Most founders walk into an automation sales call with no number in their head, which is why an AI automation cost calculator can be useful before you hear a price. They hear a price, feel a jolt, and make the decision on instinct. That is a terrible way to buy anything. It gives you the two figures that make a quote meaningful: what the work is costing you right now, and what it is worth to stop doing it.
This post walks you through building that number yourself, on paper, in about twenty minutes. No tool required. I will show you the four inputs that matter, the two most common ways founders get the maths wrong, and how to read a vendor’s price once you have your own figure to compare it against.
Key Takeaways
- An AI automation cost calculator is only useful if you feed it your real numbers: hours per week, loaded hourly rate, revenue leaking out, and a payback window you will hold yourself to.
- Your loaded hourly rate is salary plus on-costs divided by productive hours, usually 25 to 40 per cent higher than the headline wage. Most founders undercount it badly.
- Revenue leak, the deals lost to slow follow-up or dormant contacts, is almost always larger than the labour saving, and almost always the figure that gets left out.
- Automation costs split into three buckets: build, run, and change. Vendors quote the first, mention the second, and rarely mention the third.
- A sensible payback window for an operational automation is three to six months. Anything past twelve months is a strategy bet, not an efficiency purchase.
- Founder hours are worth more than staff hours, because the alternative use of them is growth work nobody else in the business can do.
- If you cannot name the task, count the hours, and describe the trigger, you are not ready to price it. You are ready to audit it.
- Knowing your own number before the call changes the conversation from “how much is it” to “does this beat the number I already calculated”.
Why You Need Your Own Number Before Anyone Quotes You
Here is what happens on most automation sales calls. The vendor asks a few questions, talks about what their system does, and lands on a price. The founder has nothing to measure that price against, so the only question available is “does that feel like a lot of money?” It usually does. The founder says they will think about it. Nothing happens.
The problem is not the price. The problem is that the founder is being asked to evaluate a number in a vacuum.
Now run the same call where the founder has already done the maths. They know that the quoting and follow-up process eats 11 hours a week across two people. They know their loaded hourly rate. They know roughly what a lost lead costs them. They walk in holding a figure, say $38,000 a year, and the vendor’s price either clears that bar or it does not. The decision takes ninety seconds instead of three weeks.
That is the entire point of an AI automation cost calculator. It is not a quoting tool. It is a decision tool. It converts a vague sense of “we’re too busy” into a number you can hold a proposal up against.
There is a second benefit that founders underrate. Doing the maths forces you to name the tasks. You cannot calculate the cost of admin overload until you have written down what the admin actually is. Half the people I do this exercise with have a small realisation partway through: the thing eating the most time is not the thing they came in worried about.
The Four Inputs That Actually Matter
Every honest automation cost calculation runs on four numbers. Everything else is decoration.
1. Hours per week, counted properly
Pick one recurring task. Not “admin”. A specific, repeating thing with a trigger and an end state. “Responding to new website enquiries.” “Chasing unsigned quotes.” “Taking booking calls.” “Rekeying job details from email into the CRM.”
Now count the hours it takes across everyone who touches it, per week. Count honestly and count the interruptions. A task that takes 15 minutes but arrives eight times a day is not two hours of work, because every arrival costs the person a context switch. Research on interrupted work suggests it takes a meaningful chunk of time to get back to full focus after an interruption, which is why a fragmented two-hour task often burns closer to three hours of usable capacity.
Most founders guess low on this input by about half. If you are not sure, spend a week writing down the start and stop times. The real number is almost always uncomfortable.
2. Loaded hourly rate, not the wage
The rate you plug in is not what you pay someone per hour. It is what that hour actually costs the business. Take salary, add employer contributions, leave, tools and software seats, insurance, training, recruitment amortised over expected tenure, and a share of the space and equipment. Divide by genuinely productive hours, which is not 40 a week; it is closer to 30 once you account for leave, meetings and the general friction of a working week.
