Why Vashti Patrick-Joseph Says the Process Comes Before the AI

Before adding another AI tool, look at the process underneath it. Vashti Patrick-Joseph shows how to identify where automation can save time, where a system needs work first, and when human judgment still matters.

There is always another AI tool promising to save time, automate a task, or make running a business easier. When something already feels harder than it needs to be, finding the right technology can seem like the obvious place to start. During her Bloom Connections Educational Moment, Vashti Patrick-Joseph challenged us to begin somewhere else. Before choosing the AI, she wants us to understand the problem we are asking it to solve. As Vashti told the group, “Instead of asking, what tool should I use, the first question should be, what problem am I trying to solve?” That distinction matters because a new capability is not automatically a business need. A tool may be able to automate something, but that does not mean the task is creating enough friction to deserve automation or that the process underneath it is ready. Vashti’s approach starts with understanding the work itself. Once you know what is happening and why it needs to improve, you are in a much better position to decide whether AI belongs in the solution.

You Can't Automate What You Haven't Defined

Vashti calls this her golden rule: “You can’t automate what you haven’t defined.”

It sounds straightforward, but think about how many processes inside a business have never actually been documented. One person knows how to do the task because they have done it for years. Someone else handles it a little differently. Part of the information lives in an inbox, another part is stored somewhere else, and the rest has simply become something the team knows.

The work gets done, so it feels like there is a process.

The problem appears when you ask technology to follow it.

If the steps are inconsistent or important decisions haven’t been defined, AI does not automatically fix what is unclear. Vashti described the result as a “faster, automated, prettier mess.”

Before automation can improve the work, the business needs to understand how that work is supposed to happen. That includes knowing what starts the process, which steps follow, what information is needed along the way, where decisions are made, and what happens when a situation falls outside the usual path.

Once those pieces are visible, something else often becomes clear: you may not need to automate everything.

You may only need to improve one part of the process.

Vashti’s customer service example showed us what that looks like in practice.

AI Needs the Context Your Team Already Knows

Imagine an online store receives an email from a customer and asks AI to draft the response. At first, that sounds like a simple use of automation. The AI reads the message, writes a reply, and someone gets time back. Except the right response depends on why the customer is writing. A customer asking about a return needs different information from someone whose shipment is delayed. A sizing question requires another response altogether. The AI also needs to know the company’s policies so it does not promise something the business would never actually offer. Then there is the way the company communicates. If customers are used to a warm or playful brand voice, an automated response that suddenly sounds stiff and corporate creates a different experience. A team member who has worked inside the company for a while may carry all of that context naturally. The technology does not. This is where documenting the process becomes useful. The business can establish how common situations are handled, provide the information the AI needs, and decide how a response should sound. Defining those expectations also forces the business to decide where the automation ends, because not every situation belongs in the same workflow.

Some Parts of the Process Still Need a Person

A standard customer question might move through an automated process without an issue. If the customer is angry, however, the same response may no longer be appropriate. The business might also decide that orders above a certain value need to be handled personally rather than automatically. Those become rules within the process. This is an important part of the automation conversation because efficiency is not only about how quickly something gets done. The outcome still has to work for the business and the person on the other side of it. An automated response can be fast and technically correct while still being the wrong response for that moment. Knowing where a person needs to step in gives the automation boundaries. Once those boundaries are established, the next question becomes whether the task is worth automating at all.

Vashti Uses Three Questions to Evaluate an Automation Opportunity

Rather than looking at everything a new AI tool can do, Vashti gave us three questions for evaluating the work already happening inside the business.

She asks whether the task is repeatable, defined, and valuable.

Is the task repeatable?

Work that happens regularly gives you something to examine. If the same task occurs every day, every week, or several times throughout the month, you can begin to see the pattern behind it.

You can identify which steps stay consistent, where someone is repeatedly spending time, and whether part of that effort could reasonably be reduced.

That is different from trying to automate a task that changes dramatically every time it happens.

Is the process defined?

This brings us back to Vashti’s golden rule. If the person doing the work cannot clearly explain the steps, AI does not yet have a reliable process to follow.

Documentation turns the knowledge someone may be carrying in their head into instructions that can actually be used. It also exposes gaps that may have gone unnoticed because an experienced person has learned how to work around them.

Once the process is defined, you can make a more informed decision about where technology could support it.

Would automating it be valuable?

This was the question that brought the conversation back to the business case.

Vashti gave the example of paying around $60 per month for a tool to automate something that a person can already accomplish with a click that takes about two minutes.

The fact that the task can be automated does not mean the business gains enough from doing it.

The time saved may be insignificant. The software becomes another expense. Someone still needs to manage the tool. The business may have introduced more technology without meaningfully improving anything.

Looking at repeatability, definition, and value together gives you a much stronger reason for choosing where AI belongs.

It also opens the door to a more practical way of thinking about automation.

