Before You Use AI in L&D, Ask These Questions

Be smart, friend

Author

Ross Stevenson
July 8th, 2025

📨 Subscribe | 🤖 AI Training | 🚀 Courses

 TODAY’S THOUGHTS ☠️

Hey there 👋,

For those of you who don’t know, I’m British, and that means moaning about the weather is a sacred pastime.

Any conversations with fellow Brits starts with commentary on the weather.

Usually it’s a negative point as it is dark, gloomy and cold on most days. But this week, we have something else to moan about - the heat 🥵.

You’d think for a place that’s cold and gloomy, the presence of sun and heat would be welcomed like a new Taylor Swift album. That’s not the case, though.

I’m writing this in a current British yearly high of 28 degrees heat. No air con (not part of homes in the UK), only a wardrobe full of black clothes and an increasingly fast-melting slab of ice on my neck.

All of this to say, I’m suffering for my art today.

Moving swiftly to today’s actual conversation…

Some of the most popular questions I get in my DMs are “What AI tool should I use for x in L&D?”, and “How can I implement AI in L&D?”.

While you might think they sound harmless, they have unwanted problems beneath the surface. It’s rare for those who ask me these questions to have ‘defined the problem to solve’, and as you can guess, that’s my follow-up question.

Experimenting with AI is great, but looking at everything as a nail and AI as your local hammer is not a smart move.

With lots of AI-powered tools available in the workplace, it’s vital to distinguish what generative AI can and can’t do.

We need to go beyond the how and what to ask ourselves why, when and where can AI be useful in L&D?

So today, we’re exploring how you can identify where to get real value from AI across your L&D workflow.

I’ll be sharing a host of research, reports and frameworks to help guide your decisions.

Get your tea or beverage of choice ready, 🍵.

We've got lots to discuss!

P.S. Your app might clip this edition due to size, if so, read the full edition in all its glory in your browser.

 IN THIS DROP 📔

  • How to assess what AI can actually solve in your L&D work

  • Why 80% of AI projects fail (and why that’s not a bad thing)

  • Everyone’s using it, but is performance improving?

🙋‍♀️ Want to reach over 5,000 L&D pros? Become a Newsletter sponsor in 2025

 THE BIG THOUGHT 👀

Before You Use AI in L&D, Consider This

It’s happening in every workplace

There’s a time and place for everything.

  • Dumb hairstyles = school and college

  • Overpriced tight-fitting clothes to impress a potential partner = your 20s

  • Not being judged for eating an entire chocolate log = Christmas

  • Using generative AI tools = ?

While I hope you agree with the rest, the last one is debatable.

Depending on your relationship with AI, your view on ‘why, when and where’ to use its delightful powers can be vastly skewed.

The ‘AI cultists’, as I like to call them, will proclaim we should use AI for everything, while ‘doomsdayers’ will warn you not to touch it as you’ll lose your humanity.

Of course, the truth of the matter is not so clear cut.

There’s an interconnected web of assessments and decisions to be made. The good thing is this is all human-powered.

The world has been so focused on ‘how’ to use new tools, that we’ve paid little attention to why, when and where.

This is one of the reasons I created the AI For L&D Crash Course. Not to just teach the ‘how’ because AI adoption is much deeper than that.

We need a wider decision-making toolkit.

So here’s a few frameworks and useful pieces of research to help you do just that.

Assess tasks, not jobs

I appreciate LinkedIn CEO Ryan Roslansky’s concept of assessing ‘tasks, not jobs’ in the context of generative AI at work.

This idea originates from Ryan’s Redefining Work article, where he explores how AI will accelerate workforce learning and amplify the importance of skills.

Ryan suggests moving away from viewing jobs as titles and instead, seeing them as a collection of tasks. These tasks will inevitably evolve alongside AI and other technological advancements. He recommends breaking your job down into its primary daily tasks.

You can bucket those tasks in this format:

  1. Tasks AI can fully take on for you, like summarising meeting notes or email chains.

  2. Tasks where AI can help improve your work and efficiency, like helping to write code or content.

  3. Tasks that require your unique skills – your people skills, like creativity and collaboration.

This sets the stage for how I currently recommend working with AI.

Explore where AI helps best

You might see a lot of glamorous uses of Gen AI tools on social media.

The reality is that the vast majority see benefits in the boring and basic tasks. I’m talking about writing better emails, summarising reports and brainstorming ideas.

This isn’t a surprise.

It’s smart to delegate the simple and mundane but incredibly time-consuming tasks.

This creates space to do more human stuff.

I don’t understand why some seem intent on getting AI to do the human stuff. What a boring life that would be. I want the AI to do the laundry with a workflow so I can build cool stuff, not the other way around.

Source: Asana AI at Work Report

Source: Gallup

A bunch of smart folks have done lots of research on this.

The above visuals come from Gallup and Asana, but I want to talk a little bit about a joint research project from Boston Consulting Group and Harvard.

These two powerhouses wanted to cut through the hype to see if AI tools like ChatGPT can improve productivity and performance. They worked with 758 BCG consultants (about 7% of their individual contributor-level staff) and split them into three groups:

  • One without AI access

  • One with GPT-4

  • Another with GPT-4 plus some training on prompt engineering

These consultants tackled 18 real-world consulting tasks to see how AI would affect their work.

The results were pretty impressive, I've got to say.

The consultants using AI managed to complete 12.2% more tasks and knocked them out 25.1% faster. But here's what really caught my attention - the quality of their work shot up by more than 40%!

