Why You Should Redesign Your Processes Before You Automate Them with AI

When a business starts to consider how and where they should implement AI, the instinct is to look at how things currently get done and ask how AI can do it faster. That is the wrong question, and it is the reason a lot of AI investments deliver a fraction of the value they could.

A better question is:

if you were building this process from scratch today, knowing what AI can now do, would you design it the same way?

In most cases, the answer is no.

This is the difference between automating a process and redesigning it. Both have their place, but the AI projects that justify investment usually improve how work happens, rather than just making it faster.

Process Redesign Vs Automation: What Is the Difference?

Automation takes the process you already have and runs it faster, with software or AI doing some of the steps a human used to do. The structure of the work, the order of the steps, and the output all stay the same, with the result simply produced faster.

Process redesign rethinks the work itself. The steps change, the order changes, sometimes the output changes, and the role of the people involved changes alongside it. AI is part of the design from the start, rather than being added onto a process that was built for humans alone.

The distinction matters because AI is not just another way to speed up existing work. It creates new possibilities, making it viable to do things that were previously too expensive, too time-consuming, or too dependent on senior human input. Designing a process around what is now possible is not the same as making the existing process run faster.

What Happens When You Automate a Broken Process

Most business processes evolve over time.

A workaround gets added because a system cannot do something, then someone creates a spreadsheet to bridge a gap between teams, an approval step is introduced after a mistake and, over time, those temporary fixes become permanent parts of the workflow, even after the original problem has disappeared.

When organisations apply AI to these workflows, the instinct is often to automate what is already happening. The result is that AI inherits all the inefficiencies along with the useful work. It processes the spreadsheet, follows the unnecessary approval chain, and performs checks that may no longer have a purpose.

The process runs faster, but it is still the same flawed process.

This is why simply automating a workflow rarely delivers the full value of AI. As Bill Gates famously observed, automation applied to an inefficient operation magnifies the inefficiency. AI is no exception.

The bigger opportunity comes from stepping back and asking which parts of the process still need to exist at all.

• If AI can analyse information in minutes rather than days, does the work need to move through so many stages?
  
  
• If AI can perform the first pass of a review, does every task need senior involvement from the outset?
  
  
• If outputs can be generated at near-zero marginal cost, are there things the business could do today that were previously too expensive to even consider?

These are redesign questions, not automation questions. They focus on what the process should become, rather than how to make the current version run faster.

What Process Redesign Before Automation Looks Like in Practice

The difference is easiest to understand in practice. Here are two processes we redesigned rather than automated, one for a client and one inside our own business.

Redesigning an Outdated Deal Brochure Process

A B2B events business produced a deal brochure ahead of each of its events. Suppliers brought deals that were exclusive to the event, the business collated them into a single brochure, and it went out to members ten days beforehand so they could order in advance. The brochure was commercially important, and producing it was a slow, manual process that had to run with every event.

The mechanics were the problem. A spreadsheet template was emailed to each supplier who filled it in, often in a different format to the one asked for and sent it back. Each return then had to be checked by hand before being passed to a design agency to be laid out, returned as a PDF, and emailed out to members. Every stage depended on a person moving the work along to the next one.

When we mapped it with the client, the obvious first instinct was to put AI on the spreadsheet step, speeding up how quickly suppliers filled it in and how quickly the checking got done. That would have delivered a small improvement while leaving the rest of the process exactly as it was.

Instead, we broke the work into three stages and designed each one from scratch with AI built in from the start.

1. Capture stopped using a spreadsheet emailed to every supplier and became a structured form, prefilled with what the supplier had offered the previous year and validating the format as it was completed.

2. Validation stopped being a person checking each deal line by line and became AI running the comparison, surfacing only the deals that needed a human decision.

3. Output stopped being a design agency laying out a brochure and became AI generating the layout from a template, with a person reviewing rather than building.

The result looked nothing like the process it replaced. None of the original steps survived in the same form, and what ran at the end was a different process rather than a faster version of the old one. The team that had been losing days to the brochure every event got most of that time back, the deals reaching members were better validated, and the design agency cost fell close to zero.

Redesigning Complex Proposal Writing

We ran the same exercise on our own proposal writing at Innovation Visual. The original process had a senior team member writing each custom proposal from scratch, drawing on previous proposals, pricing references, and what they noted from the discovery conversation. It was skilled work, but it was not the highest-value use of a senior person's time.

