How it started… and how it’s going now

Lately I’ve been writing a lot about AI from the outside in. Why rolling out AI isn’t like rolling out software. Whether organisations are ready for AI that does things. What we might be giving away when we hand over the hard thinking. And why nobody wants to be bad at something in front of other people.

All of that is about how other people and organisations take AI on. So I thought I’d take a moment to turn the lens on myself. How am I actually using AI right now?

Just over a year ago I shared my top 4 Copilot use cases. That blog was all about drafting, tweaking, brainstorming and getting AI to help me read and learn faster. It was a “how it started” moment.

So… how’s it going?

A lot has changed. My workplace is now driving Claude across the business, so Claude has become my main AI tool. I use Copilot less these days, but it still has a place in my work, especially where it comes up across M365 such as within the apps. I also want to explore Copilot Cowork a lot further, so watch this space.

The biggest shift isn’t the tool though. It’s how I work with AI. A year ago, AI helped me draft, tweak, brainstorm and summarise. Now it does the work with me, using my own methods, templates and skills. I’ve gone from chatting with AI to delegating to it.

Why Claude works for me

The simplest way I can explain it is that it flows. I can think, draft, build and design in one place without jumping between tools. I can start with a messy idea and end up with something that looks finished, not just written.

The real game changer has been skills. In my last blog on this topic I talked about structured prompting, and how more detailed prompts got better results. That’s still true. But now I can package what works into a skill: my method, my templates and the questions I’d normally ask, all saved so I don’t have to rewrite the prompt every time.

And something that I really love is that once a skill exists, it isn’t just mine. Colleagues can use it too, and get the same quality and structure I would bring. That’s when AI moves from a personal productivity trick to something that lifts a whole team.

So, here are my top use cases right now…

1. Planning adoption when I didn’t have the time

This one is a bit meta. We used Claude to help roll out… Claude.

Our organisation was taking Claude from an early adopter group to the whole organisation. As a change manager, this is exactly the kind of thing I want to be deep in. But the reality was I was busy and I couldn’t give it the time it deserved.

So I did what I could do quickly. I pumped my change and adoption templates into Claude, along with my thoughts, ideas and rough notes. Claude pulled it together into an outline of adoption activities. I then handed that to our tech experts so they could run with it. they were able to review a timeline for comms, activities, and detailed a learning path.

What I love about this is that I didn’t become the bottleneck. My change thinking still made it into the rollout, even though I couldn’t be hands-on. The tech team got structure and a starting point, rather than a blank page or a “let’s wait until Megan is free”.

2. Kicking off change (with a skill I built)

That adoption experience links nicely to this one. I built a skill to help anyone internally kick off planning for a change.

It starts with an intake interview. Claude asks the questions a good change manager would ask. What’s changing? Who’s impacted? Why now? What does success look like? From there it produces a first draft Change Overview on a page and a Lean Change Impact Assessment, both in our branded templates.

The value isn’t just speed, although going from a blank page to a solid draft is a big win. The thinking is built in. Someone kicking off a change doesn’t need me in the room to get off to a good start. They get prompted to consider the right things early, which is when it matters most. This doesn’t replace any work for a client, or remove the need for a Change Manager. It captures initial detail to help begin the thinking and contribute to the planning.

3. Learning design, end to end

In my last blog I talked about using Copilot to create learning outlines and assessment. This is that idea, grown up.

I now use two skills that flow into each other:

• Course design ideation. I give Claude a deck, a transcript or even just a course idea that might be scribbled in Whiteboard. It works through the overview, the research behind why the learning matters, the behaviours we want to see, a learner analysis and the learning goals.
• Learning journey planning. From there, it builds a Learning Design Brief, learning outcomes and a blended learning plan. That covers how it all comes together across the LMS, video, eLearning activities and live sessions.

