Learning Academy Blog

Why E-Learning Still Takes So Long to Build

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 02 Jun 2026 11:30:01 GMT

If you work in Learning & Development, you’ve almost certainly encountered this reaction from outside the profession:

“Surely modern tools make e-learning quick to produce now?”

On the surface, it feels like a fair question. After all, the tools we use today are dramatically more capable than the ones many of us started with 10 or 15 years ago. Add AI into the mix and it’s easy to assume that development times must have collapsed.

But interestingly, the reality doesn’t seem quite so straightforward.

The surprising persistence of development time

Recently, while sorting through some old folders on my computer, I rediscovered a well-known 2010 report from Chapman Alliance about e-learning development times.

The headline figure from the report was striking. At the higher end of complexity, organisations reported spending roughly 184 hours of development time for every one hour of finished e-learning.

Importantly, that figure included the whole project effort:

instructional design
SME input
media production
quality assurance
project management, and
review cycles

Not just authoring time.

Revisiting those numbers made me wonder whether they still held true today. After all, since 2010 we’ve seen huge changes in authoring tools, media production, collaboration platforms, accessibility tooling and, more recently, AI-assisted workflows

Surely development time must now be dramatically lower?

Yet when you look at more recent industry benchmarking and practitioner data, the modern estimates for complex interactive e-learning often still sit somewhere between 80 and 150 hours per finished learning hour.

Lower than 2010? Yes. But perhaps not lower in the way many people might expect.

The “more powerful, not necessarily faster” effect

I suspect something slightly counterintuitive has happened over the last decade or so. The tools haven’t simply made development faster. They’ve made it possible to produce more sophisticated output within a similar amount of time.

In other words, the capability has increased more dramatically than the efficiency.

Think about tools like Storyline. Many of the core production mechanics still involve significant manual effort:

building interactions
configuring triggers
managing layers and states
testing logic
refining navigation
checking accessibility, and
iterating through reviews

The interface may have improved. Certain workflows may be smoother. But sophisticated interactive e-learning still involves a considerable amount of detailed point-and-click production work. And interestingly, improved tooling can sometimes encourage greater complexity rather than faster production.

Features that once felt too difficult or time-consuming suddenly become achievable, which naturally encourages designers to create richer, more ambitious experiences.

The result? More sophistication. But not necessarily dramatically less labour.

AI is helping — but mostly around the edges

Of course, AI is already speeding up parts of the workflow. It can help with outlining content, drafting objectives, generating ideas, summarising SME input and producing first drafts

All genuinely useful. But most current AI tools still don’t remove the deeper production effort involved in:

building interactions
implementing design decisions
governance and stakeholder review
testing and refinement, and
organisational approval cycles

And in larger organisations, those review and governance layers can account for a substantial proportion of total project time.

So while AI may improve efficiency in certain areas, it hasn’t yet fundamentally transformed the overall production reality of high-quality interactive learning.

What's the significance of this?

I think this creates a real tension for many L&D teams.

Externally, there is often an assumption that modern technology should make production almost instantaneous. Tool vendors don’t always help here either, with messaging that suggests sophisticated learning experiences are quick and effortless to create.

Internally, this can leave L&D professionals feeling that they are somehow inefficient or behind the curve.

But perhaps the more realistic conclusion is this:

Creating genuinely thoughtful, interactive, well-crafted e-learning still requires a significant amount of human judgement, design thinking and production effort.

The technology has become more powerful. But the craft itself (and the labour behind the craft) still matters.

And perhaps that shouldn’t surprise us quite as much as it does.

A fuller exploration of these ideas — including the wider implications for e-learning production, AI and workflow support — can be found in the original Learning Re-Framed article here:

Why Some Learners Apply Training Faster Than Others

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 26 May 2026 11:30:00 GMT

Most people working in L&D have probably experienced this at some point. You design a programme carefully. The learning objectives are clear. The facilitation is strong. Learners are engaged. Feedback is positive.And yet, once people return to work, the outcomes vary.

Some learners apply what they’ve learned quickly and confidently. Others, who seemed equally engaged during the training, struggle to turn learning into action. Not dramatically. But consistently enough to notice. That unevenness can be frustrating because it often feels difficult to explain.

