Before You Add AI Tools, Understand Your Workflow

Why do you need that AI productivity tool?

A friend I’ll call George recently told me he had 100+ open to-dos. Yep, he counted them all. Many business owners in our group were recommending he add an AI productivity tool to his tool belt, specifically Grok.

I understood the temptation. When you already have 100+ things staring at you, and someone you respect tells you about an AI productivity tool that could save time and help you get more done, of course you wonder whether you should be using it too.

That pressure isn’t unique to George. Microsoft’s 2026 Work Trend Index found that 65% of AI users surveyed fear falling behind if they don’t use AI to adapt quickly.¹

But I kept coming back to a different question.

What are those 100+ things?

Are they 100+ actions George needs to complete? Are some reminders? Recurring tasks? Ideas he wanted to capture? Decisions he hasn’t made yet? Things someone else could do? Things that seemed important six months ago, but no longer matter?

Before we ask how AI can help George get more done, perhaps we need to understand the work he is trying to do.

A Long Task List Doesn’t Tell the Whole Story

I could tell you I have hundreds of tasks in Todoist, and you might reasonably conclude that I have a productivity problem.

But that’s not what those tasks represent.

I use Todoist as a tool to store recurring tasks, reminders, and ideas so I don’t have to keep them all in my head. Some recur monthly, some prompt me to make payments, and others are ideas to revisit later. They aren’t all actions that require my attention today.²

I use another tool, Sunsama, differently. That’s where I decide what deserves my attention today, what can move to tomorrow, and sometimes what doesn’t need to be done at all.³

A task list tells me what I’ve captured. It doesn’t automatically tell me what matters.

That distinction is important when we start talking about AI productivity tools. If I don’t understand the problem I’m trying to solve, adding another tool—even a remarkably powerful one—may simply give me another tool to manage.

I’ve Learned This Lesson Before

Long before we were talking about AI assistants, I spent a lot of time thinking about how people work with human assistants.

Years ago, I wrote about building a successful working relationship with a virtual assistant. My premise was that you don’t hire someone on Monday and hand over complete decision-making authority on Tuesday. You begin with clear expectations. You learn how the person works. You evaluate the quality of their research and recommendations. As trust grows, you can increase their authority.

That lesson wasn’t theoretical.

I once had a client who hired me to help with his webinars. I was responsible for getting the work done, but I didn’t have the authority I needed to make decisions and keep things moving.

So I kept going back to him with questions.

He became frustrated. I became frustrated. Eventually, I asked him what value I was bringing to his organization.

“None.”

Well.

That was a memorable performance review.

I concluded that we were not a good fit, and we ended our working relationship. Interestingly, we remain friends to this day. He loved the idea of an assistant but wasn’t ready to let go of his systems.

I can laugh about the exchange now, but there was a valuable operations lesson underneath it. He wanted me to reduce his workload, but I lacked both the authority and the information needed to move the work forward. It wasn’t that I was unwilling to make decisions; I simply couldn’t make sound ones based on the limited information I had received.

More effort on my part wasn’t going to fix the problem.

The way the work had been structured was the problem.

AI Productivity Tools Don’t Replace Operational Clarity

I’ve been thinking about that experience as I use AI more in my own work.

Recently, I worked on a policy for an international nonprofit. To evaluate it from several organizational and governance perspectives, I drew on the existing documents, organizational structure, bylaws, and constitution, as well as relevant regulatory requirements.

A Custom GPT helped me work through that information, but I didn’t simply type, “Write me a policy,” accept the answer, and call it a day.

For the tool to be useful, I first had to decide what information to provide and how to sequence it. The questions mattered. So did the lens I asked it to use when processing the information. Along the way, I reviewed the results, made changes and redirected the work.

We ended at Draft 5.

And that draft still goes to human beings—the board—for review.

AI was useful because I understood the work well enough to supervise the tool.

Notice I said supervise, not trust.

