How Much Does Business Automation Cost in 2026? The Real Math
An operator paying someone $52,000 a year to copy data between a CRM and a spreadsheet is burning roughly 25 hours a week on work a machine does for the price of a phone plan. That is the leak. The question every founder asks before signing is the wrong one. It is not "what does the software cost." It is "what does the whole thing cost once it actually runs and keeps running."
So let me answer the search you typed. The named answer is the Build-Not-License rule: the tool layer is cheap and getting cheaper, the build and the maintenance are where your money goes, and the only number that matters is payback against hours you stop buying. Below is the buyer-stage breakdown, a real cost table, a fully worked example with the months-to-payback shown, the questions to ask before you sign, and the honest part about who should not write this check.
The tool layer is the cheap part
Founders fixate on license pricing because it is the only number they can see before buying. It is also the smallest number in the deal. Two engines run the mid-market: Make.com and n8n.
The pricing models differ in a way that bites people. Make bills per operation, not per run. Every node step in a scenario counts as a billable operation, so a 10-step workflow firing 1,000 times a month is 10,000 operations, not 1,000. n8n bills per workflow execution, so the whole run counts once regardless of step count, and you can self-host the open-source core for the cost of a server. That single difference decides which engine is cheaper for your volume.
| Tier | Make.com | n8n Cloud |
|---|---|---|
| Entry | $9/mo, billed per operation | $24/mo, 2,500 executions |
| Mid | scales with operation volume | $60/mo, 10,000 executions |
| High volume | up to $299/mo | $800/mo, 40,000 executions |
| Self-host | not available | Free (open-source core) |
Check the live numbers on the Make pricing page and the n8n pricing page before you budget, because both move. We broke down the engine choice in depth in our Make vs n8n vs Zapier comparison for the mid-market. The short version: this line item is $24 to $800 a month. It is not the decision.
Here is the part the pricing page will not tell you. This layer keeps getting cheaper because connectors are now commodity. The hard engineering of talking to a Salesforce API or a Stripe webhook was written once and shared widely. The connector is free. Knowing which three of four hundred connectors your process needs, and in what order, is the part nobody has automated yet.
The four-line cost stack every quote should show
Most automation quotes show one number and hide the rest. Force the vendor to break the spend into four lines so you can see what you are buying.
Line one is the tool license: $24 to $800 a month, the engine itself. Line two is the build: scoping, integration, error handling, testing, billed once or amortized into the retainer. Line three is maintenance, the recurring cost of keeping it alive when an API changes. Line four is the cost of doing nothing, the hours your team burns today on the manual version. If a quote only shows line one, it is selling you software and calling it a solution.
The trap is buying line one and assuming lines two and three are free. A self-built automation on a $24 plan still costs you the 20 hours your operator spends wiring it, plus every hour they lose when it silently breaks. Put a real labor rate on those hours and the cheap plan stops looking cheap.
Run the four lines as one annual figure and the picture sharpens. Say the tool is $60 a month, the build is a one-time $7,000, and maintenance is $400 a month. Year one costs you $12,520; year two, with the build paid, drops to $5,520. Now hold that against line four: if the manual version eats 25 hours a week at a loaded $35 an hour, that is roughly $45,500 a year you spend whether or not you look at it. The automation costs a quarter of the leak it stops. That is the comparison that belongs on the table.
Where the money actually goes
The license rents the engine. It does not wire your stack together, handle the API that breaks at 2am, or catch the edge case that silently drops 4 percent of your leads. That work is the build, and the build is most of the cost.
McKinsey has reported for years that a large share of work activities are technically automatable with current technology, but capturing that value depends on integration and process redesign, not the tool itself. Read their automation research at McKinsey. Translation for your P&L: software is commodity, wiring is the spend. The same pattern runs through Harvard Business Review coverage of digital transformation: projects fail on process and adoption, not on the tooling budget.
Build versus maintenance
A build has two costs people conflate. The first is standing it up: scoping the process, mapping integrations, writing error handling, testing edge cases. The second is keeping it alive: when Shopify changes a webhook or your CRM renames a field, an unmaintained automation breaks quietly and you find out from an angry customer. Maintenance is the recurring cost most quotes hide.
A luup managed automation build runs $3,500 to $10,000 a month and ships live in 14 days. That band covers the build, the monitoring, and the human who fixes it before you notice. It is not a vague retainer where you pay for strategy and slides. It is a working system on Make or n8n that you own. We wrote about why open-ended retainers bleed operators in this breakdown of agency retainer math.
