Buying an AI licence isn’t an AI strategy
A lot of businesses are at a very strange point with AI.
They know it matters. They know people are using it. They know tools like ChatGPT, Claude, Gemini and Microsoft Copilot are not going away. Some have bought licences. Some have written an acceptable use policy. Some have had a one-hour awareness session. Some have a few confident people quietly racing ahead while everyone else hovers at the edge wondering what they are supposed to do with it.
And then the leadership team looks at the monthly licence cost and asks a very fair question:
Why hasn’t this changed anything yet?
That’s the bit we need to talk about.
Because AI adoption behaves much more like a modern way of working than a normal software purchase.
You can buy a CRM, set up users, move data, train people on the screens, and call that implementation. The project might still be painful, because CRM projects have a special talent for becoming everyone’s problem, but at least the shape is familiar. There is a system. There are fields. There are workflows. There are reports. There is usually a fairly clear before and after.
AI is different.
AI touches judgement, communication, delegation, process, confidence, risk, curiosity and habit. It can help someone write an email, prepare for a meeting, summarise a document, analyse feedback, plan a project, role-play a difficult conversation, draft a policy, question a process, build a training outline, explain a concept, compare options, or turn a messy thought into something usable.
That is useful.
It is also messy.
Because when a tool can touch almost every part of someone’s working day, adoption can’t just be about switching it on.
Access is not adoption
Giving people access to AI and calling it adoption is like handing someone the keys to a car and assuming they can drive.
Some people will be fine.
Some will be nervous.
Some will press every button to see what happens.
A few will accidentally reverse into the bins.
That is roughly where a lot of businesses are with AI at the moment. One person is using it for everything and proudly telling everyone they have saved 6 hours before breakfast. Another person tried it once, got a strange answer, and quietly decided it was all nonsense. Someone else is pasting sensitive information into whichever free tool appeared first on Google. A manager has asked for an AI policy because that feels like the responsible thing to do. The IT provider is being asked which licences to buy. The business owner just wants to know whether any of this will actually save time.
That is not a tool problem.
That is an adoption problem.
The technology might be available, but people still need to understand what it is good at, where it falls over, what they should avoid, how to ask better questions, how to check what comes back, and how to apply it to the work in front of them.
That is AI literacy.
Article 4 of the EU AI Act has made AI literacy harder to ignore
There is now a legal and governance reason to take this seriously too.
Article 4 of the EU AI Act is about AI literacy. The official text says providers and deployers of AI systems must take measures, “to their best extent”, to ensure a sufficient level of AI literacy for staff and other people dealing with the operation and use of AI systems on their behalf. It also says that technical knowledge, experience, education, training, the context of use, and the people affected by the AI system all matter.
The Act itself applies in stages. It entered into force after publication in the Official Journal, with general application from 2 August 2026, while Chapters I and II, which include Article 4, applied from 2 February 2025.
Article 4 of the EU Artificial Intelligence Act is about AI literacy. It says providers and deployers of AI systems should take measures, to their best extent, to ensure a sufficient level of AI literacy among staff and others dealing with AI systems on their behalf.
Now, because we are based in the UK, I want to be careful here. This is not legal advice, and not every UK business will be affected in exactly the same way. The scope of the Act includes providers placing AI systems on the EU market, deployers established or located in the EU, and providers or deployers in third countries where the output produced by the AI system is used in the Union.
So the legal position depends on your business, your customers, your market, your systems, and how AI is being used.
But even if your business is not directly caught by the EU AI Act today, the direction of travel is obvious.
Businesses are going to be expected to show that people using AI have been trained properly. That does not mean everyone needs to become technical. It means people need enough understanding to use AI safely, sensibly and in context.
And frankly, that is a good thing.
Because “we gave everyone access and hoped for the best” is not a serious adoption plan.
What should AI literacy actually include?
This is where a lot of training gets it wrong.