The automated version of this would have been to give that same person an AI-generated first draft to edit. It would have been quicker without changing anything that mattered. The proposal would still follow the same structure, the senior person would still own the document, and proposal writing would still be a core part of their job.

Instead, we changed who, or what, handles each step. An AI workflow now takes the transcript of the discovery call and writes a draft of the proposal, drawing on a curated set of previous proposals and pricing references. Another person in the team then reviews the draft, focusing on the commercial framing and the specific recommendations where judgement is required, rather than writing the whole document from scratch. Proposal writing has effectively been lifted out of that role, and the role has been rebuilt around higher-value work the person was not previously able to get to.

The outcome was better for the business than a faster version of the old process would have been. Proposals go out faster, the document draws on a wider base of relevant prior work, and the senior team member now spends their time on higher-impact work.

How To Tell Whether You Should Redesign or Just Automate

There is a quick way to check whether you should automate or redesign. Look at the steps in your current process and ask which of them would still exist if you were designing this today. If most of those steps survive, you are automating; if the shape of the process changes, you are redesigning.

The other check is the output. A result that looks like a faster version of what you had means you have automated. A result that looks like something you could not have built before means you have redesigned.

Both have a place. There are plenty of processes where automation is fine and redesign would be overkill, particularly where the process is stable, well-understood, and already runs reasonably well. The projects that drive revenue, though, tend to be the ones where the bigger gains come from redesigning the work entirely. Those are usually the processes that are painful, that have evolved over years, that involve multiple hand-offs, or that depend on expensive senior time for steps that no longer need it.

How To Start Redesigning a Process Before You Automate It

Start with a single process where the current approach is obviously not working and map it from end to end. Capture every step, every hand-off, every system, and every person involved. Identify the steps that exist because of historical constraints rather than genuine present need, then consider what the work would look like if those constraints were removed. That picture is usually your redesigned process.

The mapping itself matters more than most people expect. A documented process and a process as used in practice are often quite different, and the redesign only works when it is based on what actually happens. This means mapping the work alongside the people who do it, rather than the people who believe they understand it.

With the redesigned process defined, the question of where AI belongs becomes far easier to answer. You are no longer adding AI onto a process that was never built for it. You are designing it into a process built around what AI does well.

At that point, you have a process genuinely worth automating.

Find the Processes Worth Redesigning First

The AI Workshop for Leaders is a full day built around exactly this question. We work through where AI fits in your business, where it does not, and where the biggest gains come from redesigning the work rather than speeding it up.

You will spend time mapping the processes where redesign would make the biggest difference, and we will help you find practical ways to start once you are back in the office.

Or if you would rather have a direct conversation about your own processes, you can book some time with me using my meeting link.

Frequently Asked Questions

These are some of the questions that come up most often when leaders start thinking through process redesign and AI together.

Should you automate or redesign a business process first?

For most processes that are obviously painful or that have evolved over years, redesign should come first. Automating an inefficient process simply produces the same poor result faster. For processes that are already well designed, automation on its own can be the right answer.

How do you know if a process needs redesigning before automation?

Three signals usually indicate that redesign should come first:

• The process has evolved over time rather than being deliberately designed.
• It includes workarounds that exist because of constraints that have since gone away.
• The steps would not all exist if you were designing the process today from scratch.

If two or more of these are true, redesign before automating.

Can AI fix a broken process?

No. Applying AI to a broken process tends to make the problems worse, because it runs the same flawed process faster and at greater scale. AI performs well on top of a process that has been designed around it, and poorly as a substitute for fixing one that has not.

What is the difference between process improvement and process redesign?

Process improvement makes small, continuous changes to an existing process to reduce friction. Process redesign rethinks the shape of the work itself, including the steps, the order, the hand-offs, and the roles involved. Improvement keeps the process and makes it better, whereas redesign changes the process itself.

Where should you start with process redesign?

Pick one process that feels obviously broken, and map it end to end with the people who actually do the work. The first redesign project is best chosen for impact and clarity rather than scale. A process that affects revenue or customer experience, that has clear steps, and that is painful for the team is usually a good first candidate.