Our first instinct is usually to look at the training itself. Was the design strong enough? Did the learners get enough practice? Was the content too complex? Those questions absolutely matter. But even when we get many of those things broadly right, the pattern often remains.

Which raises an interesting possibility. What if part of the difference sits not inside the training programme itself, but in how learners approach learning once real work begins?

The Part We Rarely Talk About

Research into workplace learning has repeatedly shown that learner characteristics play an important role in whether learning transfers successfully into workplace performance.

Things like how:

learners respond to setbacks
they build confidence
they monitor their own progress
they turn intentions into action under pressure

This isn’t about labelling learners as “good” or “bad.” Nor is it about blaming learners when transfer doesn’t happen. It’s simply recognising that some people arrive with stronger workplace learning habits than others. The interesting part is that many of those habits are teachable.

Which creates an important opportunity for L&D.

Perhaps part of our role is not just helping people learn a specific skill, but helping them become more effective at using learning in the workplace.

Confidence Often Comes After Action

One of the strongest predictors of workplace application is something called self-efficacy — a learner’s belief that they can successfully perform a task in real conditions.

Importantly, this isn’t abstract confidence. It’s highly contextual.

Can I still do this when:

I’m under time pressure?
things don’t go smoothly?
the stakes feel higher?
someone is watching?

Learners with lower self-efficacy rarely opt out consciously. More often, they hesitate. They delay using the new skill. They quietly fall back into familiar habits.

What’s interesting is that confidence usually grows through successful use, not through explanation alone. That has practical implications for L&D. Rather than encouraging learners to “feel confident first,” we can help them:

break a skill into smaller usable steps
identify low-risk opportunities to apply it quickly
repeat those actions deliberately, even imperfectly

Sometimes the smallest successful action is what starts building real momentum.

Helping Learners Manage Their Own Learning

Another important factor is what psychologists often call self-regulated learning. In simple terms, this is a learner’s ability to manage their own learning process once formal support reduces. Effective workplace learners tend to:

plan how they will apply learning
notice when something isn’t working
make adjustments
reflect briefly and purposefully

Less effective learners often wait for:

more instruction
more clarity
more confidence
more time

Unfortunately, modern workplaces rarely provide perfect conditions for any of those things. The encouraging part is that this kind of learner self-management does not need to be taught in a complicated way. Even a lightweight three-step loop can help:

What will I try next time?
How will I know if it’s working?
What will I adjust if it isn’t?

Small reflective prompts like these can significantly increase learner ownership and application.

Good Intentions Are Surprisingly Fragile

Most learners leave training with positive intentions. The problem is that good intentions often collapse under the pressure of real work. Not because learners don’t care. But because workplaces are busy, cognitively demanding environments. People forget. Priorities shift. Familiar habits return.

One surprisingly effective technique for bridging this gap is something called if-then planning. Instead of vague intentions like: “I’ll try to use this more", learners create clear situation-action links such as: “If I receive a difficult customer complaint, then I’ll use the three-step response structure before replying.”

This works because it reduces the mental effort required in the moment. The situation itself becomes the reminder. For L&D, this creates a simple but practical opportunity: help learners leave training with a few realistic workplace application plans rather than just good intentions.

Supporting Recovery, Not Perfection

Early attempts at applying learning rarely go perfectly. Some learners interpret that as:

“I’m obviously not very good at this.” Others interpret it as: “That approach didn’t work particularly well — what can I adjust next time?”

That difference matters enormously because it shapes whether learners persist or quietly disengage. As L&D professionals, we can help normalise imperfect early application by encouraging learners to ask:

What part of this was under my control?
What would I change next time?
What support or resource would help?

This shifts the focus away from personal failure and towards gradual improvement.

A Small Shift That Can Have a Big Impact

None of this replaces good instructional design. Design quality still matters. Practice still matters. Manager support still matters.

But increasingly, workplace performance also depends on whether learners know how to use learning effectively once they are back in the flow of real work.

That may be one of the most overlooked parts of workplace learning.

The encouraging part is that many of these learner habits are highly teachable. And once learners begin building them, every future learning experience has a better chance of translating into meaningful workplace performance.