AI literacy is not a one-hour “look what ChatGPT can do” session. Those sessions can be useful. I deliver plenty of them. Sometimes people need to see the art of the possible before they understand why it matters.
But awareness is only the first step.
If someone is going to use AI in their role, they need more than a few prompt tricks. They need to understand the behaviour of the tool well enough to make sensible decisions with it.
That means they need to know things like:
- What AI is actually doing when it generates an answer
- Why it can sound confident and still be wrong
- How to brief it properly
- How to check the output
- What data should never be shared with public tools
- The difference between consumer AI and business AI
- When Microsoft Copilot is the safer option
- When automation is better than AI
- How bias can appear in outputs
- Where a human needs to stay in the loop
- How to use AI without letting it sand down their own skills
That last point matters more than people think.
There is a lazy way to use AI. Get it to write every email. Get it to answer every question. Get it to think through every problem before you have even tried. It feels productive for a while, but it can quietly erode the very skills that made someone good at their job in the first place.
The better way is to use AI as a coach, a critic, a drafter, a challenger, a translator, a research assistant, a planning partner, or a second pair of eyes.
That takes training.
Prompting is delegation
One of the biggest mistakes people make with AI is treating prompting like a magic phrase.
They go looking for “the best prompt”.
The better question is: have I given the tool a good brief?
When you ask AI to do something, you are delegating. That is why we teach our ROAR framework:
- Role: Who do you want the AI to act as?
- Objective: What do you want it to do?
- Appearance: What should the output look like?
- Restrictions: What rules or boundaries should it follow?
That might sound simple, but it changes how people think.
If you asked a colleague to produce a report, you would probably explain what the report was for, who it was for, what it should include, how long it should be, what format you wanted, and what to avoid.
Then people open AI and type:
“Write me a report.”
And somehow the computer is supposed to guess the rest.
That is why so many people get poor results. The tool is not reading your mind. It is working from the brief you gave it.
This is where I often use an amazing classroom example from America, involving a peanut butter and jelly sandwich. The teacher asks the class to write instructions for making the sandwich, then follows those instructions literally. It falls apart almost instantly because humans skip steps all the time. We assume shared context. We miss details. We say “put the peanut butter on the bread” and forget to mention opening the jar, picking up the knife, or which side of the bread to use.
Computers are painfully literal.
AI is more flexible than traditional software, but the principle still matters. If the instruction is vague, the result is a roll of the dice.
Better delegation gets better results.
AI and automation are not the same thing
This is another big one.
A lot of businesses say, “We want AI,” when what they really need is automation.
That is understandable. AI is the noisy word. It is the thing everyone is talking about. It is the thing vendors are putting in every product name, whether it belongs there or not.
But AI and automation do different jobs.
Automation is brilliant when the task is repeatable. A thing happens, then another thing happens. An email arrives. A form is submitted. A file is saved. An approval is needed. A reminder is sent. A record is updated.
Automation is the steady, safe pair of hands.
AI is more useful when language, judgement support, summary, comparison, drafting, explanation or interpretation is involved.
For example, if you want every completed form to create a task in Planner, that is automation.
If you want a long customer email summarised into the key points, that is AI.
If you want invoice data moved from one system to another, that is automation.
If you want a rough meeting transcript turned into a readable summary with actions, that is AI.
If you want a report dashboard with accurate measures and proper calculations, that is business intelligence.
If you want a narrative explaining what the numbers might mean, AI can help.
Businesses need to understand these differences before they start buying tools or building agents. Otherwise they end up using AI where automation would have been safer, cheaper and easier to maintain.
The “bloke I don’t know” test
Here is one of the simplest ways to think about AI risk.
When you say, “I want AI to do this,” replace the word AI with “a bloke I don’t know”.
“I want AI to go through my inbox and delete anything I don’t need.”
Becomes:
“I want a bloke I don’t know to go through my inbox and delete anything I don’t need.”
Suddenly, the risk feels very different.
