How much does it cost to automate a business process?

“Automate this” is the most common request we hear, and the price for it runs from a subscription fee to a five-figure build. That spread is three different kinds of work hiding behind one phrase. Here is how to tell them apart, what each costs, and what a quote should contain.

All dollar and hour figures below are market ranges as of August 2026, gathered from what independent consultants, small agencies, and automation platforms publicly quote or list. Where we give our own price, we say so.

What “automate one process” actually means

For a business with 5 to 100 people, a process is usually a chain of handoffs: something arrives (a form, an email, a job, an invoice), someone reads it, re-types part of it somewhere else, and then tells someone else it is done. Automating the process means removing the re-typing and the reminding. It rarely means removing the person.

That matters for cost, because the expensive part is not the software. It is deciding which steps can run unattended, which need a human to check, and what happens when the input is not what the automation expects. We wrote about why the process has to be settled before any software touches it. An automation built on a process three people do three ways costs more and breaks sooner.

The three cost bands

As of August 2026, nearly every small-business automation request we hear lands in one of three bands.

Band 1: a connection between two mainstream tools. Market range: $0 to $1,500 in setup, plus $20 to $100 per month. This is a job in the scheduling tool becoming an invoice in QuickBooks, or a form submission becoming a row in a spreadsheet and a message in Slack. Tools like Zapier, Make, and the native integrations inside Google Workspace, Microsoft 365, QuickBooks, and most CRMs cover this. Setup is hours, not weeks, and you may not need a consultant at all.

Band 2: a multi-step automation with human checkpoints. Market range: $2,500 to $12,000 fixed price from an independent consultant or small practice. This is where most paid work happens. Three or four steps, two or three systems, and at least one point where a person approves something before it moves on. Often there is a messy input in the middle: an email to sort, a PDF to turn into fields, a note to summarize. That is where a small amount of AI earns its place, and where the testing time goes. This is the shape of our Implementation Sprint, and the general consulting cost post has the four-week breakdown of where those dollars go.

Band 3: a small custom tool when nothing fits. Market range: $8,000 to $30,000 and up. Sometimes one of your systems has no integration, or the process is specific enough that no connector handles it. Then someone has to write software: a small internal app, a script that runs on a schedule, a tool that sits between two systems that refuse to talk. This band is legitimate but should be the last resort, and a quote here should explain in plain language why bands 1 and 2 would not work. Our custom software and automations page covers when we think it is justified.

What drives the cost inside each band

Four things move a quote, in roughly this order.

1. Number of systems. Each system is a connection to build, test, and watch. Two systems is a connection. Four is a project. The count matters more than which tools they are.
2. Messy inputs. A form with fixed fields is cheap to process. A free-text email, a scanned PDF, or a photo of a handwritten work order is not. Messy inputs are where AI helps and also where most of the testing hours go, because real data is stranger than the sample you hand over.
3. Exceptions. Ask how often the process does not go the normal way. If the answer is “about one in twenty,” the automation needs a path for that one, and that path is often half the build. A process with no exceptions is rare, and a quote that assumes there are none is optimistic.
4. Who maintains it. A no-code connection can be owned by whoever set up the email accounts. A custom tool needs someone who will notice when it stops and knows who to call. If that person does not exist on your team, the ongoing cost is either a support retainer or a quiet failure six months in.

A fifth driver others rarely mention: how settled the process is before anyone quotes it. Cleaning that up first lowers every quote you will get.

A short worked example

Say you run a 20-person HVAC company along the Wasatch Front. Jobs get booked in a scheduling tool, the tech finishes and texts the office a photo of the paper invoice, and someone in the office types it into QuickBooks and emails the customer. Call it 15 minutes per job, 30 jobs a week. That is roughly 7 hours a week of re-typing.

Band 1 version: the scheduling tool already integrates with QuickBooks, so the job creates a draft invoice automatically. The tech still texts a photo, and the office still fills in the line items, but the customer record and job details are already there. Market cost as of August 2026: a connector subscription and maybe a few hundred dollars of setup. Saves perhaps a third of the time.

Band 2 version: the photo goes to a shared inbox, an AI step reads the line items and drafts the invoice, and the office person approves it in QuickBooks with one click before it goes out. Market range for a build like this: $4,000 to $9,000 fixed, with two or three weeks of testing against real photos, because real photos are crooked and smudged. Saves most of the 7 hours.

Band 3 version: only if the scheduling tool is an old one with no integration at all. Then someone has to build the bridge, and the range climbs to $10,000 and up. Before paying that, ask whether switching to a scheduling tool that does integrate is cheaper. Often it is.

If you put a conservative $35 per hour on office time (your number will differ; use yours), 7 hours a week is around $12,000 a year. The band 2 build pays back inside the first year, and the band 1 version pays back in weeks. That arithmetic, not the technology, is what should decide the band.

What a quote should include

Whatever band you are in, a quote you can trust has these lines.

• The process, written out, step by step, including where the human checkpoints are. If the consultant cannot write it down, they have not understood it yet.
• Which systems are touched and which one is the source of truth when two disagree.
• What happens with exceptions. Not “we will handle edge cases.” The actual path: it stops, it flags someone, it goes to a queue.
• Testing against your real data, with a stated amount of time for it.
• Who on your team owns it afterwards, what training they get, and what documentation is handed over.
• A fixed price, and what would change it.
• The monthly cost of any subscriptions it depends on, because that is the part that keeps billing after the consultant leaves.

A quote missing the exceptions line or the ownership line is cheaper for a reason.

Where we start

We do not quote an automation from a phone call, because the band depends on things we cannot see over the phone. We start with a half-day on-site Opportunity Assessment from $750, which ends with a ranked findings memo and one quick win already working. If your process is a band 1 connection you can set up yourselves, the memo says so and you are done for $750. If it is a band 2 build, you get a fixed quote with every line above filled in.

Common questions

What is the cheapest way to automate a process?

Connect two tools you already pay for. Most mainstream software has a built-in integration or works with a no-code connector, and that usually costs a subscription fee plus a few hours of setup. It only gets expensive when the inputs are messy or the tools do not have a usable connection.

Why do quotes for the same process vary so much?

Because the quote depends on things you cannot see from the outside: how many systems are involved, how clean the data is, how many exceptions the process has, and whether testing and training are included. Two honest quotes can differ by a factor of three if one includes handover and the other does not.

Should we automate or buy a new platform?

Automate the handoff between the tools you have first. A new platform replaces everything, which means re-training everyone and migrating data, and the old process problems usually come along for the ride. A small automation fixes the one step that is actually broken.

Who maintains the automation after it is built?

Someone on your team should, and the quote should say who and how. A no-code connection needs a few minutes a month. A custom tool needs a named owner and a plan for when it breaks. If nobody will own it, do not build it.