For most small businesses in New Zealand and Australia, this pushes the effective rate 25 to 40 per cent above the headline wage. A $70k admin role is not $36 an hour. It is closer to $48 once you do it properly.
For your own hours, the calculation is different and more brutal. Your hourly rate is not what you pay yourself. It is the value of what you would do instead. If an hour spent on the business, on pricing, on a partnership, on a new service line, is worth more than an hour of quoting, then every hour you spend quoting costs you the difference. Most founders I talk to value their own hours at three to five times their admin rate, and they are usually still being modest.
3. Revenue leak
This is the input founders skip, and it is usually the biggest one.
Labour saving is what you stop spending. Revenue leak is what you stop losing. They are not the same, and the second number is nearly always larger.
Three places to look. First, response time. Research from InsideSales and Harvard Business Review on lead response found that the first business to respond wins the deal roughly 78 per cent of the time. If you take four hours to call a new enquiry and a competitor takes four minutes, you are not losing on quality. You are losing on the clock. Multiply your monthly enquiry count by your close rate gap by your average deal value, and you have a number.
Second, the dormant database. Every business over two years old has contacts who enquired, got busy, and were never followed up. Nobody has time to work that list, so it sits there. We reactivated a finance broker’s dormant list, 319 contacts his team had written off entirely, and it produced $49,000 in recovered revenue. That was not new marketing spend. That was a list he already owned.
Third, missed calls and after-hours enquiries. Count the calls that ring out. Multiply by your close rate and average job value. Most service businesses find a five-figure annual number sitting in the calls nobody picked up.
4. The payback window you will actually hold
Decide, before you see any price, how long you are prepared to wait to get your money back. Write it down.
For an operational automation that touches a known process, three to six months is reasonable. Six to twelve months is defensible if the automation also removes a key-person risk or unblocks something else. Beyond twelve months, you are not buying efficiency anymore; you are placing a strategic bet, and that is a legitimate decision, but it should be made consciously, not accidentally.
Writing this down before the call is what stops you from rationalising backwards once you are emotionally invested in a solution.
Running the Calculation
Here is the sequence. Do it on paper.
Step one: annual labour cost of the task. Hours per week, multiplied by loaded hourly rate, multiplied by 48 working weeks. A task taking 11 hours a week at $48 an hour costs $25,344 a year. If four of those hours are yours at $150, split the calculation: 7 hours at $48 plus 4 hours at $150 is $936 a week, or $44,928 a year. Notice what happened. The founder’s hours more than doubled the cost of the same task.
Step two: annual revenue leak. Response-time losses plus dormant database value plus missed-call losses. Be conservative. Halve whatever you first write down, because a recovered lead is not the same as a closed deal. If the halved number is still significant, it is real.
Step three: add them. That total is your current annual cost of not automating. This is the figure you carry into the sales call.
Step four: decide what share is actually removable. No automation takes 100 per cent of a task. A realistic target for a well-scoped process automation is 60 to 80 per cent, with a human staying in the loop on exceptions and anything involving judgment. So multiply your total by 0.7 for a working estimate of the annual benefit.
Step five: set your ceiling. Annual benefit divided by twelve, multiplied by your payback window in months, gives you the maximum you should pay in the first year, including both the build and the running cost.
An example. A plumbing business with 11 hours a week across quoting and follow-up, of which 4 are the owner’s, plus a conservative $22,000 a year in response-time and missed-call losses. Total exposure: roughly $67,000. At 70 per cent removable, the annual benefit is around $47,000. With a six-month payback window, they can justify spending up to about $23,500 across build and first-year running costs, and they should expect to be ahead from month seven onward.
That is a real number, derived from their own business, ready to hold a quote up against.
The Three Cost Buckets Vendors Do Not Always Separate
Your calculator needs a cost side as well as a benefit side. Automation costs come in three distinct buckets, and they behave very differently.
Build
The one-off cost to design, configure and install the thing. Mapping the process, writing the logic, connecting the systems, testing it against real scenarios, training the people who will supervise it.
This is the number most vendors quote, and most founders fixate on. It is also the least important of the three, because it happens once.
