Move From Playing With AI to Getting Value From It

Published: 1 July 2025

By now, most people I meet in workshops have used ChatGPT, Microsoft Copilot, Gemini or another generative AI tool at least a few times, and many of them are using one of these tools almost every day.

They might use AI to rewrite an email, summarise a document, come up with some ideas for a presentation or help improve something they have already written. These are all useful ways to get started and, for many people, this is where their use of AI begins.

The question is what happens next and how you move from using AI for individual pieces of work to using it properly across the work you are responsible for.

That is where the value starts to become much more interesting.

Using AI is easy to start

One of the reasons generative AI spread so quickly is because there is very little stopping somebody from getting started.

You can open ChatGPT or Copilot, ask a question, get an answer and then keep the conversation going from there.

Within a few minutes, somebody can feel that they understand how to use the tool and to some extent, they do because they have learned how to have a conversation with it.

Things become more interesting when they want to use AI for something more substantial, such as preparing a proposal, reviewing a customer account, developing a recruitment plan, working through an RFP, producing a management report, analysing several documents or building a presentation using information from different places.

These tasks normally have more going on because there is a result that needs to be achieved, information that needs to be used, decisions that need to be made along the way and a final output that somebody needs to review before it can actually be used.

This is where people need a clearer way of working with AI.

Start with the work

When we build practical AI training, one of the first things we want to understand is what people actually do in their jobs and where their time and effort are going.

What work takes time? What gets repeated? What involves bringing information together? Where are people reading, analysing, comparing, preparing, writing, reviewing or making decisions?

Those questions give us somewhere useful to start because they bring the conversation back to the work rather than starting with a list of AI features.

Take Procurement as an example.

A team might regularly need to compare supplier responses against an RFP, which could involve several proposals, scoring criteria, technical requirements, commercial information and previous correspondence.

Someone may already be using ChatGPT or Copilot to summarise one of those documents, which is useful, but there is normally a much wider task around that summary.

We would look at questions such as:

• What needs to be compared?

• Which information should be used?

• What criteria need to be checked?

• Where are gaps or missing answers important?

• What should the final comparison look like?

• What does a person need to review before using it?

Now AI is supporting a larger piece of work where the person understands what they are trying to achieve, the information they need and the steps they need to work through.

The same approach applies across many different roles.

HR might be preparing a recruitment plan, Sales might be getting ready for an account review, Marketing might be reviewing campaign performance and deciding what to change next, Operations might be comparing activity across several locations, while a leadership team might be working through information before making a decision.

The tools people use may change depending on what their company has chosen to provide, but the way we think about the work stays very similar.

A workshop example

One example from a workshop with an events business has always stuck with me because the task was so normal and everybody could immediately understand why it mattered.

One of the participants was planning the food and beverage menu for an upcoming event and, as part of that work, they needed to look back through previous event feedback, find comments relating to food and beverage, think about the profile of the audience and then use all of that information to help develop the next menu.

There were several pieces of work involved.

They needed to find the previous feedback, pull out the useful comments, understand what people liked and disliked, bring that information together, consider the audience for the next event and then develop some initial menu options.

We used Copilot to work through those stages, which meant the participant could locate relevant feedback, bring the comments together and use that information as part of developing the next set of menu ideas.

From there, we continued working through the result by asking what would make the options stronger, what needed changing, whether there were cost considerations, whether some dishes might be more practical to prepare and what other information should be considered before anybody made a final decision.

The useful part was seeing AI being used across a task that the participant already understood very well.

Because it was their work, they could immediately see where the tool was helping and they could also see where the result needed to be questioned, checked or improved.

Generic examples only take you so far

There is still a place for general AI awareness because people need to understand what the tools are, what they can do and some of the basics of working with them.

That gives people a starting point, and after that you need to get closer to the work they actually perform.

Someone in HR is likely to get much more from an exercise involving workforce planning, employee communication or recruitment because those are tasks they recognise and can immediately relate back to their job.