You May Only Need AI for One Part of the Work

One thing I appreciated about Vashti’s Educational Moment was that she did not frame AI as an all-or-nothing decision. The entire workflow does not have to be automated for the technology to be useful. Vashti encouraged us to notice work that repeatedly takes too much time or makes us think there has to be an easier way to do this. Once the current process is documented, you can identify where the friction actually occurs and determine whether AI could reduce it. That might mean using AI to brainstorm, summarize information, create a first draft, organize ideas, or repurpose something that already exists. Vashti used meeting notes as one example. When you are trying to participate in a conversation while also capturing everything being said, part of your attention is always divided. An AI note taker can handle the documentation so the people in the meeting can remain present in the conversation. She also described using voice features when she has ideas but has not organized them yet. Instead of forcing herself to immediately turn those thoughts into polished writing, she can talk through them and use AI to help organize what she said into something she can continue developing. In both examples, AI supports the work without becoming the work. That distinction becomes even more important when we consider what the technology cannot bring to the process.

AI Can Support Your Expertise Without Replacing It

Vashti was clear that using AI does not remove the value of the person behind the business.

The technology does not have the years of experience that inform how you recognize a problem. It does not have the relationships you have built with customers. It does not bring your judgment to an unusual situation or understand why you might make an exception when the standard process says otherwise.

As Vashti reminded the group, “If you’re selling, you’re not selling to machines, you’re selling to human beings.”

That is an important consideration when deciding what to automate.

Saving time matters, but the experience created by the process matters too. There are moments when a customer needs judgment rather than a generated response. There are decisions where experience changes what happens next. There are conversations where the relationship itself is part of the value.

The goal is not to keep a person involved in every task simply because that is how it has always been done. It is to understand why that person is involved and whether their contribution is necessary at that point in the process.

Once you understand that, AI can take repetitive work off someone’s plate without removing the part that makes the work valuable.

And in some cases, looking this closely at the workflow reveals that AI was never the first problem.

Sometimes the Process Is What Needs Attention

When something inside a business feels frustrating, our instinct is often to look for something to add.

A new platform might fix it. A better tool could make it easier. An automation could finally take it off someone’s plate.

Mapping the process can reveal a different problem.

Perhaps the steps are not clear. Maybe two people are doing the same task differently. Important information could be scattered across several systems. There may be unnecessary steps that accumulated over time, or no one has decided what happens when the usual process does not apply.

In those situations, adding AI does not address the source of the friction.

The business first needs to decide how the work is supposed to function.

That is why Vashti’s order matters. Her approach is to define the problem, put it into a process, and then consider where AI belongs.

As she told us, “AI doesn’t replace good systems, it makes good systems more powerful.”

That puts technology in a very different role.

Instead of asking AI to create order, you create the order first and then decide whether technology can make it work better.

The Better AI Decision Starts With the Work

The next time a new AI platform promises to automate hours of work, the first decision does not have to be whether you need the tool. Look at the work you want to improve. Understand how it happens today and where it begins to feel heavier than it needs to be. Determine whether the task occurs often enough for automation to make a difference and whether the process is defined well enough for technology to follow it. Then consider what you would actually gain by changing it. You may find a strong opportunity for AI. You may discover that only one small part of the workflow needs support. You may realize that human judgment needs to remain at a particular point. Or you may uncover a process that needs to be worked out before automation enters the conversation at all. That is what I took from Vashti’s Educational Moment. The question is not simply what can AI automate? A more useful question is what does this business need to work better, and where can AI genuinely help us get there? When you start there, the technology has a reason to be part of the process.

About Vashti Patrick-Joseph

Vashti Patrick-Joseph smiling in a blue striped shirt against a cream background

Vashti Patrick-Joseph helps growing organizations and businesses simplify their operations and adopt practical AI where it makes sense, giving teams more time to focus on work that matters. Her work includes individual projects, ongoing operational support, AI workshops, and training. Her background includes leading operations for a CPG beauty brand sold at retail scale across the U.S., U.K., and Canada, along with experience in process improvement and systems implementation.

Frequently Asked Questions About AI Automation for Business

What should a business do before automating a process?

Before automating a process, identify the problem you are trying to solve and document how the work currently gets done. Understanding the steps, information, decisions, and exceptions involved makes it easier to see where AI could improve the workflow and where a person still needs to be involved.

How do you know whether a task is worth automating with AI?

Vashti Patrick-Joseph recommends evaluating whether the task is repeatable, whether the process is defined, and whether automating it would create meaningful value. Those three considerations help distinguish a useful automation opportunity from something that is merely possible to automate.

Can AI fix an inefficient business process?

AI can make parts of a workflow faster, but it does not automatically resolve an unclear or poorly defined process. Vashti recommends working out how the process needs to function before adding automation so the technology is supporting something that already makes sense.

What kinds of business tasks can AI help with?

Vashti discussed practical uses such as brainstorming, summarizing, drafting, repurposing content, organizing information, and capturing meeting notes. The right use depends on the problem the business is trying to solve rather than simply what a particular AI tool is capable of doing.

When does AI automation still need human oversight?

Human involvement remains important when a situation requires judgment, empathy, an exception to the usual process, or a personal response. Vashti recommends defining those boundaries so an automated workflow knows when something needs to be handed back to a person.

Learn Something. Share Something. Grow Together.

Vashti’s Educational Moment reminded us that adopting AI does not have to begin with finding another tool. It can begin by looking more closely at how the work gets done. When the process makes sense first, you can make a much better decision about where technology belongs. Want to experience the next Educational Moment? Join us at an upcoming Bloom Connections gathering.

Share this article :