The honest range here is wide. A single, well-defined automation on top of tools you already own sits at the low end. A connected system touching several parts of the business, with a proper process audit in front of it, sits considerably higher. What drives the difference is almost never the AI. It is how messy the process is underneath.
Run
What it costs per month to keep operating. Platform subscriptions, the CRM layer, per-message or per-minute usage costs for voice and SMS, model API costs, and any hosting.
Running costs for AI automation have collapsed over the last two years and keep falling. This is genuinely one of the cheaper parts of the equation now. But watch the usage-based lines: a voice system priced per minute is fine at 200 calls a month and a different conversation at 2,000. Ask the vendor to model your actual volume, not their example volume.
Change
The bucket nobody quotes and everybody pays.
Your processes will shift. A new service line appears. A supplier changes their format. Someone realises the automation should also do the thing next to it. Somebody needs to own supervising it, reading what it did, and correcting it when it gets something wrong.
I have seen more automation projects fail on this bucket than on any technical problem. The build works. Nobody owns it. Six months later it is quietly switched off, and everyone concludes “AI didn’t work for us”. The MIT NANDA research on enterprise AI found that around 95 per cent of AI initiatives produce no measurable return, and the pattern in the failures is not that the technology was inadequate. It is that the work around the technology never happened.
When you are pricing automation, ask directly: who owns this after it goes live, what does ongoing support look like, and what happens when the process changes. A vendor who has a real answer is worth more than one who is cheaper on the build.
What the Calculator Cannot Tell You
Three things will not show up in your spreadsheet, and all three matter.
Key-person risk. When one person holds a process in their head, the business carries a liability that does not appear on any P&L until the day they resign or get sick. Automating the process does not just save hours; it moves the knowledge out of a head and into a system that cannot quit. There is no clean dollar figure for that, but if you have ever had a critical person leave, you know what it costs.
Compounding. The first automation is a discrete saving. The second is cheaper to build because the groundwork exists. By the fourth or fifth, you are not automating tasks anymore; you are building the brain of your business: a layer that knows how you operate, sees your numbers, and handles the repetitive work without being asked. Each piece makes the next one easier and more useful. A straight per-task cost calculation always undervalues the later ones.
What you do with the recovered time. Fifteen hours a week back is worth whatever you point it at. Point it at growth, and it is the best money you ever spent. Point it at nothing in particular, and it quietly fills back up with different admin within a month. I have watched both happen. The founders who get the most out of automation decide in advance what the freed time is for, and they write it down.
Reading a Quote Once You Have Your Number
Walk into the call with three figures written down: your annual cost of the status quo, your realistic annual benefit at 70 per cent removal, and your ceiling.
Then ask the vendor four questions.
“What exactly is in the build price and what is not?” Process mapping, integration with the tools you already run, testing, training, and the handover should all be in there. If the proposal is a single line item with no breakdown, ask for one.
“Model the running cost at my actual volume.” Give them your monthly call, message and enquiry counts. Make them do the sum in front of you. If they cannot, they have not thought about your business.
“Who owns this in month four?” Listen for whether there is a real answer involving a real person and a real process, or a vague reference to support being available.
“What is your estimate of the benefit, and how did you calculate it?” If their number is wildly larger than yours, ask which input differs. Sometimes they are right, and you have undercounted your own leak. Sometimes they are quoting a best case from a different business. Either way, you will learn something in the next thirty seconds.
If the quote clears your ceiling comfortably, buy it. If it sits just above, the question is whether the build is bigger than the one you priced, which is often true and worth understanding. If it sits far above with no explanation, walk. You now have the ability to walk without second-guessing yourself, because the decision is anchored to a number you calculated from your own business rather than a feeling you had on a call.
One last thing. If you run the calculation and the numbers are small, do not automate. That is a legitimate outcome and a good calculator has to be able to produce it. Plenty of businesses are better served by fixing the process first, or simply by doing the task differently. The point of working out the number is to find out whether the thing is worth doing, and “no” is an answer the exercise is allowed to give.