Someone in Procurement is going to learn more from comparing supplier information, while someone in Sales starts seeing more possibilities when they work through account preparation, opportunity planning, customer follow up or a QBR.

Relevance matters because people start connecting what they are learning with the tasks sitting in front of them when they go back to work, which makes it much easier for them to use those skills again.

This is one of the reasons we build Microsoft 365 Copilot training and ChatGPT training around the work teams actually complete.

People also need a way of working

A useful task gives people something relevant to practise with, and they also need a way of approaching AI that they can use again when the next task is different.

At Koshima, we keep coming back to a few basic things.

You need to be clear about what you are trying to achieve, what the final result should look like and who it is for.

You then need to think about the information the AI needs in order to help you properly. That might include documents, emails, meeting notes, reports, spreadsheets or other source information, along with anything important about the situation that is not already obvious.

The instructions then need to be clear enough for the tool to understand what you want it to do with that information, what needs to be included and what it should pay particular attention to.

Larger pieces of work will normally need several stages, so you might first understand the information, identify gaps, develop the work, challenge what has been produced and then improve it.

Throughout that process, the person doing the work still needs to review what comes back.

Is it accurate? Is anything missing? Does it make sense? Has AI made assumptions? Can the important information be traced back to the source?

That review is part of the work because the final result still needs to be good enough for somebody to use.

Once people become comfortable working this way, they can start using the same approach across many different tasks rather than relying on a collection of prompts that only work in one situation.

Access gives people somewhere to start

A company can provide Microsoft 365 Copilot licences, give employees access to a ChatGPT Business workspace or approve another AI tool, which gives people somewhere to start.

The next question is what they are actually able to do with that access.

This is something we see regularly with Microsoft Copilot. People know they can draft emails, summarise meetings and create presentations, and then you show them how Copilot can support a wider piece of work involving several emails, documents, meetings or other sources.

At that point, the conversation starts moving towards their job and the work they are responsible for.

This is why giving employees Microsoft 365 Copilot licences is only the beginning.

The same applies to ChatGPT because somebody can use ChatGPT every day and still only use a relatively small number of ways it could help them in their work.

Capability develops through experience. People learn something, use it in their work, see where it helps, find where it struggles, improve the way they approach it and then try it on something else.

That is also why continued AI enablement can be useful after an initial workshop, particularly when people have the opportunity to bring back real tasks, questions and problems from their own work.

Measure what happens to the work

If the reason for introducing AI is to improve how work gets done, then that is what we need to look at.

Take a task before AI is introduced and understand how long it normally takes, where the effort goes, where mistakes are likely to happen, what normally needs reworking and how consistent the final result tends to be.

Then look at what changes once somebody learns to use AI effectively as part of that task.

Did the work become faster? Did the quality improve? Was the result more consistent? Did somebody spend less time searching for information?

Were they able to work through more information than they could before?

Did they still need the same amount of rework afterwards?

Those questions tell us much more about value because they connect AI use directly to the work people are responsible for completing.

Usage data can still tell you something useful because it shows whether people are using the tool, but the more important question is what changed as a result.

Start with something people already need to do

If you are trying to move AI use forward in your company, you do not need to begin with the biggest process you can find.

Start with a useful task that people already complete and where saving time, improving quality or making the result more consistent would genuinely matter.

Understand how that task works today, then work through it with the AI tools your company has available and look at where those tools can help.

• What information does the person need?
• Where can AI add value?
• Which stages still require judgement?
• What needs checking?
• What does a good final result look like?

Once you have worked through one task properly, you can learn from it and move on to the next.

Over time, people start becoming much better at looking at their own work and identifying where AI could help, which means they rely less on somebody else giving them a list of use cases.

They start asking a much more useful question: How could I do this better with AI?

That is when AI starts becoming part of how people actually work, and that is where the advantage starts to build.