Automate or Redesign? How to Decide Before You Build

When a process frustrates your team, the instinct is to automate it. Slow approvals, repetitive data entry, and endless follow-ups all seem like perfect candidates for AI. Sometimes they are. But automating a flawed process doesn't fix it. It simply makes the flaw run faster, at greater scale, and with less visibility.

Before you build anything, you need to answer one question: should this process be automated as it is, or redesigned first?

Why this decision matters

Automation locks a process in place. Once a workflow is built into software, changing it requires time, money, and retraining. If the underlying process was inefficient, you've now invested in preserving that inefficiency.

Redesign, on the other hand, asks why the process works the way it does. Often, the answer reveals steps that exist only because of habit, approvals no one remembers the reason for, or workarounds created to compensate for another problem entirely.

Five questions to ask first

1. Does the process produce the right outcome?

If the end result is frequently wrong, late, or inconsistent, automation won't correct it. Fix the outcome first.

2. Why does each step exist?

Walk through the process step by step and ask what each one accomplishes. Steps without a clear purpose are candidates for removal, not automation.

3. Where does work wait?

Delays often come from handoffs and approvals, not from the work itself. If a quote takes three days but only an hour of actual effort, the problem is waiting, and automation may not be the answer.

4. How often are there exceptions?

Automation handles predictable patterns well. If your team spends most of its time managing exceptions, the process may need to be simplified before it can be automated.

5. Would you design it this way from scratch?

If you were building the process today, knowing what you know now, would it look the same? If not, that's a strong signal to redesign.

When to automate

A process is usually ready for automation when it is:

• Stable: it works well and rarely changes
• Repetitive: the same steps happen frequently
• Rules-based: decisions follow clear, consistent criteria
• High volume: enough instances to justify the investment

When to redesign first

A process usually needs redesign when:

• Outcomes are inconsistent or frequently corrected
• Staff rely on workarounds to get it done
• Multiple approvals slow it down without adding value
• Exceptions are the norm rather than the exception
• Nobody can clearly explain why it works the way it does

An example: the quotation process

Consider a business where preparing a quote takes several days. The obvious solution is to automate quote creation. But a closer look often reveals the real delays: waiting for pricing confirmation from a manager, searching for product information in several places, and reworking quotes because requirements weren't captured properly at the start.

Automating the document itself would save some time. Redesigning the process, by capturing requirements clearly upfront, centralizing pricing, and defining which quotes genuinely need approval, addresses the actual causes. Automation then becomes far more effective, because it's built on a process that works.

Keep people where judgment matters

Redesign is also the moment to decide where human oversight belongs. Not every step should be automated, even in a well-designed process. Decisions that carry financial, legal, or reputational risk benefit from a person reviewing the output before it reaches a client. Good design places AI where it removes effort and people where they add judgment.