Business process automation services, built around how your work runs today

Who are business process automation services for?

A growing business often runs on a handful of systems that don’t talk to each other: a CRM, an accounting tool, a job or order system, a shared inbox and a few spreadsheets. People carry the work between them. An order arrives by email and someone types it into the job system. A finished job becomes an invoice because someone copies the details across. A new customer is entered in three places, with a typo in one of them. How that plays out depends on the size of the business:

• Established local businesses: a single location where the owner signs off on changes, and an office team that keys the same details into several tools. A good first project is one process the owner can name without thinking.
• Multi-location and regional businesses: several locations or a wide service area, where each office has its own way of doing the same task. Automating it means agreeing on one version of the process first.
• National businesses: high volumes, several teams and systems that each have an owner. The automation has to fit change control, access rules and the reports those teams already use.

If a process runs a few times a month, automating it may cost more than it saves, and you’ll hear that from us on the call. Our custom software page covers the rest of what we build, and the page on custom software cost explains what moves the price.

Which processes are worth automating?

The processes worth automating repeat often, follow rules someone could write down and move data from one system to another. Two numbers decide whether a process is worth the build: the hours it takes each week, and what a mistake costs when one happens. A task that takes ten minutes a day but sends a wrong invoice once a month can be worth more to fix than one that takes an hour and never goes wrong.

They tend to share three traits:

• They start from something structured: a form, a record in one system or a file in a fixed layout. Free text needs a different kind of step, covered further down.
• The rules fit on a page: if the person who does the task can explain every decision in it, those decisions can be written as rules.
• The result lands in a system: a record created, a status changed, a message sent or an invoice raised.

The manual work cost calculator on our AI automation page turns the hours a task takes each week into a yearly figure, and the free Operations Audit worksheet turns up to eight tasks into hours a year and a yearly cost, then scores how rule-shaped each one is. Our guide on what to automate first walks through choosing the first one.

How is a process mapped before it is automated?

We go through the process with the people who do the work, one step at a time. For each step we write down where the information comes from, which system it goes into, who makes the call and what happens when something is missing. That map becomes the plan for the build: each step turns into a rule, a connection or a hand-off to a person.

Mapping tends to turn up the parts nobody wrote down. A side spreadsheet one person keeps to track rush jobs. A step done differently for one large customer. A check that happens only because someone remembers to do it. An automation built from the official version of the process breaks on exactly those parts, so they go on the map too.

Sometimes the map shows the process should change before anything is automated: two approvals where one would do, or a field nobody reads. Automating a step that shouldn’t exist only makes it run faster, so we point those out on the map and you decide what stays.

What would workflow automation services build for your business?

It depends on the process, so these are examples of the shape the work takes at each level, written as what we would build:

• For an established local business: a quote request from the website creates a record in the CRM, posts a message to the team’s chat and opens a job in the scheduling tool, so nobody retypes it. When the job is marked done, a draft invoice appears in the accounting tool for someone to check and send.
• For a multi-location or regional business: new work is routed to the right location by its address or service, each location’s queue lives in one shared system, and a daily summary goes to the managers.
• For a national business: orders arrive from several channels and end up in one system of record, with a log of every change and a queue for any order that fails a check.

Some of this is plumbing between tools you already pay for. Some of it calls for a system of its own: a custom CRM when your records don’t fit the tools you have, or dashboards and reporting when the numbers you need sit in five places. Our industry pages set out what we would build for logistics companies, IT service providers and title companies.

What happens when an automated step meets a case it can’t handle?

It should stop and say so. An automation that guesses does more harm than one that pauses, because its mistakes look like finished work. What we would build sets the case aside with a note saying what was missing or odd, and sends a message to the person who handles it. The rest of the queue keeps moving while that one waits.

Three things make that work day to day:

• A record of each run: what came in, what the automation did and where the result went, so a wrong result can be traced back to the step that made it.
• Alerts when something stops: a message when a connection fails or a run doesn’t finish, sent to someone who can act on it.
• A way to pause it: a switch that stops the automation without stopping the systems around it, for the day a supplier changes the layout of its order form.

When does a process need an AI step?

When a step means reading something written for people. A rule can copy a field from a form. It can’t read an email in which a customer describes an order in their own words, or pull the figures out of a PDF that arrives in a different layout from each sender. Those steps need a model, and they belong with our AI automation work, where a model does the reading and code checks what it produces before anything is posted.

The steps around the reading usually stay rule-based: creating the record, routing it, sending the message, raising the invoice. So one workflow can mix the two. We build the rule-based steps as process automation and the reading step as AI automation, and the proposal names each part.

Should the automation run in Zapier, Make or n8n, or in custom code?

That turns on the volume, on how much the process has to remember and on who will look after it. A tool such as Zapier, Make or n8n suits a process that moves data between apps with ready-made connectors, at a volume the tool’s plan covers. Custom code suits a process that needs its own data or screens, rules too tangled to keep readable in a visual editor, or a volume that would send a per-task bill climbing.

• The tools: our Zapier, Make and n8n work builds on them, picking one for each workflow by how it bills and where it runs.
• Low-code apps: low-code development on platforms such as Airtable and Bubble suits a process that needs screens and records of its own but not a system written in code.
• Direct connections: API integrations link systems in code, for volumes or rules the tools handle badly.

Our guide to build vs buy software sets out that choice in more depth. When an off-the-shelf tool already does what the process needs, we tell you to buy it.

What should you ask business process automation companies before you hire one?

The answers tell you how the automation will behave on a bad day. Five questions to put to any automation agency, us included:

• What happens when a step fails? Ask to see the alert, the log and the place unclear cases go. If the answer is that it won’t fail, keep looking.
• What will it cost to run for a year? The build is one figure. The tools’ plans, the usage they bill and the monthly fee for upkeep make up the rest.
• Who can change it? A workflow only one person understands is a risk. Ask for a written list of what runs, what triggers each run and who owns it.
• Whose accounts does it run in? Ask which accounts each tool runs under and who pays for them. Then get in writing what you own, what they will run on your behalf and how leaving works.
• Will you tell me when not to automate? A good automation agency turns down processes that won’t repay the build.

What does process automation look like in a system we run?

AlvyOne is an operations system we built with Alvy Logistics and still run with them, and DSI Transportation shares it. Alvy is owned by our co-founder Noah Alvarez, and DSI is an Alvy company. Its order intake reads emailed orders with a model, which makes that part AI automation, covered on our AI automation page, and the AlvyOne case study tells the whole story. Around that reading step, the system runs process automation of the kind on this page:

• Posts to the team’s chat: each job the system builds is posted to the dispatch team’s chat, and a summary of the day’s builds is posted there as well.
• Reminders on held work: a job set aside for a person gets a reminder every hour until someone picks it up.
• A health report: every 15 minutes the system sends a report to our own admin system, and a report that arrives late is flagged.
• Two switches: one stops it posting orders and the other stops it reading the inboxes, each taking effect at once.

None of these steps assigns a driver on its own: that decision stays with the dispatch team.

How is an automated process kept running after launch?

Automations break for reasons outside them. A supplier changes the layout of its order emails, a tool renames a field, an API retires an old version or the business opens a new location. Any of those can stop a step that worked yesterday, so an automation needs someone reading its runs and fixing what changes.

Once a process we automate is live, we look after it on a monthly plan, and what that plan takes on is written into your proposal. When a system or form the process depends on is about to change, telling us beforehand saves a repair afterwards.

The market’s figures for a build and its upkeep at each level are on our custom software cost page. To find out whether a process of yours is worth automating, book a 15-minute call and bring the one you’d most like to stop doing by hand.