and better data often produce revenue effects, and they are excluded here because they cannot be estimated honestly from four inputs.
What this model cannot tell you
It cannot tell you what share of the work is genuinely automatable. That number comes from watching the process, and it is almost always lower than the first estimate.
It excludes the internal cost of the project: the time your team spends specifying, testing and supervising the parallel run.
It assumes the process is stable. Automating a process that is about to change is how automation budgets get written off.
It ignores the cost of failure. An automation that silently does the wrong thing is more expensive than the manual process it replaced.
Automation build and retainer cost, observed 2026 bands
These are observed 2026 market bands, published so you can locate yourself inside them. They are not TrainYourAgent quotes, and nothing here is a promise about what your project will cost.
Single workflow build: $5,000 – $15,000 — One process, mapped, built, instrumented and handed over.
Multi-workflow programme: $15,000 – $50,000 — Several connected processes, usually sharing integrations and a data layer.
Ongoing retainer: $3,000 – $20,000 / mo — Operating the automations: monitoring, exception handling, changes as systems shift.
How is automation ROI actually calculated?
Four numbers, in order. How many hours a week does this consume across everyone who touches it. What share of those hours can genuinely be automated. What is an hour of that person's time actually worth to the business, fully loaded. And what does the build and the ongoing operation cost. Multiply the first three to get a monthly gross saving. Subtract the retainer to get a monthly net. Divide the build cost by that net and you have a payback period in months, which is usually the only number a decision-maker actually wants. The reason to compute payback rather than a percentage return is that percentages hide the shape of the risk. A project with a 300% three-year return and a nineteen-month payback is a materially different proposition from one with the same return and a four-month payback, because you only have to be right about the first four months in the second case.
What is a fully loaded hourly cost and why does it matter so much?
Because it is the multiplier on everything, and it is the input people get most wrong. Fully loaded means the total cost to the business of having that person available for an hour: base pay, employer taxes, benefits and insurance, paid leave spread across working hours, the software seats they consume, their share of management time, and recruiting cost amortised over expected tenure. Using a bare hourly wage typically understates the real figure substantially, which inflates the payback period and can kill a project that was actually sound. Using a billable rate does the opposite and inflates the case. The number you want sits between them and comes from your own finance function. This calculator does not default it to anything, because a default here would be a number we invented that then propagated into someone's business case.
Is hours saved the same thing as money saved?
Only if something changes as a result. This is the honest objection to every automation business case and it deserves a direct answer rather than a footnote. If automating a process frees eight hours a week across a team and nobody's role changes, you have bought capacity. That capacity is real and often valuable, because it absorbs growth without hiring, or it moves people from data entry onto work that generates revenue. But it does not reduce next month's payroll, and presenting it as though it does is how automation projects lose credibility with finance. Say which one you are buying before you build. If it is cash, the plan has to include the headcount or contractor change that realises it. If it is capacity, the business case should be framed against the hire you are avoiding, and that is a legitimate and defensible frame. The calculator computes the money figure because that is what people ask for, and labels it as a model of your inputs rather than a saving that will appear on a P&L.
What does automation cost to build and to run?
Observed 2026 bands put a single-workflow build at $5,000 to $15,000, a multi-workflow programme at $15,000 to $50,000, and an operating retainer at $3,000 to $20,000 a month. The retainer is the line that surprises people, and it is not padding. Automations break because the systems around them change: an API version deprecates, a form gains a field, a vendor changes an export format, a rule changes at the end of the tax year. Something has to notice, and something has to handle the exceptions the automation refuses. The reason the retainer band is so wide is that it covers two different products. At the bottom it is monitoring and break-fix. At the top it is a team continuously extending the automation estate as the business changes. Ask which you are being sold before comparing two retainer numbers. TrainYourAgent publishes its own automation pricing rather than quoting on request, and it sits inside these bands: $5,000–$50,000+ to build · $3,000–$20,000/mo to operate.
Which process should I automate first?
The one that is high volume, low judgement, stable, and already documented. Those four conditions together are rarer than they sound, and a process that fails any one of them will overrun. High volume, because the saving scales with repetition and the build cost does not. Low judgement, because encoding a rule someone applies by feel takes far longer than automating a rule someone can write down. Stable, because automating a process that is about to be redesigned wastes the entire build. Documented, because if nobody can describe the current process precisely, the first month of the project is discovery and the estimate was wrong. The most common expensive mistake is automating the process that annoys people most rather than the one that costs most. Irritation and cost correlate poorly.
Which process should I automate first? — in detail
High volume: the same task many times a week, not a hard task once a month
Low judgement: rules that can be written down without hedging
Stable: nobody is planning to change this system or process this year
Documented: someone can describe every step and every exception today
Measurable: you can count how often it runs and how long it takes now
How do I calculate ROI on automation?
Hours saved per month times fully loaded hourly cost, minus the monthly retainer, gives net monthly value. Divide the one-off build cost by that net and you get payback in months. Everything else is presentation.
What does a single workflow automation cost to build?
The observed 2026 band is $5,000 to $15,000 for one workflow mapped, built, instrumented and handed over. A multi-workflow programme runs $15,000 to $50,000.
Why is there a monthly retainer at all?
Because automations sit between systems that change without asking. Observed retainers run $3,000 to $20,000 a month, covering monitoring, exception handling and the changes required when an upstream system moves.
What hourly rate should I enter?
The fully loaded cost: pay, employer taxes, benefits, leave, software seats and supervision. Not the bare wage, which understates it, and not a billable rate, which overstates it.