To be honest, I don’t trust AI to get everything right. I’ve used it enough to know that small errors creep in. Sometimes it misunderstands what I’m trying to accomplish. Sometimes it gives me something technically responsive but misaligned with what I need. Sometimes its attempt to be helpful simply isn’t helpful.

That’s not just my experience. The National Institute of Standards and Technology warns that generative AI can confidently produce false or inconsistent information, including fabricated logic and citations. Those risks become especially important when the work requires significant context, domain expertise or consequential decisions.⁴

So perhaps the question isn’t simply, 

Can AI do this?

A better question may be:

Do I understand this work well enough to supervise the tool—and recognize when it gets something wrong?

Every Tool Requires Something from Us

AI can be fast. It can sort information, identify patterns, help us think, generate possibilities and handle repetitive work.

I’ve experienced that value myself.

I’ve also experienced the opposite: spending time correcting the tool, redirecting it, repeating context, and troubleshooting instructions, only to realize that the thing meant to reduce friction had become the source of it. In one case, it filled in missing information instead of telling me it couldn’t read a document. Well, duh! Tell me that.

Time saved isn’t the only measure of whether a tool is helping. I also want to know what it costs me in attention, energy and mental capacity. Everyone’s capacity is different, but for me, if I walk away frustrated after three iterations, it is not bringing me value.

Verification is part of that cost. Check, double-check and even triple-check. Even OpenAI advises users to verify important information and to check citations and source links rather than relying on a response simply because it sounds authoritative.⁵

A tool I use for Olinda Services may not be appropriate for a client project because of privacy or confidentiality requirements. When the work involves confidential information, organizational decisions or other people’s information, I have additional questions. What am I authorized to share? How is the information handled? What belongs in this system and what doesn’t? Where does human judgment need to remain firmly in place?

Those aren’t reasons to avoid AI.

They’re reasons to use discernment.

Start With the Work, Not the Tool

This is where I come back to George and his 100+ to-dos.

Maybe Grok would be enormously helpful to him. Maybe another AI tool would be. Maybe what he really needs is a better process, some automation, human support, or an afternoon spent deciding which of those 100+ things no longer deserves space on his list.

I don’t know.

And that’s the point.

Earlier, I shared how I use Sunsama. It’s useful to me, but that doesn’t mean it’s right for you. Even advice from people we respect must pass through the filter of our own work, responsibilities, capacity and season of life.

Before you add AI—or any other tool—to your toolkit, perhaps the better questions are:

• What problem am I trying to solve, and what outcome do I need?
• What part of this work requires my judgment?
• What could reasonably be delegated or automated?
• Do I understand the work well enough to supervise the help and recognize when something isn’t right?
• Will this tool create capacity—or consume it?

Remind Yourself

You don’t have to adopt every useful tool.

And you don’t have to get everything done simply because you once put it on a list.

Sometimes the most productive thing you can do is understand the work in front of you, decide what matters, and then choose the right kind of help.

That’s not resistance to technology.

That’s alignment.

Notes

1. Microsoft, “Agents, Human Agency, and the Opportunity for Every Organization,” 2026 Work Trend Index Annual Report, May 5, 2026. The report surveyed 20,000 workers using AI across 10 countries and found that 65% feared falling behind if they did not use AI to adapt quickly. Read the report.
2. Todoist is the tool I use to capture recurring tasks, reminders, and ideas.
3. I use Sunsama to plan my day. If you’d like to explore it, my referral link includes a free month. I receive account credit if you subscribe.
4. Chloe Autio et al., Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1 (Gaithersburg, MD: National Institute of Standards and Technology, 2024), sec. 2.2. NIST describes “confabulation” as confidently presented erroneous or false content and notes heightened concerns in contexts requiring domain expertise or consequential decisions. Read the report.
5. OpenAI, “Does ChatGPT Tell the Truth?,” OpenAI Help Center, accessed October 16, 2026. OpenAI advises users to verify important information from reliable sources and specifically recommends opening cited sources and checking details before relying on responses. Read the guidance.