Which Business Processes Should You Automate First?

A purchase request lands in a manager’s inbox on Monday. She is traveling, so it sits. On Wednesday it gets buried under forty newer emails. By Friday, the person who asked is not sure whose turn it is, so finance steps in and sends a reminder. 

The request was approved the following Monday, six days after it started. The supplier’s price has moved in the meantime.

Nobody in that story was lazy or careless. The request simply had no way to move by itself. Everyone was waiting for a human to remember.

That gap is what workflow automation closes. It does not replace anyone’s judgment. It makes work that follows a predictable route actually travels that route, with reminders, handoffs, and escalations that happen on time. 

The hard part is not the tooling. It is choosing what to automate first, and that is what this piece is about.

Workflow Automation Moves Work, Not Just Data

Workflow automation sits one layer above in your data. Instead of moving a customer record from one tool to another, it moves a task from one person to the next. An approval. A review. A handoff. A follow-up.

You can have one without the other, but they work best together. A purchase request that routes itself is still slow if the approver has to open three systems to see what the request is for.

Skip the Scary Number

Almost every article on this topic opens with the same statistic. Asana’s Anatomy of Work research says knowledge workers spend about 60 percent of their time on what it calls work about work. 

That means chasing updates, searching for information, switching between tools, and sitting in status meetings. It comes from a worker survey, so it reflects how people feel their time is spent. It is still a useful signal.

It is a poor guide to what to automate, though. Sixty percent spread across hundreds of small, varied interruptions is not an automation opportunity. It is just what a job looks like. 

One automation consultancy put the useful version well. The hours worth recovering are the ones that clump: the same sequence, on the same trigger, many times a month. Those are the ones to hunt for.

Four Questions That Find Your First Candidate

Run any process you are considering through these four questions. The best first candidates answer yes to all of them. Do not automate anything you expect to change in the next year or two.

1. Does it happen often, the same way? Same trigger, many times a month. A task that happens twice a year is not worth the setup.
2. Can you write the rules on one page? If an approval depends on the amount, department, and a few thresholds, a workflow can follow it. If the honest answer is “it depends, ask Sam,” a person needs to stay in the loop.
3. Has it stayed the same for six months? Automating a process that keeps changing means rebuilding the automation every time. Let the process settle first.
4. What does a delay or a mistake cost? This is the tiebreaker. A late approval that lets a price move, or a missed follow-up that loses a lead, is worth far more than a slow internal reminder.

Here is how some common candidates tend to score. Treat it as a starting point rather than a ruling. Your own business may score differently.

Notice the monthly report row. If assembling that report takes a day because the numbers live in three systems, the real problem is a connection problem, not a workflow problem. We covered that fix in how to automate manual data entry between business systems, and it usually comes first.

Do You Have to Fix the Process First?

In that same piece, we said not to automate a broken process. That still holds, but it needs one refinement. You do not need a perfect process. You need an agreed one.

UiPath, which sells automation software and so has a stake in the answer, argues against waiting for one golden process. Pick the path most work already takes, automate that, and send the unusual cases to a person. We think that is right. The goal is one clear main route, not a flawless one.

Where AI Fits, and Where It Does Not

Rules first. If a process can be written as if-this-then-that, a plain rules-based workflow is cheaper and more predictable than anything with AI in it.

AI earns its place where the input is messy. Reading an invoice that arrives as a scanned PDF. Sorting free-text support requests into categories. Pulling a date and an amount out of an email. In those cases, AI does the reading and a rules-based workflow does the routing.

Keep a person in charge of anything that moves money or reaches a customer. AI can be confidently wrong, which we explained in why AI chatbots make things up. Let it triage and draft. Let a person approve.

Give It an Owner Before You Give It a Tool

Automations rarely fail because the software is bad. They fail because nobody owns them. The approval rules change. The person who built the workflow leaves. A quiet failure goes unnoticed for weeks. Name one person who owns each workflow, knows what it should do, and gets the alert when it does not.

Why This Comes Last

We think of this as the fourth step in what we call the Digital Progression Framework. Stabilize your systems. Connect them. Run them well. Then orchestrate the work across them. 

Workflow automation sits last for a reason. It rides on top of your systems, so if they are fragile or disconnected, the automation inherits every problem underneath it. If your approvals depend on a system nobody dares touch, start earlier in the sequence.

Now think of the last thing that sat in someone’s inbox for a week. A purchase request, a new-hire checklist, a lead nobody followed up. That is probably your first workflow.

We sell this kind of work, so weigh our advice with that in mind. The four questions above cost nothing, and they will get you most of the way on their own.

Not sure which process clears all four questions? 

Bring us your shortlist. We will tell you which one to start with and which ones to leave alone. 

The Author

LN Webworks