What used to take me days of structuring now gives me a strong draft much faster. That frees me up for the parts AI can’t do well: understanding the learners, knowing the organisation and making the judgement calls about what will actually land.
It even gathers research on the challenges employees are facing, which helps build the need for the learning and the data behind the ‘why’. All of this is created as a very detailed starting point. We then sift through and built upon it. We don’t use this to substitute our work, it supplements. It gathers and structures, for us to then review, discuss, and extend. This also helps us because at times you stare at a blank page and start thinking about what the elements could be. Working this way gets the juices flowing rapidly for ideation and review to be deep and valuable.

4. Turning a two-day offsite into a 12-month plan

We recently had a two-day offsite as a team. Lots of great conversation, lots of ideas, and the usual worry: how much of this will survive past Monday?

As we’ve found to work well previously, we recorded every workshop session and discussion. We then used the transcripts to capture the ideas, goals, actions and issues that came up along the way.

Claude turned all of that into a detailed program of work, including our goals and the projects to get us there over the next 12 months. It even absorbed messy whiteboard ideation from a photograph which I found so useful in itself!

What I love about this is that nothing got lost. The quiet comment in the last session carried as much weight as the big idea on day one. And nobody had to spend a week writing it all up, so we could get straight into the work. I also turned the spreadsheet into a single page programme of work and easily translated sections into our EOS business framework.

5. Working with my brain, not against it

A while ago I shared how much I love using AI to summarise long, dry content. As a neurodivergent person, I’m not always in the best head space to absorb and focus, and that support makes a real difference. If I am going to read a 30-page journal article about learning theory, I like to read the summary and overarching insights or points, then dive into the full article with that context.

This has gone up a level from helping with detailed content to helping me across my role and life. Claude is now connected to my Microsoft 365, so it can look across my email, calendar, Teams and files. Instead of trawling through everything to work out what needs my attention, I can just ask.

I can use a morning brief that pulls my priorities together in one place to start the day. It’s the same principle as before: lower the effort it takes to get started and take in information. The difference is that now the information comes to me. I can offload mundane admin to then spend more time in the strategy and deep thinking.

Beyond work: Claude at home

It’s not just a work tool for me. At home, it’s become my go-to for the life admin and curiosity that never quite fits into a busy week.

• Research: from the best shoes for a specific injury to making sense of health reports and recovery plans before I see a professional.
• Garden planning: what to plant in spring, which fruit trees suit a small yard, and the ongoing battle to keep possums off the passionfruit.
• Kids’ education: comparing secondary school options in Victoria, including online schooling and when TAFE becomes an option.
• Travel prep: for Japan, everything from locations and driving routes to visa requirements and which medications are allowed into the country. It even created a high level list of key phrases and etiquette to use on our recent trip.

The common thread? It saves me hours of searching and turns scattered questions into a clear plan.

What I’ve learned this year

• Context beats clever prompts. Give it your templates, methods and examples. The more of you it has to work with, the more useful it is.
  
• Delegate the first 70%, keep the judgement. AI is brilliant at structure and first drafts. The thinking about people, culture and what will actually work is still mine. And as I wrote about recently, I want to keep doing that hard thinking on purpose.
  
• Skills turn one person’s know-how into a team capability. If you do something the same way every week, it’s probably worth packaging up.
  
• I’m still accountable. I review everything, especially anything that goes to a client or out to the business.

Then vs now

Last time I left you with a list, so here’s the updated version.

• Then: drafting, tweaking, brainstorming, reducing reading time, engaging to deepen learning and creating learning outlines.
• Now: delegating whole pieces of work, packaging my methods as skills, kicking off change, end-to-end learning design, turning workshops into plans, a connected daily workflow, and a fair bit of life admin at home.
• Next: exploring Copilot Cowork, and how it fits alongside Claude. I have hesitated with this area, knowing it can do a lot of work and I really may not have full control. I do think it is time to dip into and understand this further.

I’d love to hear from you. What have you moved from chatting with AI to delegating to it?