It takes five things to make AI work at work: preparation, practicality, purpose, people, and patience

I gave a talk on AI recently, and as I was answering questions, something occurred to me: a lot of the questions people have are fundamental to how AI will actually function at work.

They weren’t simple or basic questions; they’re fundamental questions because they get to the core of why and if you should use AI. Questions of privacy, security, sustainability, and cost are the important things we need to be talking about. Things you need to think about if you’re planning on officially brining AI tools into your organization (psst, people are already using the tools even if you have “approved” them).

Where to start? I think there are five things critical for bringing AI into your organization and making it successful:

• Preparation
• Practicality
• Purpose
• People
• Patience
  

Skipping or skimping on any of these might not torpedo your AI pilot, but it will certainly make it harder to bring it to port. Let’s get into each of these and what you need to look at for each of them.

Preparation

The first P is preparation, and this is bigger than just planning. Is your organization even ready from a legal and privacy standpoint? Is there a budget? I’ve heard stories of companies bringing in AI for coding, only to get hit with a massive bill because a junior dev didn’t realize how many tokens they were burning through with every AI request.

You have to do your research. “Testing” doesn’t mean using the free version for a couple of days and going, “Yeah, that’s good.” It means getting a core group of testers across your organization—your future AI evangelists, who are probably already there, you just don’t know it yet—and paying for a pro license for a month or two. Get them into Gemini, ChatGPT, and Claude to really kick the tires.

Treat it like evaluating any other piece of software. What are your needs? What are the trade-offs (there are always trade-offs). It’s going to cost a little money to do this right, but it will save you a fortune later.

Practicality

Related to preparation is the practicality of it all. Lots of companies handle extremely sensitive information: health records, financial data, you name it. You have to decide if it’s practical for everyone to have access. Are the guardrails in the AI tools themselves good enough, or are you going to have to erect so many internal barriers that the whole thing just isn’t worth the effort?

This isn’t a simple “yeah we think it’s okay” exercise, you need to really think about it and ask yourself—what are the risks and can you manage them. Maybe not every department can use an external AI tool. Maybe Finance and HR need an internal system with tighter controls to make sure private data doesn’t wander its way onto the public internet.

On the flip side, don’t make it impractical for teams like Marketing, IT, Sales, and Operations to use AI, because even if you “ban” AI at work, people will still keep using it. Shadow IT has always been a thing—and (cough) I may have done a little of it in the past—but with AI there are real risks to your company with shadow AI use. Saying “we’re not going to use AI here” is like saying, we’re not going to use email anymore, it’s faxing from now on. Yeah how well do you think that would work out?

Purpose

Next is purpose. You have to have a real reason to do this. Because “it’s new, shiny and everyone is doing it” is not a purpose. Will it create direct cost savings? Will it save time so your team can do more creative, human work because the busywork is handled? You need a solid business case. You need a purpose for using AI—and how you’ll know if you’re achieving it.

And each part of your business will have a different purpose for AI. Your developers will use it to draft, test, and secure code. Finance will work with spreadsheets and reports. Marketing will use it to analyze data, generate ideas, draft content, and do competitive research. Don’t expect everyone to have the same purpose or success metrics—but everyone needs to have some sense of purpose before you start.

Measuring outcomes is going to be tough. People are terrible at estimating time. Asking “How much time did you save this week by using AI?” is like asking, “What did you have for lunch two weeks ago on Tuesday?”—you’re not going to get a useful answer. As part of your preparation step, think about what practical metrics you can gather to assess how AI is going—reduced dev time, fewer bugs per release, more optimized digital spends—things you can figure out without asking people fuzzy questions.

People

You can’t just flip a switch one day, announce everyone has access to a ChatGPT Pro plan, and expect them to succeed. That’s not how people work. Sure, a few folks—the ones like me who are already using it on the side—will hit the ground running. They’ll be your internal evangelists, but you can’t expect them to train the whole company; they’ve got their own jobs to do.

You have to plan for real training. Start with a general overview for everyone covering the basics and your internal rules on privacy and data. Then get real value from AI, you have to tailor the in-depth training to each department. Finance doesn’t care about brainstorming 20 email subject lines. Marketing doesn’t need a deep dive on securely deploying Python modules—or maybe they will if they become vibe coding junkies. IT doesn’t what to learn how to create a brand voice handbook. Each group will have their own use cases—and you might need different experts for each group. Better to have a team of trainers than forcing everyone down the same path.

Patience

The final, and maybe most important P, is patience. Yes, you’ll see some amazing, quick-win breakthroughs in the first few weeks. That’s the low-hanging fruit, picked while the tool is new and shiny. But the real transformation takes months. It happens in that post-new-and-shiny phase when people are learning to prompt better, building new workflows, and moving beyond just doing old tasks faster to doing entirely new things they never thought possible.

AI is not magic; there is no magic wand here. That oft-misquoted MIT study about 95% of enterprise AI pilots failing? It probably took them three months just to get going, giving people on the ground maybe three months to actually do anything. That’s barely long enough. Your pilot doesn’t start the day you decide to use AI. The clock starts the day after your first training session.

Check in with people—the AI evangelists in each department—on how it’s going. Monitor costs like API calls and cloud computing. Listen to what people are excited about and celebrate the spectacular failures as much as the wins. The spectacular failures are the stepping stones to going from saving time to transforming your business.

But be patient for the real results to come in.

So there you have it. The five P’s of bringing AI into your organization (or solo work):

• preparation
• practicality
• purpose
• people
• patience

If this sounds like a framework you need, reach out. I can come into your organization to help with training and defining those five P’s.

I might not be able to make you more patient, but I’ll at least try to help you get there.