A fuller exploration of these ideas — including the psychology behind learner confidence, self-regulation and workplace transfer — can be found in the original Learning Re-Framed article here.

Work with the Brain - But Don't Stop There...

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 19 May 2026 11:30:01 GMT

Over the last few years, neuroscience has become increasingly influential in the world of learning and development. Concepts around attention, memory, cognitive load and emotion now appear regularly in conversations about learning design — and with good reason.

One of the first neuroscience-related concepts that really stuck with me was the role of the amygdala in how we respond to challenge and uncertainty. Michael W. Allen references this in his book Designing Successful E-Learning, explaining how the brain can trigger a kind of cognitive “flight mode” when we encounter something stressful, difficult or emotionally uncomfortable.

Most of us have experienced this in learning situations.

You reach a moment in a course, workshop or e-learning module where something suddenly feels difficult, awkward or mentally draining — and you quietly disengage. Your attention drifts. You stop processing properly.

This matters because learning design doesn’t happen in a vacuum. It happens within the constraints of how human brains actually work. And when you start exploring books like John Medina’s Brain Rules, that reality becomes even clearer. Medina highlights several principles that feel especially relevant to anyone working in L&D:

The brain doesn’t pay attention to boring things
Attention is limited and selective
Stress interferes with thinking and memory
Multitasking reduces effectiveness
Emotion strongly influences what we notice and remember

The practical implication is obvious enough: if learning design ignores how people actually process information, outcomes become unpredictable. Which is why many good learning design practices already compensate for these realities. We:

chunk information into manageable sections
repeat important ideas
simplify complex tasks
use prompts, scaffolds and job aids
reduce unnecessary cognitive overload

All sensible and valuable approaches. But I think there’s an interesting tension hidden inside all this. Because taken too far, the “work with the brain” perspective can unintentionally become limiting. It risks framing learners primarily in terms of their constraints and weaknesses.

Anyone who has worked in L&D for a while will probably recognise the following situation. Two people attend the same programme. They receive the same content, the same practice and the same support.

Yet what happens afterwards is often very different. One person experiments, adapts, recovers from mistakes and gradually improves in the workplace. The other hesitates, delays, loses confidence or reverts to old habits under pressure.

We often explain this through motivation, personality or mindset — and all of those factors matter. But I suspect there may also be something slightly deeper involved: what we might call learning capability.

This is about the ability to:

act despite uncertainty
recover when things don’t go smoothly
monitor whether something is working
adapt rather than abandon

And this raises an interesting possibility for L&D. What if understanding how the brain works isn’t just about designing around human limitations? What if part of our role is also helping people become better at managing those limitations themselves?

That doesn’t mean ignoring cognitive science or pretending attention and memory don’t matter. Quite the opposite. It means building capability on top of that understanding.

Because perhaps the long-term goal isn’t simply to design learning that compensates for fragile attention, unreliable memory and uncertainty. Perhaps it’s also to help learners become more resilient, adaptive and self-aware when facing those realities in the workplace.

And that, I think, may be a slightly bigger and more interesting challenge for modern L&D.

A fuller exploration of these ideas — including the tension between cognitive constraints and learner capability — can be found in the original Learning Re-Framed article on Substack.

Why L&D Should Consider Learning to “Speak Visually”

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 12 May 2026 11:45:00 GMT

Most of us in L&D have had plenty of practice in using words to communicate clearly.

We know how to structure content. How to explain ideas. How to sequence information logically. How to write learning objectives, facilitator notes, e-learning copy and assessment questions.

But visual communication? That’s different.

Despite working in a profession that relies heavily on visuals — slides, graphics, scenarios, video, interfaces, layouts and e-learning interactions — very few of us have ever been taught how visual communication actually works.

And yet we instinctively know when something visual feels right.

We recognise strong visual storytelling immediately in films, television, photography, comics and advertising. We can usually tell when a slide deck looks polished or amateurish. We notice when an e-learning course feels visually polished and engaging versus visually flat.

The strange thing is this: most of us consume visual language fluently every day… while feeling far less confident about creating it ourselves.