You start asking better questions. What is he allowed to see? What rules is he following? What happens if he deletes the wrong thing? Is he preparing suggestions or making final decisions? Who checks his work? What data does he have access to? Can he forward things outside the business? Does he understand what matters?
That is the mindset we need.
AI can be incredibly useful, but the human still owns the outcome. If it writes something wrong and you send it, that is on you. If it summarises a policy badly and you act on it, that is on you. If it invents a source and it ends up in your proposal, that is on you.
The tool can help.
It does not take responsibility.
Training needs to be practical
The best AI training is built around real work.
Emails. Meetings. Planning. Research. Admin. Customer communication. Sales. Marketing. HR. Finance. Operations. Policies. Reports. Processes. Presentations. Follow-up. Notes. Ideas that are stuck in someone’s head and need turning into something useful.
That is where people start to understand the point.
AI becomes much easier to understand when someone sees it help with something they already do.
That might be:
- Turning meeting notes into actions
- Drafting a difficult email without sounding blunt
- Creating a first version of a process document
- Comparing 2 policy documents
- Summarising a tender
- Preparing questions for a client meeting
- Creating a social media post from rough notes
- Explaining a technical concept in plain English
- Role-playing a conversation with a member of staff
- Creating a checklist from a long document
- Building a first draft of a training outline
This is why generic training rarely sticks. People need examples that feel like their working day.
That is also why AI adoption has to include curiosity. The people who get good at this are not always the most technical. They are the people who keep asking:
Could I use AI to help with this?
Crawl, walk, run
At Techosaurus, we use a crawl, walk, run approach.
The crawl stage is about confidence.
People need to understand what AI is, what it can do, where it fails, and how to use it safely. They need to practise with low-risk tasks. Everyday tasks are often a good place to start because the stakes are lower. Meal plans, travel ideas, explanations, personal admin, rewriting a paragraph, making a checklist. This helps people get comfortable without worrying about breaking a business process.
The walk stage is about applying those skills to real work.
This is where people start using AI for emails, meetings, planning, research, document summaries, customer communication, internal notes, first drafts and business tasks. They learn to use better prompts. They learn to check outputs. They learn when to ask AI to interview them before producing an answer.
The run stage is where adoption becomes more structured.
This is where a business starts thinking about approved tools, shared prompt libraries, internal champions, AI policies, process documentation, automation, Copilot adoption, reporting narratives, and where agents might safely fit.
The order matters.
If a business jumps straight to “run” without the foundations, it usually creates confusion. People start building things before they understand the process. They buy licences before they understand the use case. They ask IT for answers to questions that are really about operations, HR, sales or management.
Crawl first.
It sounds slower. It usually saves time.
Evidence matters
This is where certified training starts to become useful.
If a business is trying to show that it has taken AI literacy seriously, it helps to have something more concrete than “we had a chat about it in a team meeting”.
You want evidence.
Evidence that people completed training. Evidence that the training covered practical use, risk, checking, prompting and business context. Evidence that learners received a certificate. Evidence that the business took reasonable steps to help people use AI properly.
That is one of the reasons we built our online course the way we did.
Practical AI for Everyday and Business Tasks is our self-paced online AI course. It is built for people who want proper practical AI training without taking 10 full days away from the business.
It includes:
- 20+ hours of Certified CPD
- 50+ hands-on tasks
- 11 business areas
- 12 months’ access
- A certificate on completion
- A LinkedIn badge
- Tool-agnostic methods that work across ChatGPT, Claude, Gemini, Copilot and other AI tools
- Our crawl, walk, run approach
- The ROAR prompting framework
- Practical examples people can use in everyday work
The course is designed to help people build confidence first, then apply AI to real business tasks, then start thinking more clearly about how AI fits into work.
It is not there to turn everyone into a technical expert.
It is there to help people become competent, careful and useful with AI.
That is the bit businesses need.
