Does Your Business Need AI or Ordinary Automation?

A customer enquiry arrives. Someone copies the details into a spreadsheet, forwards it to the right colleague and asks the customer for information they already supplied.

You want that work improved. A proposal promises AI, but you are unsure which part actually needs it.

I would break the task into steps before choosing the technology.

“Does my business need AI or automation?” is a useful buying question. My answer would depend on whether each step involves moving known information, interpreting someone’s wording or making a business decision.

One workflow may contain all three.

Start with the information you already receive

Consider a hypothetical catering business.

Customers can complete a form with fields for event date, guest count, venue and service type. Others send an email saying, “We’re organising lunch for roughly thirty people next month. Could you deliver to our office?”

The form and email create different work for the coordinator.

With the form, the required details are already separated. A workflow could copy them into the existing enquiry spreadsheet and assign the request using agreed rules.

With the email, someone must identify what has been supplied and what remains unclear. “Next month” is not an event date. “Roughly thirty” is an estimate.

I would preserve those distinctions when assessing where AI could help.

Use rules where the next action is already clear

For this first implementation, ordinary automation could handle several steps.

A submitted form could create an enquiry record. A service area selected from a list could determine which coordinator receives it. A missing event date could place the enquiry in a list awaiting details.

Those instructions can be described before the request arrives.

I would also check whether the form, inbox or customer system already includes suitable features. There may be a useful improvement available through configuration or a small connection between tools.

The choice between existing software and custom automation is worth considering before commissioning anything new.

Adding AI would need a reason tied to the task. Moving an agreed field from one system to another would not be that reason on its own.

Consider AI for the wording that varies

The catering emails are a more plausible place to test AI.

A proposed AI step could suggest the guest count, venue and event description from the message. It could prepare a short list of missing information, with the original email beside it.

The coordinator would check the suggestions before accepting them into the enquiry record.

For example, “We might have another ten people” should appear as a possible increase. It should not silently replace the original count with a confirmed larger booking.

Microsoft’s documentation on using prompts in Power Automate shows a concrete way to combine approaches: a prompt can be one action inside a workflow, followed by human review and a condition based on the reviewer’s response.

That supports an implementation with a limited AI step. It does not establish that any particular email will be interpreted correctly. I would test that using examples from the business.

Keep the business decision explicit

Preparing an enquiry record is different from accepting the event.

The coordinator still needs to check availability, delivery arrangements and whether the requested service is something the company can provide. Dietary requests would follow the business’s existing checking process.

A draft reply could ask for the exact date and explain which details are needed next. It would not promise capacity or a final price.

I would also decide where difficult cases go. Conflicting dates, messages discussing several events and revisions to an existing booking would remain visible for individual attention.

This is partly an implementation choice and partly a process question. If staff disagree about what counts as a confirmed enquiry or who can accept a booking, software needs that clarified first.

I discuss that problem in Don’t Build a Tool for a Moving Process.

Compare the work left for your team

When reviewing a proposal, I would ask the implementer to identify the exact step that uses AI and show the alternative.

For the catering business, that alternative might be a clearer enquiry form and an automatic acknowledgement. It might also be a simple manual checklist for emails.

The comparison should include the work staff still do.

If checking an AI suggestion requires rereading every message and retyping the same details, I would investigate whether the proposed design helps enough. If the form creates confusion for customers, I would examine that too.

I would ask for setup costs, continuing software charges and support to be described separately. The choice should reflect the complete working process, including the review effort.

Test one small decision before buying a larger system

An attainable first step is to collect ten representative enquiries, with sensitive details removed for the initial discussion.

Include completed forms, informal emails and a few awkward messages. Write down what a good enquiry record should contain for each one.

Then mark the steps that follow fixed instructions and those that require interpretation.

For the hypothetical caterer, I would first connect the form to the enquiry list. I would test AI on email drafts separately, with a coordinator checking every result.

That gives the business a practical way to decide whether the extra step is worthwhile.

At Novus Automate, I would start with that task and the tools your team already uses. If you want help choosing between a straightforward automation and a small AI implementation, get in touch and show me one recurring piece of work.