← All insights

Automation

Three processes worth automating first, and two that aren’t

By Hazel River Digital

November 3, 2026

3 min read

A priority grid of blue squares with one large orange square in the top-left quadrant marking the first process to automate.

Interest in automation and AI has led many businesses to ask what they should automate before asking whether a given process is suitable. The result is often an impressive demonstration of a workflow that nobody uses, or an automation that silently handles a task that should have involved a person.

Selecting well largely comes down to three questions about the process itself, and you can answer them before choosing any tool.

Three tests for a good candidate

Volume. The process happens often enough that the time saved, and the errors avoided, outweigh the effort to build and maintain the automation. A task performed twice a year rarely qualifies.

Rules. The steps can be written down, and two competent people would carry them out the same way. If the answer depends on judgment that has never been articulated, the first task is to articulate it, and automation comes later.

Risk. An error is inexpensive and easy to correct, or a person reviews the result before it has consequences. The higher the cost of a mistake, the more a human check belongs inside the process.

A process that passes all three is a strong candidate. One that fails the rules test is not ready, and one that fails the risk test needs a different design.

Three processes worth automating first

1. Inquiry intake and routing. When a form is submitted, the details are recorded in the CRM, the inquiry is assigned to the appropriate person, a task is created, and the sender receives an acknowledgment. The volume is regular, the rules are simple, and the cost of an error is a misrouted message that a person can correct. The benefit is also easy to see: no inquiry waits because the person who handles it is away from their inbox.

2. Hand-offs between systems. In many businesses, the same information is typed into several places: a won deal becomes a project, a project becomes an invoice, and a customer record is updated in more than one tool. Each manual re-entry is a chance for error and a delay. Moving the data between a CRM, a project tool, and accounting software, with a person approving anything involving money, removes the repetition without removing oversight.

3. Routine internal reporting and reminders. Weekly summaries of open inquiries, overdue tasks, or upcoming renewals are assembled from existing data. They are predictable, follow fixed rules, and surface errors quickly because someone reads the report. They also build confidence in automation without risking anything important.

Two that are not good starting points

1. A process whose rules are unclear or still changing. Automating an inconsistent process makes the inconsistency faster and harder to see. If the people who perform a task disagree about how it is done, they need to agree on the method and document it. Automation can follow once the process is stable.

2. Decisions with significant consequences for a customer or the business. Pricing exceptions, refunds, disputes, credit decisions, and hiring are cases in which accountability matters and context is hard to capture. AI can usefully prepare the material, for example by summarizing a case history or drafting a response, but a person should make and own the decision.

Count the cost before you build

Automation platforms measure usage differently, and the same workflow can cost very different amounts depending on how it is built. Zapier counts a task when an action completes successfully and does not count triggers. Make counts credits, with most actions consuming one, so a workflow with many steps consumes many. n8n's cloud plans count an execution as a single run of an entire workflow regardless of how many steps it contains. A workflow that runs thousands of times a month with several steps each should be priced against the platform before it is built. We discuss the choice in more detail in our article on selecting a connector.

The short version

Choose processes that are frequent, rule-based, and low-consequence, and keep a person responsible for decisions that carry significant risk. Start with intake and routing, handoffs between systems, and routine reporting, and resist automating anything nobody can yet describe. If you would like our team to review your processes and identify the strongest candidates, reach out—we'll assess them transparently before any commitment.

Have a system you want your website to talk to?

Book a free discovery call. We will tell you where your applications stand before you commit to anything.

Book a discovery call