A small company I work with decided to bring AI into the business. Once the whole team was on one approved AI tool, it took less than a month before a few people started building their own internal solutions. One of them built a social media calendar and a first draft of 350 posts across several channels.
In a large organisation, I have seen people wait for sign-offs just to get access to a test environment. By the time access arrives, the idea is stale.
Same technology, very different pace. So who is going to win?
TL;DR
Small businesses win on speed. They want real results fast, and fewer people stand between an idea and a test.
Enterprises win on scale. The long-term gain is bigger, but they start with policy, data and governance, so everything moves slowly.
Both have the same hidden problem: people already use AI without telling anyone.
In my experience, a few people use AI every day, a big middle group tries it and goes back, and some never use it. Leadership usually thinks the picture is better than it is.
Action beats planning.
What does a small business do differently?
The company I mentioned had a simple goal: grow revenue without hiring more people. That is easy to measure, and it makes every process a plain question. Can AI help us do this with the people we have?
The steps were not complicated. Audit the real processes. Write simple governance, so people know what they can use. Move the whole team onto one default AI tool in a safe environment, and show them how to use it. Even a small business has proprietary data and its own way of working, and that has to stay inside.
After that, people are creative. Give them a safe place and a few examples, and they find ways to make AI help with their own day-to-day job.
The business is growing without adding headcount, but it is still early, so I will not call it a result yet.
What holds small businesses back?
Mostly one thing: they do not know where to start.
People inside are already experimenting, so shadow AI is already there. Leadership knows something is happening but not what to do about it. They often have nobody to lead the journey, and sometimes no budget.
The US Census Bureau’s business survey fits this: fewer than 20% of firms with four or fewer employees use AI, against 37% of firms with 250 or more. I think the small ones are not behind because they are not interested. A lot of them have no map.
My advice: define the goal, audit where AI could help, write the policy first, then pick one team plan or the AI already inside your Microsoft or Google workspace. Start with a few workstreams, find the curious people, and let it grow.
Why is it so slow in a large organisation?
The shadow AI problem is the same. In a Deloitte survey of UK workers, 31% of GenAI users said they use it without their employer knowing. Enterprises have the most to protect and often no way to see it.
So they start with the safe things. What data do we have, who owns it, which tools are mandatory, which are banned. All reasonable. But from idea to test to production, everything is slow.
In my experience, most enterprise AI projects die between the idea and the first prototype. Too many people, too much red tape, too many opinions. Some people also prefer the status quo, and they can slow things down without saying so. If access to test an idea takes months, the idea is gone.
Some large companies do it differently. In 2024, one large pharmaceutical company reported around 750 custom assistants built by its people within about two months of rolling out an enterprise AI tool, with 40% of weekly users building one. They made a safe sandbox and let people experiment inside it. Is that the best approach? I do not know. But it is better than nothing.
Scale is where enterprises are strong. In McKinsey’s 2026 survey, 54% of respondents from organisations with $1 billion or more in revenue said AI is scaling across the enterprise, against one third of smaller ones.
Who is actually using it?
It depends on the person. In my experience, any organisation splits into three groups.
Pioneers, about 5 to 10%. They use AI every day, on every task. In a usage report it is less about how many tokens they burn and more about how often they come back.
The messy middle, about 60 to 70%. They try something, go back to the old way, try again. They listen to their peers, so pioneers are the best people to lead them.
Non-users, about 20 to 30%. The most common reason is “we are too busy to try something new”. Others dislike AI or worry about their jobs. Staying with the old way is not a winning strategy long term, but I understand it.
Most people use it for simple things: write an email, fix grammar, do some research, summarise a long document or pull data out of it. Leadership often thinks AI is embraced. The truth is that changing a workflow is the hard part. In the same survey, 31% of all workers said they used GenAI at work but saved no time, about half of its users.
What did not work?
I build a lot of tools. Some reached production and were never used.
It usually goes like this. A team says they really expect a tool. I build it. They try it once, it does not give the result they expected, and they go back to the old process. Nobody tells me why. When I do get feedback, it is vague, and two people in the same team can want completely different things from the same tool. Sometimes, before I even start building, someone says they do not need it.
I do not have a fix. I think it is mostly communication, and people sticking to processes they know. Expect the first version to disappoint, and ask for feedback earlier than feels comfortable.
So who wins?
I think small businesses will challenge some big ones. They can now buy a high level of intelligence at a low price, they are more flexible, and they are bolder. They can build quickly what a big organisation cannot. Enterprises, I assume, will win on scale if they get it right, and gains of that size can make a whole business more sustainable.
AI Champions Free Virtual Conference on 29 October 2026!
4 hours of AI cases to help you become AI-native and build an AI-first organization!
5000+ attendees and 20+ speakers from Google, Amazon, DoorDash and more
What can you do on Monday?
In a small business:
Find out what leadership thinks and what is already in place. If something exists, start using it, even a simple copilot
If nothing exists, open one free account, test a few ideas from your own work, and show the output. Ask: do you think we should do more?
They probably know already and just do not know where to start. Be the first one
In a large organisation:
Check what exists. If something does, use it, share it, and write down use cases
If adoption is low, try it yourself and ask your line manager if there is a policy
Be the squeaky wheel if you need to. It is a bigger battle, but every path starts with one step
FAQ
Do small businesses adopt AI faster than enterprises? In my experience they move faster once they decide. Survey data shows large firms ahead overall, with 37% of firms with 250 or more employees using AI against fewer than 20% of the smallest in the US Census Bureau’s May 2026 data.
Why do enterprise AI projects stall? Red tape, many people involved, many opinions, and slow access to test environments.
What is shadow AI? Employees using AI tools without their employer knowing. In a Deloitte survey of UK workers, 31% of GenAI users said they do.
Final Thoughts
The numbers say large firms are ahead. What I see day to day is small teams doing in a month what some big teams plan for a year. Both can be true, and I find that exciting. Strong AI is now available to almost anyone at a low price, so a small team with a clear goal and a safe place to work can build things that were out of reach not long ago. Enterprises that get to scale will probably gain even more. I think there is room for both.
So try something today, for your work, your business or just yourself. You do not need a perfect plan to take the first step, and people who start usually learn faster than people who wait.
You will fail at some point. Everyone does, and it does not matter. You cannot fail if you do not try, so try, fail, learn and keep going.