Why online learning matters
Live training is brilliant. I love being in a room with people, answering questions, running demos and watching the lightbulb moments happen.
But live training is not always practical.
Some businesses cannot take a whole team out for a day. Some have shifts. Some are spread across different locations. Some have budget constraints. Some want people to learn at different speeds. Some need a record of completion. Some want new starters to go through the same material later.
That is where online learning makes sense.
Our Practical AI for Everyday and Business Tasks course gives people a structured route through the basics and into real business use. They can pause, revisit lessons, complete tasks in their own time, and build evidence of learning as they go.
For businesses, it also means AI training does not have to be a one-off event. It can become part of onboarding, team development, management training, or an internal AI adoption plan.
A one-hour session can create interest.
A proper course can build capability.
What about in-person learning?
Online learning works well for many people, but some people learn best in a room.
That is why we have also launched Open Learning.
Open Learning is our way of making practical digital skills training available without funding criteria, waiting lists or complicated eligibility rules. We have spent the last 2 years delivering AI, automation and digital skills training across the South West, including work with more than 150 businesses. Open Learning takes that practical approach and makes it easier for individuals, small groups and local businesses to book onto focused in-person courses.
Our first Open Learning courses are:
AI Fundamentals is for people who want to understand how AI works, how to prompt it properly, how to use it safely, and how to apply it to real business tasks.
Copilot 365 Fundamentals is for businesses already using Microsoft 365 who want to understand what Copilot can do, how to use the business version safely, and where it fits into everyday work.
Automation Fundamentals is for people who want to understand triggers, actions, conditions, approvals, process mapping and the logic that sits behind good automation.
These in-person courses are a good fit for people who want guided learning, questions in the room, and a practical day away from the usual noise.
But for businesses that want wider AI literacy across a team, especially where time and geography are awkward, the online course is usually the easiest place to start.
Buying Copilot still needs training
Microsoft Copilot is going to be a big part of AI adoption for many businesses.
That makes sense. If you are already a Microsoft 365 business, Copilot is sitting close to the tools you already use. It has the benefit of your Microsoft environment, your security setup, and your existing way of working.
But Copilot still needs training.
People need to know which version they are in. They need to understand the difference between consumer Copilot and business Copilot. They need to know what is included in their Microsoft 365 subscription and what needs a paid upgrade. They need to know when the paid licence is worth it. They need to understand what Copilot can and cannot see. They need to learn how to brief it properly.
Otherwise, businesses risk spending money before people are ready to get value from it.
That is why I often say: exhaust what you already have before you rush to buy more.
The paid tools can be brilliant.
They are far better when people know what they are doing.
AI adoption is a people development challenge
This is the uncomfortable bit.
The businesses that get AI right will probably not be the ones with the longest tool list.
They will be the ones that train people properly.
They will give people safe ways to experiment. They will explain what good use looks like. They will document their processes. They will create internal guidance. They will choose approved tools. They will build champions. They will know when AI is the right answer and when automation is the better one.
They will also keep humans in the loop.
Because the goal is not to remove human thinking from the business. That would be a terrible use of AI.
The goal is to give people more room to do the work that actually needs them.
Less time fighting blank pages.
Less time digging through long documents.
Less time rewriting the same email from scratch.
Less time trying to remember what was agreed in a meeting.
More time thinking, deciding, talking to customers, supporting staff, solving problems and doing the work humans are actually good at.
Start with literacy
If your business is using AI already, even informally, it is time to take AI literacy seriously.
That does not mean panic.
It means getting people trained. It means giving them a shared foundation. It means helping them understand risk without making them scared of the tools. It means showing them real examples. It means giving them a route from curiosity to confidence.
That is exactly why we built Practical AI for Everyday and Business Tasks.
It is practical, self-paced, Certified CPD, and built around the real tasks people deal with every day.
If you want your team to use AI well, start there.
Because adoption does not happen when someone gets access to a tool.
It happens when they know what to do with it.