What Is the Work Nobody's Doing Costing You?
Most ROI maths for a hire starts from a salary you were never going to pay. This starts from the misses instead: the invoices nobody chased, the follow-ups that died, the renewal that auto-renewed on bad terms. Price those, then compare three ways to cover them. Everything runs in your browser.
There are two sides to it. Current prices what leaving the work undone costs you today, and every line can be checked against your own records. Growth indicates what installing a named AI role could add, which is a scenario rather than a number you can verify. They stay separate on purpose.
Calculator
Base salary. Employer NI, pension, recruitment and kit get added for you.
What is going uncovered?
These start filled in with sensible figures so you have something to argue with. Correct anything you've got a real number for, untick what doesn't apply, and add your own at the bottom. The "how sure are you" setting widens or narrows the final range, so guessing is allowed as long as you say that's what you're doing.
Add your own category
Want the category-by-category breakdown?
The free result above gives you the total and the three-way comparison. The breakdown ranks every category biggest first, shows which of the six AI roles could own each one, flags the rows you marked as guessed, and gives you the order I would work them in. Optional, and the number stays yours either way.
What could a named role add?
This side is a different kind of number and it is worth saying so plainly. The Current tab prices work that is already going undone, and you can check every line against your own records. This one indicates what a role could add, which nobody can verify in advance.
So it never gives you a single figure. Every role shows a range, anchored to a mechanism you describe with your own numbers, and states what would have to be true for it to hold. Two of the six roles have no honest revenue mechanism at all, and they say so rather than inventing one.
Where the percentages come from: they are deliberately conservative and they are not drawn from a dataset. I would rather say that than dress a guess up as a benchmark. The mechanism above each range is the part worth arguing with, because those are your numbers, and the percentage is the widest band on the page for a reason.
Want to know how to test this before you spend anything?
The ranges above are only worth what their conditions are worth. The breakdown turns each condition into something you can actually check against your own data, tells you which role to test first and why, and gives you the one number to watch per role so you can tell afterwards whether it worked. It is closer to an experiment plan than a sales case.
Where these numbers come from
Two kinds of figure sit behind the defaults, and it matters which is which. Employer National Insurance at 13.8%, pension at 4%, a recruitment fee near 20%, replacement cost at 50 to 150% of salary, and unused software spend in the 10 to 30% range are all public, published numbers. You can check them without taking my word for anything.
The rest come from my own engagements: how often follow-ups really get missed, how many deals go cold while nobody is watching, and how much of each category an AI role realistically covers. No public dataset exists for those. They're my benchmarks, they're the ones you should argue with hardest, and every one of them is editable above.
The coverage percentages are deliberately unkind to my own product. Retention scores 10%, bad hires 20%, strategy 30%. An AI role can watch, draft, and remind. It can't keep someone who's decided to leave. If a calculator tells you otherwise, it's a brochure with arithmetic on it.
One warning on double counting: don't add "hours I get back" as a category alongside these. The time you save chasing invoices is already inside the invoice row. Counting it twice is the fastest way to build a number that collapses the first time somebody checks it.
Join the Drift Digest
One short email every couple of weeks: practical fractional CTO and founder-AI notes, new AI tells as models change, and a quarterly nudge to re-measure your voice.
Something went wrong. Please try again, or email web@ctoondemand.co.uk.
Common questions
How do you calculate the ROI of hiring someone?
Price the work that isn't getting done, then compare it with the cost of covering it. For each category, estimate how often the miss happens in a year and what it costs each time, and multiply. Some categories are continuous drift rather than countable events, so those use a base amount times a percentage. Revenue misses need a conversion step: chances per year times close rate times average value. Sum the categories and you have the annual cost of the gap. Compare that with the year-1 all-in cost of a hire, which is salary plus employer National Insurance at 13.8%, pension, a recruitment fee of around 20%, and equipment.
Why not just compare an AI subscription with a salary?
Because it assumes you'd actually have made the hire, and most founders wouldn't have. The role just went undone. A percentage-of-salary saving falls apart the moment someone asks whether the person was ever in the budget. Pricing the missed work directly is harder to argue with, because each line can be checked against your own records.
Can an AI agent replace an operations hire?
Not entirely, and any tool claiming otherwise is selling you something. Agents cover the watching and drafting well: chasing invoices, tracking renewals, catching unused software, preparing meetings, surfacing stale deals. They cover people-shaped work badly. Retaining someone good, making a hiring judgement, and deciding strategy all stay human. This calculator scores each category separately for that reason, and it will tell you to hire a person when the numbers say so.
What does a bad hire or a departure actually cost?
Published HR benchmarks commonly put the cost of replacing an employee between 50% and 150% of their salary, covering recruitment, lost productivity, and ramp time for the replacement. A bad hire caught late is often benchmarked near 30% of first-year salary. Both are ranges rather than fixed figures, which is why this calculator asks how confident you are in each number and widens the result accordingly.
Got a number that needs a second opinion?
Book a 30-minute call. If the answer is that you should hire a person rather than install anything, I'll say so, and it happens more often than you'd expect.