GuidesContact
Calculator

What one repetitive task costs you a year

Five answers about one task, and this page works out the hours and the dollars in it.

The result is an illustrative estimate, not a quote. It runs on your five answers and on our own scoping assumptions, and both are printed in full below the calculator. Some answers come back saying the task is not worth automating.

What the time costs
How much time it takes
Your estimate

Fill in the five fields and the estimate appears here. Nothing is sent anywhere — the arithmetic runs in this page.

Every assumption behind that number

Employer burden
The rate or salary you enter is multiplied by 1.25. Payroll tax, benefits, equipment and paid leave cost real money on top of the wage.
Weeks a year
52, with no deduction for holidays or leave. A weekly task usually still gets done in those weeks, by somebody.
Salary to hourly
An annual salary is divided by 2,080 hours, which is 40 hours across 52 weeks, before the burden multiplier is applied.
Automatable share
A range for the task type, multiplied by a factor for the platform. Both tables are printed below, in full.
Payback
Measured against the $4,500 one-workflow price. The faster end of the payback uses the optimistic end of the saving range.
Hours cap
60 hours a week per person. Above that, the entry is almost always a typo or a whole-team total in a per-person field.

None of this is a quote. A real number needs a look at your actual data and the systems it moves between, which is what the $1,500 audit is for.

The model

Where the ranges come from

These are CloudBuilder AI's own scoping assumptions, taken from what a build of each kind has usually been able to take over. They are not drawn from published research, and no study is cited for them anywhere on this page.

By task

The share of a task a build can usually take over, before the platform factor is applied.

Rekeying data between systems

70–90%

The same information typed into a second system. The mapping is mechanical, so most of it moves. Matching records that do not line up is the part that stays.

Document and invoice processing

60–85%

Reading a document and putting the values somewhere. Clean digital files go through. Poor scans, handwriting and genuine exceptions still need a person.

Recurring reports and exports

70–95%

Pulling the same figures on the same schedule. The pulling and the formatting go almost entirely. Deciding what the numbers mean does not.

Quote generation from a price list

60–85%

Assembling a priced document from known rules. Standard jobs go through. Anything needing a judgement call on price comes back to a person.

Follow-up and chasing sequences

70–90%

Sending the next message when a condition is met. The sending and the tracking go. A real reply is a conversation, and that stays with you.

Scheduling coordination

40–70%

Finding a slot several people can take. Calendars and confirmations automate well, the negotiating around them does not, which is why this is the lowest range here.

Answering repeat customer questions

50–80%

The same questions, answered from information you already publish. The repeat half goes. Anything specific to one account needs a person.

Data cleanup and deduplication

60–85%

Finding and merging records that should be one record. Rules catch most of it. The ambiguous pairs need somebody to decide.

By platform

What the platform does to that share. A system with no documented way in cuts it hard.

Mainstream software with an API

×1

Documented, supported interfaces to read from and write to. The share for the task stands as written.

A web app with no documented API

×0.6

Reachable through a browser session rather than a supported interface. It works, it breaks more often, and part of the task stays manual.

Paper or phone

×0.5

Nothing can be automated until the information is captured digitally. A capture step comes first and takes part of the saving with it.

Desktop-only, legacy or on-prem

×0.3

Automation means driving the screen, which breaks on every software update. Most tasks in this category are not worth automating.

Something else

×0.7

Feasibility cannot be read without knowing the system. This assumes there is a documented way in, which is the first thing an audit checks.

Some answers say do not automate it

A task can be genuinely annoying and still not be worth building anything around. Two conditions produce that answer here: less than an hour a week of total team time in the task, or a payback longer than two years even at the optimistic end of the range. Desktop-only software reaches the second condition often, because driving a screen breaks on every software update. The calculator returns that verdict as readily as the other one, and the arithmetic that produced it is printed on this page either way.

What the estimate cannot know

It has not seen your data, your exceptions, or the state the two systems are actually in. Those are what move a real project inside or outside a range, and they are the difference between an estimate and a price. A field filled in inconsistently for six years, an approval step nobody wrote down, or an API that exists but is switched off on your plan will each change the answer, and none of them are visible from a form with five boxes on it.

What to do with a number that looks worth chasing

The automation audit is $1,500 for half a day with your team and a written map of every repetitive task, with hours-per-month on each, ranked by what is worth automating first. It comes off any build in full. One workflow is $4,500, built and running in three weeks. An operations build covering three to four connected workflows is $12,000. You keep the map either way, including the parts of it that say a task is not worth touching.

Read next

Common questions

Where do the percentages come from?
They are CloudBuilder AI's own scoping assumptions, based on what builds of each kind have usually been able to take over. No study is cited for them because none is being cited. They are printed in full on this page so you can disagree with a specific one.
Why a range instead of one number?
Because one number would be false precision. How much of a task a build can take over depends on how clean the data is and how many real exceptions the task has, and neither of those is visible from five form fields.
Why multiply the wage by 1.25?
Payroll tax, benefits, equipment and paid leave are real costs of an hour of somebody's time. 1.25 is a common, conservative figure for that burden. The real one at your company may be higher.
Is this a quote?
No. It is an illustrative estimate built from your five answers and the assumptions printed on this page. A price needs a look at the actual systems, which is what the audit is for.
Does anything get sent anywhere when I use it?
No. The arithmetic runs in your browser and nothing leaves the page. The email box is optional, appears only after the result, and sends you a link to the scenario.
What if the system we use is not in the list?
Pick Something else from the list and name it. That case assumes there is a documented way in, which cuts the range by a factor rather than taking it at full value.