Most companies are starting to make decisions about their 2027 technology budgets, and AI is going to be somewhere near the top of the list.
The conversations are already happening. Should we standardize on ChatGPT? Is Copilot the right answer because we are already a Microsoft company? Should we be looking at Claude? Do we need one platform or several?
Before getting too far into those conversations, I think there is something every CEO should do first.
Use the tools yourself.
Not sit through a demo. Not have someone show you ten prompts that worked really well. Actually use AI on work that matters to you.
Why Personal AI Experience Skews Platform Decisions
Ask a group of people who use AI every day which platform is best and you will get very confident answers. The person who uses ChatGPT will tell you why ChatGPT is better. Someone who has spent the last six months in Claude will make a great case for Claude. If your company lives in Microsoft, someone will tell you Copilot is the obvious answer.
They may all be right for themselves. That doesn't necessarily mean they are right for the company.
The more time you spend with one of these tools, the better you get at working with it. You build projects, instructions, examples and ways of giving it context. You learn how to push back when the first answer isn't good enough. Over time, the experience gets better because the tool has more of the right context and you have gotten better at using it.
That makes it difficult to separate two things: Is this actually the better tool, or am I just better at working with it?
I see this in my own use. There are tools I naturally gravitate toward because I know how to get what I want out of them. That doesn't automatically mean I should tell an entire company to standardize on the same thing.
From Personal AI Use to Enterprise AI Strategy
I think about this as the difference between Localized AI and Enterprise AI.
Localized AI is AI that gets better around the individual. I have my AI. You have yours. We each develop our own way of working with it, build up context around what we do and figure out where it makes us better.
A lot of the progress we have seen over the last couple of years has happened this way. Someone starts using ChatGPT or Claude, gets better at it, finds a few ways it improves their work and keeps going.
There is a lot of value in that, and companies should encourage it.
How does what an individual learns become something the company can learn from?
That is where Enterprise AI starts.
Not every prompt someone writes needs to become a company process. Not every personal workflow needs to be standardized. But when someone figures out a better way to do important work, and that improvement is valuable and repeatable, the company should have a way to capture it.
Which workflows should be shared? What data does AI need access to? What knowledge should belong to the company instead of sitting inside someone's account? What needs governance? What should stay flexible? What becomes part of the way the company operates?
And what happens when the person who figured it all out leaves?
Does the next person benefit from everything that was learned, or do they start over?
This is why one of the most important questions going into a 2027 AI budget discussion is: Where will our company's context live, and who owns it?
That question is much bigger than whether ChatGPT beats Claude on a particular task.
It also changes the way I think companies should budget for AI.
The cost of AI is not just the cost of the licenses. Obviously, the licenses matter, especially once you start rolling them out across a company, but that is only one part of the investment.
There is the data the AI needs access to. There are the workflows you want to improve. There are SOPs and instructions that explain how your company actually works. People need to learn how to use the tools. Someone needs to think about governance, security and permissions. Someone needs to measure whether people are actually using any of this and whether it is making the work better.
And someone needs to own it.
I also think the budget has to leave room for both Localized AI and Enterprise AI.
You don't want to lock everything down so tightly that people stop experimenting. A lot of the best ideas are going to come from individuals trying new things and finding better ways to work.
At the same time, you can't leave everything localized forever. The things that repeatedly make the work better need a path to become company capabilities.
That doesn't mean turning every good idea into a giant technology project. It means having a way to recognize what is working, capture the learning and decide what is worth making repeatable.
Localized AI is where people learn. Enterprise AI is how the company learns.
Don't Spend Too Much Time Picking the Winner
This is also why I don't think companies should spend too much time trying to pick the permanent AI winner.
The technology is moving too quickly. Models get better. New capabilities show up. Pricing changes. Enterprise controls change. Different tools are going to be better at different kinds of work, and I expect that to keep changing.
I would spend less time trying to predict which platform wins and more time figuring out how the company manages the data, permissions, workflows and knowledge that it is going to need regardless of which tools are being used.
How to Evaluate AI Platforms Before You Buy
If I were a CEO heading into the 2027 planning process, I would spend some real time using AI before making those decisions.
Pick something you would normally have a hard time delegating. A board memo. A strategic question you have been wrestling with. A difficult client situation. Something where there isn't an obvious right answer.
Then work through it with AI.
Try more than one tool. Give them real context. Push back on the answers. Tell them what they got wrong. Ask another question. Keep working with them.
You don't need to turn this into a giant platform evaluation. The point is to understand the experience for yourself.
You will see where AI is better than you expected and where it still falls short. More importantly, you start to understand how much the result depends on context, information and the way the person works with the tool.
Once you experience that yourself, I think the questions you ask your team change.
Instead of just asking which platform we should buy, you start asking which work we actually want to change. What have our best users already figured out? What should stay with the individual and what should belong to the company? Where do we want people experimenting and where do we need tighter controls? Which of these individual gains are worth turning into something repeatable?
Those are the conversations I would want to have before deciding what goes into the budget.
A lot of boards are telling management teams they need to use more AI right now. I get it. There is a lot of change happening and the pressure to move faster is real.
But "use more AI" is not much of a strategy.
The bigger opportunity is figuring out how to take what people are learning individually and turn the best of it into something the company can build on.
Use the tools yourself first.
Then figure out what should stay Localized, what should become Enterprise and what the company actually needs to own.
Then have the budget conversation.
Larry
