AI agents or workflow automation: which does your business need?

Every vendor you've spoken to uses both words, and most of them use them to mean whatever they sell. So you're left trying to decide between two things nobody has defined for you, for a process you know better than any of them.

The distinction is simpler than the sales decks make it. It comes down to one question about each step in the process: is the input the same shape every time?

What each one does

Workflow automation

Automation runs a fixed sequence. A form is submitted, so a record is created in the CRM, a confirmation goes to the customer, and a task lands with the account manager. Every input has the same fields in the same places, and every run does the same thing with them. It is fast, cheap to run, easy to test, and it never gets creative. When something arrives that doesn't fit the rule, it stops or it does the wrong thing, and it won't know which.

AI agents

An agent is given a goal and the tools to pursue it, and works out the steps itself. Read this invoice, whatever its layout, and tell me the supplier, the amount and the purchase order it belongs to. Read this email and decide whether it's a complaint, a question or a sales enquiry. Draft a reply using these three documents. The input varies, so the handling has to vary with it. That flexibility is the point, and it is also the risk: an agent can be confidently wrong in a way a rule never is.

A test you can run on any process

Write the process down as a numbered list of steps, the way you'd explain it to a new hire. Then go through the list and ask of each step: could I write the rule for this on one line? "If the total is under the approval limit, approve it" is a rule. "Work out what this customer is actually asking for" is not.

The steps with a one-line rule are automation. The steps without one are where an agent earns its place. In most business processes the split is lopsided: eight or nine steps out of ten are rules, and one or two need reading or judgment. That's the shape to build, with the agent handling only its steps and the automation carrying the rest.

If every step has a rule, you don't need an agent at all, and adding one makes the system slower, dearer and harder to trust. If no step has a rule, the process probably isn't ready to automate yet, because nobody can say what "done correctly" looks like.

Where each one fits in a growing business

The same handful of processes come up in the businesses we talk to, and they split cleanly.

• Lead intake. Capturing the form and creating the record is automation. Working out whether a free-text enquiry fits your criteria and drafting the first reply is an agent step. Both belong in Revenue Growth.
• Invoice handling. Routing, matching to the order number and queuing for approval are rules. Reading the invoice itself, in whatever layout the supplier used, is an agent step. See Operational Efficiency.
• Support triage. Pulling the order from your system and sending the reply are automation. Classifying the message and drafting the answer from your approved information are agent steps. See Customer Experience.
• The weekly report. Almost entirely automation. The numbers come from fixed places and the layout doesn't change. An agent adds little here beyond a written summary, and that summary needs checking.

One enquiry, step by step

Take a new enquiry from your website form. Here is how it splits. The form arrives and a record is created in the CRM: a rule. The company name is checked against the records you already hold, so a repeat customer isn't treated as new: a rule. The free-text message is read to work out what they want, how big the job is and whether it fits the criteria your sales team set: an agent step, because the message could say anything. A first reply is drafted from that reading and your approved wording: an agent step. The draft and the record land with the right account executive, chosen by region or size: a rule. They read the draft, change a line, and send it: a person.

Six steps, two of them agent steps, one of them a person. The automation carries the enquiry from the form to the rep's screen without anyone touching it. The agent reads and drafts. The rep decides what goes out. If the reading is wrong, the rep sees it before the customer does, and the system learns from the edit. That shape, with the agent inside the automation and the person at the end, is what nearly every good first system looks like.

Where the person goes

An agent's step is the one most likely to be wrong, so it is the one that most needs a check. The rule we hold to is simple: anything an agent produces that could reach a customer, a supplier or an account lands in front of a person before it does. The person sees the draft, the source it was built from and the confidence the agent had in it, and approves or edits in a few seconds. Routine runs still take no time from them. The exceptions get their attention.

This is also what makes the system trustworthy enough to extend. Once a team has watched the agent's drafts for a while and seen how often they need a change, they can decide for themselves which categories could go out unchecked. That decision stays with them, not with the vendor. We've written more on this in how to keep a person in charge of an automated process.

Questions that cut through the pitch

When someone proposes a system, three questions tell you quickly whether they've thought about your process or just their product.

1. Which steps are rules, and which steps does the agent handle? If they can't split the process, they haven't looked at it.
2. What happens when the agent is wrong? A good answer names the person who sees it, what they see, and how the system learns from the correction.
3. Why does this step need an agent rather than a rule? If the only answer is that the agent is more capable, the rule would have been cheaper and safer.

What to do this week

1. Pick the process you'd most like off your plate and write it as a numbered list of steps.
2. Mark each step as a rule or a judgment. Count both.
3. If the judgments are one or two, you have a system worth building: automation with an agent inside.
4. Write down who would check the agent's output, and what they'd need to see to check it in under a minute.

If you'd like someone to look at the list with you, that's what the first call with us is for. Bring the steps and we'll tell you which ones are rules, which ones need an agent, and whether it's worth building yet.