Why maintenance is not optional
Operators treat maintenance as insurance they can skip. It is closer to oil changes. Every automation depends on systems it does not control, and those systems change on their own schedule. A payment processor deprecates an API version. A CRM renames a field your workflow reads by name. None of these announce themselves. The workflow keeps firing, but it now writes to the wrong field or skips records it cannot parse, and the failure is silent because the automation has no opinion about whether its output is correct. The first signal you get is a customer asking why they never received the thing your system was meant to send. Maintenance is the cost of watching the seams before the customer does.
A worked example: the lead-routing build
Numbers in the abstract do not close a budget. A mid-market services firm gets 600 inbound leads a month across a web form, a chat widget, and a shared inbox. A coordinator reads each one, scores it, tags it in the CRM, and routes it to the right rep. Loaded cost of that coordinator: $4,500 a month. Time on this task: about 60 percent of the role, so call it $2,700 a month of routing labor.
The automation reads all three channels, applies the scoring rules the coordinator was applying by hand, writes the record, and routes it. Built on n8n at roughly 600 executions a month, it sits inside the $60 tier. The build, scoped honestly, is around $7,000: two days of process mapping, four days of integration and error handling, two days of testing against real historical leads to confirm the scoring matches what a human would do.
| Line | Before automation | After automation |
|---|---|---|
| Routing labor | $2,700/mo | $0 (redeployed) |
| Tool license | $0 | $60/mo |
| Maintenance | $0 | $400/mo |
| One-time build | $0 | $7,000 |
| Monthly net saving | - | $2,240/mo |
The monthly saving is the $2,700 of routing labor minus the $460 in ongoing tool and maintenance cost: $2,240 a month. Divide the $7,000 build by $2,240 and payback lands at 3.1 months. From month four on, the firm keeps $2,240 a month it was previously spending, and the coordinator is back on work that touches revenue. That is the math that signs the deal, and the math you should demand a vendor show you before you pay.
One-time build versus managed retainer
There are two ways to pay for the build, and they suit different operators.
| Model | One-time build | Managed retainer |
|---|---|---|
| Upfront | $5,000 to $25,000 per workflow | $3,500 to $10,000 per month |
| Maintenance | Yours to handle, or pay hourly | Included and monitored |
| Best for | Stable process, in-house operator | Multiple workflows, no internal owner |
| Risk | Breaks quietly, you find out late | Someone is accountable for uptime |
A one-time build looks cheaper on the invoice and often is, if you have someone in-house who owns it. The hidden cost shows up in month 4 when an integration breaks and nobody is watching. The retainer prices that risk in. Pick the one-time build when you have one stable workflow and a capable operator. Pick the managed band when your time is worth more on revenue.
There is a third pattern worth naming, because it is where most operators land: a one-time build with a thin maintenance contract bolted on. You pay the build once, then a smaller monthly fee, often a few hundred dollars, that buys a named person to fix breaks and a guaranteed response window. You are not carrying the full retainer, but you are also not betting your lead flow on nothing upstream changing. If a vendor offers only the two extremes, ask for this middle option by name.
What to ask before you sign
The fastest way to separate a real automation partner from a software reseller is to ask questions the reseller cannot answer. Bring these to the call.
Who owns the account? If the workflow lives in the vendor's Make or n8n account, you are renting access to your own process. Insist the build sits in an account you own, so ending the relationship does not end the automation.
What happens when it breaks at 2am? Ask for the monitoring setup and the response window in writing. "We'll keep an eye on it" is not a maintenance plan. A real answer names how failures are detected, who gets paged, and how fast a fix lands.
Can you show the payback math on this specific process? A partner who has scoped your process can put numbers in the four-line stack and show a months-to-break-even figure. A reseller pivots to feature lists. The pivot is the answer.
What is the run cost at my real volume? Make the vendor estimate operations or executions against your true monthly volume, not a demo. The gap between Make's per-operation billing and n8n's per-execution billing can be a factor of ten at mid-market volumes.
Common mistakes that wreck the math
The cost model breaks in predictable ways. Each of these turns a project that should pay back in a quarter into one that never does.