It’s one thing to do something at speed, but another to do it at such high quality too.

That’s the trap I see happening in every industry right now. Too many prioritise speed over quality. You can have both, if you craft the right skills to collaborate with AI.

There was a catch, though (when is there not!).

When consultants tried to use AI for tasks it wasn't really built for, their performance dropped by 19%.

I don’t see this as a negative.

It’s very helpful to know where the limitations are. You cannot have a balanced approach without this. Another particularly interesting outcome was how the consultants ended up using AI.

Some folks took a hybrid approach, blending AI with their own expertise, while others went all-in and relied heavily on AI. Both styles seemed to work, but context was key.

While those marked as novice employees found the biggest performance gains, this dropped with those classed as experienced workers. Those in the latter category still saw a modest boost of 15% in most tasks.

TBF, I’d take that on most days.

Identify problems AI can actually solve (aka don’t kill a fly with a bazooka)

A lovely and, in my opinion, quite eye-opening piece of research on why most AI projects fail revealed that too many people try to get AI to work on problems it can’t solve.

Yes, I know that might be shocking to read, but AI is not a cure-all.

The sweet sauce exists in applying AI to problems it can support and solve.

Likewise, don’t go into overkill mode. If the task is simple, keep it that way. Don’t spend days building an AI-powered workflow to do it.

To do that, we need to understand the range of options at your disposal.

We have:

The list goes on.

This video will give you clarity on the AI model landscape, and what you can use vs your task:

Where AI is not your friend

I know it will break some hearts to hear this, yet…AI is not your saviour.

Life is about opportunities and pitfalls.

The research from BCG and Harvard provides an important lesson: Gen AI works really well when it’s used for tasks it can handle…but…outside of that, it’s the wild west.

As always, context is key in this type of decision-making, and tools are always improving. This is the part where I like to call on everyone’s common sense. Yet, as I’m continually pressed on, common sense, it seems, is not so common these days.

There’s no way I can cover every task in every industry you will encounter.

So, below is a general framework to help you consider when and when not to use Gen AI based on the difficulty of the task.

The summary is pretty simple: AI works well with tasks with pre-defined guidelines and less severe consequences of a f**k up. It should not be relied upon in what I class as ‘mission critical’ matters, aka the human stuff.

The over-reliance on AI is already a big threat to education, work and life.

We spoke about this in an edition covering the “Hidden Impact of AI on Your Skills”.

This is why I always come back to helping people with the mindset and behaviours to use AI intelligently. [Note: I define ‘using AI intelligently’ as one who understands the why, what, how and when of AI application vs tasks].

Adoption can very easily become addiction.

Choose wisely

A quick way to identify tasks AI can help with

This is the thing we all need help with.

Where can and can’t AI help me?

There’s no clear-cut answer to this. I’d love to give you some fancy 2x2 framework, but I don’t believe that will serve you well. Each scenario is context specific, and generative AI tech is evolving so fast.

I tend to think about my tasks in a macro and micro view.

Your tasks can easily be broken down into sub-tasks (micro). We’ve talked about continuing to invest in your thinking in this era of AI. This is something that requires deep thought and reverse engineering your ideal outcomes.

As an example, I use a little table like this.

It’s not fancy, but it does the job.

We have two macro tasks:

  • Presenting insights and actions on the L&D function’s performance to senior leaders

  • Launching a new internal course

For our first task, my outcome is to deliver a presentation to senior leadership on L&D performance.

So, I break down (in my mind) the micro tasks to reach that, as you see above. I then assign each of those to a column. Note: The first column can be automated without AI.

I don’t use this for every task, only those that I believe, with my current experience of Gen AI, could be an opportunity to work smarter.

What’s key is that the AI components are always low to mid-level, and the mission critical parts are always done by me (the human).

Final thoughts

Knowing how to use AI tools is useful.

But understanding why, where and when to call upon their power is an advantage.

I say this sooo often, but it’s a damn good quote and continues to be relevant in this space:

“With great power comes great responsibility”

Uncle Ben (Spiderman’s uncle)

Think wisely about when to wield that power.

→ If you’ve found this helpful, please consider sharing it wherever you hang out online, tag me in and share your thoughts.

Till next time, you stay classy, learning friend!

PS… If you’re enjoying the newsletter, will you take 4 seconds to forward this edition to a friend? It goes a long way in helping me grow the newsletter (and cut through our industry BS with actionable insights).

And one more thing, I’d love your input on how to make the newsletter even more useful for you!

So please leave a comment with:

  1. Ideas you’d like covered in future editions

  2. Your biggest takeaway from this edition

I read & reply to every single one of them!

 👀 ICYMI (In case you missed it!)

  1. I released a AI For L&D Readiness Assessment tool. You might have seen this shared on LinkedIn or over on learning news. This tool has one goal = help you assess where you are on the AI in L&D maturity curve vs other L&D pros, and give you the guidance to keep growing. It’s free to use and doesn’t need any personal data.

 VIDEO THOUGHTS 💾

Everyone’s “using AI”, but few know how to REALLY use it

Give everyone a license to the latest large language model, and you’ve made everyone adopt AI, right?

❌ WRONG

Getting real ROI from AI in L&D and beyond is a long game. In this video, I explore what the research tells us about the reality of AI skill development globally and a few ideas to accelerate your efforts in this space.

Enjoy!

🙋‍♀️ Want to reach over 5,000 L&D pros? Become a Newsletter sponsor in 2025

P.S. Wanna build your L&D advantage?

Here’s a few ways I can help:

Reply

or to participate.