Fix the Process, and the Culture Follows

Field Notes is our ongoing look at what's actually happening in AI and what it means for small businesses in Spring, TX and across Greater Houston. No hype, no noise, just the signal.

Imagine three people in your business preparing the same kind of quote. One keeps the current prices in a spreadsheet. Another works from an older copy with a few corrections they remember. The third asks you whenever something looks unusual.

You could give all three an AI tool to help write quotes. Before you do, though, somebody needs to decide which prices are right and what happens when a request falls outside the usual work.

I help small businesses figure out where AI fits, and this is the kind of question I want answered before we talk about tools. What does the work look like today? Where does it get held up? Who knows how to get it moving again?

Writing that down is process mapping before automation at its simplest, and it gives you something useful whether you add AI or not. You can see the process, and you can start to see how the team works together inside it.

Why better AI tools don't always mean a better workflow

In "You're Probably Using AI Wrong. Do This Instead," Keith Ferrazzi talks with Peter Diamandis about three ways companies approach AI:

• Giving people tools
• Changing the business model
• Redesigning workflows

Ferrazzi argues, and I'd have to agree, that companies are spending too much attention on the first and too little on the work itself.

Those three people now generate quotes faster. But the quotes may sit two days waiting for approval, or carry wrong prices because everyone has their own version.

The tool helped with a task, but the rest of the process still needs work.

Before even thinking about anything AI related, take some time and map out the process. Document each step from start to finish. Start with one page about one recurring job. Write down what starts it, who handles it, what they need, and what counts as finished. Include the parts people normally explain with "then I check with somebody" or "it depends."

Process mapping before automation can be that plain and simple. You don't need a special diagramming tool to find out that two people are waiting on each other. You just need to be thorough and specific. It can be tedious, but the tedium now helps avoid headaches later.

Writing down the work makes teamwork visible

Many family-run businesses have processes that were created and implemented by staff members who are long gone using tools that have long since been replaced, yet the process remains the same. Writing down the work helps to surface bottlenecks, workarounds, and ad-hoc fixes your team may have had to invent. Giving your team an opportunity to have input in how to make things work better not only gives them a sense of ownership in the process but it also gives you more insight into the level of care your team has in what they do and how they work together to solve problems.

Consider a handoff where one person believes their job ends when they send an email. The next person believes the job hasn't reached them until a form is complete. Both can be doing what they understand their job to be, while the customer waits.

Once you write down that handoff together, you can agree on what has to arrive, who owns the next step, and how to flag missing information. You have a specific problem to work on instead of another conversation about communicating better.

You may also notice someone who quietly keeps work moving. They catch missing details, help coworkers find answers, or remember the follow-up nobody assigned. Those kinds of contributions deserve to be visible as well as recognized, and the process shouldn't depend on them being available to make sure the process executes successfully.

This is part of why a full workflow engagement interests me beyond the automation. Understanding how a business operates also means understanding how its people work together. A written process gives the team a chance to explain what helps them do good work and what gets in their way.

As you capture each step in a process, encourage your team to be open about what they do, why they do it, and how they think it could be done better. Assure everyone that nobody is on trial. The point is to find what doesn't work so those issues can be fixed and everyone's experience and efficiency improves.

That is how fixing the process ends up fixing the culture. The team can finally see the same work, agree on who owns what, and solve specific problems instead of trading frustration.

One operator's example: fix the information before drafting the quote

In a separate ChiefAIOfficer conversation with Justin Watt, Watt describes a logistics company's quoting process. According to him, this company's account managers were working from different rate sheets and different ways of preparing quotes.

Before adding AI, his team brought the pricing information together. AI then helped prepare a first draft, with people still reviewing it.

Watt says the eight account managers went from spending about 160 hours a week on quotes to roughly 24 hours a week. He also says the team wasn't reduced and had more time for customer relationships and closing work.