Automating the exception, not the rule. Operators often point automation at the weird 5 percent of cases because those annoy them most. The 5 percent is also the hardest to encode and the most likely to change. Automate the boring 80 percent that runs the same way every time, and leave the exceptions to a human. The boring cases are where the payback lives.
Buying tooling before scoping the process. Picking Make or n8n before mapping the steps is like buying lumber before you have a drawing. The engine choice should fall out of your run volume and step count, not the other way around.
Skipping the test against real history. An automation that scores or routes should be checked against months of past data before it goes live, to confirm it makes the same calls a human made. Skip this and a silent 4 percent error rate ships to production, showing up as lost revenue three months later.
The only number that closes the deal
Payback. Most automation projects return their cost in 3 to 8 months, measured against the ops hours you stop paying for. The math is blunt: take the fully loaded cost of the person or hours doing the work today, subtract the build cost, and count the months to break even.
If you pay a coordinator $4,500 a month and automation removes 70 percent of their manual load, you free roughly $3,150 a month in capacity. Against a $7,000 build that is not instant payback, but redeploy that person to revenue work and the equation flips fast. Use real labor rates from the Bureau of Labor Statistics if you want a defensible input. Run your own number first with our Open Loop Tax calculator, which prices the hours leaking out of your business right now.
One caution on the math: count freed capacity at its redeployed value, not the raw hourly rate. Saving a coordinator 20 hours a week only matters if those hours go to work that moves the business. The payback figure assumes you redeploy the time. If you do not, the only saving you bank is the gap between labor cost and tool cost, a far smaller number. The build does not generate the return. What you do with the hours it frees does.
Who this is not for
Honest part. Do not automate a process that changes every week. If the steps are still moving, lock them first, then automate. Automating a moving target means you pay to rebuild the same workflow over and over, and the ROI never closes. We see this constantly. When we audited 50 mid-market AI stacks, most broken automations were built on processes nobody had stabilized.
Skip the managed build if your volume is tiny, your process is one stable step, and you have an in-house operator who enjoys maintenance. In that case, a $24 n8n plan and a weekend is your whole project. The managed band is for operators whose time is worth more on revenue.
Skip it too if the process is not a process yet, just a habit living in one person's head. Automation forces you to write down every step and decision rule. If those rules do not exist in a form two people would agree on, you are not ready to buy a build. You are ready to spend a week documenting the work, which is cheaper and often reveals that half the steps were never necessary.
Your next move
Start with the math, not the demo. Take our free Closed Loop Audit to see where your stack leaks and whether automation pays back inside a quarter. See what 14-day builds look like in our case studies, read the full automation service breakdown, then book the build when the numbers make sense. The cheapest automation is the one you scope correctly before you pay for it.
Frequently asked questions
How much does business automation cost per month?
Tool licenses run $9 to $800 per month depending on volume. A fully managed build that someone designs, runs, and maintains for you costs $3,500 to $10,000 per month. The license is rarely the line item that decides the budget. The build, the ongoing maintenance, and the cost of the manual work you pay for today decide it, and that last figure is usually the largest of the four.
Is n8n or Make cheaper for automation?
n8n bills per workflow execution and self-hosting is free, so high-volume jobs stay cheap. Make bills per operation, meaning every node step counts, which surprises buyers running multi-step scenarios. Volume and step count decide which one wins. A high-step workflow at high volume favors n8n; a low-volume scenario where you value the polished interface can favor Make. Estimate both against your real volume first.
What is the ROI timeline on a business automation project?
Most automation projects pay back in 3 to 8 months, measured against the ops hours you stop paying for. If a build cannot show that math inside a quarter, the process is probably too small or too unstable to automate yet. Remember to count the freed hours at their redeployed value, since capacity you do not reuse on revenue work is capacity you have not really banked.
Why is the build more expensive than the software license?
The license rents the engine. The build is the wiring: integrations, error handling, edge cases, and a human who fixes it when an API changes. Maintenance is the recurring cost most quotes hide, and it is where unmaintained automations quietly break. The connectors are commodity; what you pay for is the judgment about which ones your process needs and in what order.
When should I not automate a process?
Do not automate a process that changes every week. Lock the steps first, then automate. Automating a moving target means you pay to rebuild the same workflow repeatedly, and the ROI math never closes. The same applies to a process that lives only in one person's head: write it down first, then decide whether a build is worth it.
The license is cheap. The build is the spend. The payback is the only number that matters - run yours before you sign.