Conclusion
The AI automation cost calculator that actually matters is the one you build yourself, with four inputs: real hours, loaded rate, revenue leak, and a payback window you commit to before you see a price. It takes twenty minutes on paper, and it changes every automation conversation you have afterwards, because you stop reacting to numbers and start comparing them.
The bigger problem sitting underneath all of this is that most founders cannot answer these questions quickly, because the information lives in their head rather than in a system. The task list is not written down. The lead leak has never been counted. The dormant database has never been sized. That is the real constraint, and it is why the first useful step is usually not buying an automation. It is getting the business out of your head and into something that can see it. An AI brain that knows how your business runs, and an AI workforce that does the repetitive work, is what comes after that. Not before.
If you want a second opinion on your number, or you want help counting the inputs properly, book a 30-minute discovery call. Bring whatever figures you have, even rough ones. Half an hour is usually enough to tell you whether the thing you are considering is worth the money, and I will tell you plainly if it is not. If you would rather start with the leak side of the equation, the Revenue Recovery Calculator will size your dormant database in a couple of minutes.
For more on the ground underneath this, read how to get from $1 million to $5 million with AI automation, and if you have been burned before, what to do after a bad AI automation vendor covers how to reset without writing off the category entirely. The 90-day AI automation checkpoint is the follow-up exercise once something is live.
Frequently Asked Questions
How much does AI automation cost for a small business?
It depends almost entirely on how messy the underlying process is, not on the AI. A single well-defined automation built on tools you already own sits at the low end. A connected system touching several parts of the business costs considerably more to build. Running costs have fallen sharply and are now a small monthly figure for most businesses. The useful question is not the price; it is whether the price clears the annual cost you calculated.
How do I calculate ROI on automation?
Add your annual labour cost for the task, using a loaded hourly rate rather than the wage, to your annual revenue leak from slow response, missed calls and dormant contacts. Multiply the total by roughly 0.7, since no automation removes a whole task. That is your annual benefit. Compare it against build plus first-year running costs, and divide to get your payback period in months.
What is a realistic payback period for AI automation?
Three to six months for an operational automation on a known process. Six to twelve months is defensible if it also removes key-person risk or unblocks something else. Past twelve months, you are making a strategy bet rather than an efficiency purchase, which can be the right call but should be a conscious one. Decide your window before you see a price, not after.
Is AI automation cheaper than hiring someone?
Usually, but that is the wrong comparison to lead with. A hire brings judgment, relationships and the ability to handle the unexpected. Automation handles volume and consistency without fatigue. The better framing is sequence: install the system first so the knowledge lives somewhere, then hire into it. A new person joining a business that has documented context becomes useful in days rather than months.
What costs do automation vendors usually leave out?
The change bucket. Build and running costs get quoted. What rarely gets discussed is who supervises the system, what happens when your process shifts, and who fixes it when something breaks. That gap is where most failed projects die: the build works, nobody owns it, and six months later it is switched off. Ask directly who owns it in month four.
How do I know which task to automate first?
Pick the one with the highest hours-times-frequency score that involves the least judgment. Repetitive admin with a clear trigger and a predictable output beats a complex decision-heavy process every time, even if the complex one feels more valuable. Most founders try to automate the hardest thing first, hit problems, and conclude automation does not work. Start where the win is quick and obvious.
Should I automate before I fix my processes?
Fix anything genuinely broken first, but do not use process improvement as a reason to delay forever. Automating a slightly imperfect process that runs consistently usually beats waiting eighteen months for a perfect one. The exception is a process nobody agrees on. If three people describe the same task three different ways, map it properly before you build anything on top of it.
About Octavius
Titus Mulquiney is the founder of Octavius AI, where he builds AI brains and AI workforces for founder-led businesses stuck running everything out of their own head. Twenty years in marketing, ex-Sony product manager, ex-GM Zeal NZ. Based in Auckland, working with operators across NZ, Australia, and the US. Connect on LinkedIn.