Those are Watt's reported results, not an independently verified benchmark or a promise about your business. The useful part is the order: agree on the information, sort out the process, then use AI for a defined part of the work.

The host offers a practical way to get started: have the person who owns the process record themselves explaining what they do. Keep asking what happens next, where the file goes, and what they do when something is missing.

If staring at a blank document slows you down, talking through the work is a perfectly good starting point. Get the person's agreement to record, leave customer details out where possible, and check what you can share before putting the recording or transcript into an AI tool.

How to document a workflow before adding AI

For small business process improvement, I'd rather see one job understood and tested than a folder full of plans nobody uses. Work through these steps with the people who actually do the job.

1. Choose one recurring workflow. Pick a job that comes up often and takes more time or chasing than it should. Give it a clear start and finish. "A customer asks for a quote, and we finish with an approved quote ready to send" is something you can work through. "Improve sales" is too broad. If you need help choosing, What Should a Small Business Automate First? covers that first decision.

2. Write down what happens today. Have the person doing the work talk through a recent, ordinary example. Note each step, the information used, where it's stored, who handles it, and who receives it next. Ask about missing information and unusual requests too. Workflow documentation should describe the job as it happens, including the workarounds, before you decide how it ought to work.

3. Measure the work and the waiting. Record how much hands-on time the job takes, how long it takes from start to finish, and how often it needs correction. Those are different things. A quote might take twenty minutes to prepare and two days to get approved. Saving five minutes of typing won't fix the approval delay. Use your own numbers rather than another company's advertised savings.

4. Fix unclear steps before automating them. Agree on the current information, remove duplicate entry where you can, and make each handoff clear. Then ask which remaining steps need ordinary automation and which need AI. Moving a completed form into the right folder may need a simple rule. Turning approved notes into a readable draft may be a better job for AI. AI workflow redesign doesn't mean putting AI into every step.

5. Give "the intern" a defined job and a supervisor. Think of AI as a smart, eager intern. Show it the approved information, what a good result looks like, and when to stop and ask. Keep human approval for decisions involving money, customer commitments, safety, or reputation. Name the person who reviews the output and the person who can pause the workflow when something changes.

6. Run a contained test. Choose a fixed test period, such as two weeks or a month, and keep the old way available while you learn. Check normal requests and awkward ones. Missing information, an outdated price, and an unusual customer request should have somewhere to go other than an AI guess. Include the time spent checking and correcting the output in your results.

7. Compare the results with the old process. Look at hands-on time, total turnaround, corrections, and the team's experience. Ask whether the work got easier or whether you've moved the effort to somebody else's desk. Count software and upkeep costs too. More AI use isn't the goal; a better result with a manageable amount of supervision is.

8. Expand only when the improvement holds. Keep the instructions current and review them when prices, responsibilities, or customer needs change. Add the next task when the first one works reliably and the team understands it. If a simpler process solves the problem without AI, keep the simpler process.

The same principle runs through What a Small Business Should Never Fully Automate: make the follow-through reliable while people keep responsibility for the decisions and relationships.

A warning before you buy

One word of advice before you shop. If someone tries to sell you AI to fix a problem without taking the time to understand how the work gets done today, what they're selling is the tool, not the fix. Either the sale matters more than your result, or they haven't been around the block enough times to know better. Either way, get a second set of eyes before you sign anything. My 15-minute Strategy Call is free, and I'm happy to look at what you've been offered and tell you what I'd ask first.

You don't have to turn your whole business into an AI project. One clearer workflow can give your team fewer things to chase, make their contributions easier to see, and help you decide where an intern would actually be useful.

Sources

• Keith Ferrazzi and Peter Diamandis: "You're Probably Using AI Wrong. Do This Instead". Source for the three approaches to AI.
• ChiefAIOfficer: "How AI Workflow Automation Changes the Way Your Business Works with Justin Watt". Source for Watt's reported quoting example and the host's process-recording suggestion. The results are the speaker's account, not independently verified outcomes.