Why This Matters for L&D

If you’ve ever:

struggled to make slides feel visually engaging
created visuals for scenario-based e-learning that felt a bit unnatural, or
found yourself thinking, “I know the look I want, but don't seem able to achieve it”

then you’ve already experienced this gap.

And historically, there’s been a practical problem sitting behind it. Even if someone in L&D developed stronger visual instincts, actually bringing ideas to life often required access to:

graphic designers
illustrators
photographers, or
video specialists.

Most small L&D teams simply didn’t have those resources available.

Why AI Changes the Equation

That resource barrier is suddenly much lower than it used to be. AI image generation tools now allow L&D professionals to create visuals that would previously have required specialist design support.

But there’s an important catch. The quality of the output depends heavily on how clearly you can describe what you want visually. And this is where many people hit frustration.

A prompt might feel perfectly logical in everyday language but still produce disappointing results because AI image tools “think” more like photographers, directors and visual storytellers than instructional designers.

They respond much more effectively when prompts include things like: framing, composition, camera angle, lighting, perspective and visual focus.

In other words, the more fluent you become in visual language, the better your AI-generated visuals are likely to become.

Learning the Vocabulary of Visual Communication

The good news is that this visual “grammar” is surprisingly learnable. You don’t need to become a professional film-maker or illustrator. Even a small amount of familiarity with visual storytelling principles can significantly improve your ability to:

direct AI tools more effectively
create stronger learning visuals
design more authentic scenarios, and
communicate ideas more clearly.

Two resources I’ve personally found useful are: Making Comics by Scott McCloud and The 5 C’s of Cinematography by Joseph V. Mascelli.

Neither was written specifically for L&D, yet both contain extremely practical ideas about visual storytelling, framing, composition and sequencing that transfer remarkably well into learning design and AI prompting.

And importantly, they help develop something many of us were never formally taught: the ability to think visually — not just verbally.

A Final Thought

For years, many of us in L&D have quietly accepted that visual design was something best left to specialists. AI is beginning to change that.

But making the most of these tools will require more than simply learning how to write prompts. It may also require us to become more fluent in the language of visual communication itself. And that feels like a genuinely useful new skill for modern L&D professionals to develop.

A fuller exploration of these ideas — including the connections between AI prompting, cinematography and visual storytelling — can be found in the original Learning Re-Framed article on Substack.

Why Learning Styles Feel Right — Even When They Miss the Point

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 06 Jan 2026 12:30:01 GMT

In my previous post, I explored why learning styles fail to stand up to scrutiny when examined through the lens of evidence. Despite decades of popularity, there is no reliable research showing that matching instruction to a learner’s preferred style leads to better learning outcomes.

But that only addresses part of the problem

A more interesting question is this: if learning styles are so poorly supported by evidence, why do they still feel so intuitively right to so many people in L&D?

Learning styles emerged as a genuine attempt to fix a real and persistent problem — but in doing so, they quietly pulled L&D’s attention in the wrong direction.

Why learning styles gained traction

Learning styles rose to prominence at a time when L&D was beginning to confront an uncomfortable reality: one-size-fits-all training wasn’t working.

Courses were generic. Learners were disengaged. Completion rates looked fine, but application and impact were questionable at best. Learning styles offered a compelling response:

People are different
Those differences matter
Learning shouldn’t look the same for everyone

At face value, that instinct was (and still is) entirely reasonable. The problem wasn’t the intent. It was how the idea evolved.

From design challenge to classification exercise

Learning styles transformed a complex design problem into a categorisation problem. Instead of asking, “What does someone need to be able to do?”, we started asking, “What type of learner is this?”.

That subtle shift matters. It moves the focus away from performance and toward cognitive processing preferences. In doing so, it anchors L&D firmly in the knowledge-acquisition phase of learning — not in the messy, pressured reality where learning needs to be applied.

Modality is not style

This is where a lot of confusion crept in. Research by Richard Mayer on multimedia learning — later synthesised and extended by Ruth Clark — demonstrated something important: how information is presented does matter.

But Mayer and Clark were never talking about learner preferences. They were talking about matching instructional design decisions to the:

nature of the content
type of task, and
limits of working memory

For example:

Visual-spatial tasks benefit from diagrams
Procedural tasks benefit from step-by-step visuals with minimal text
Language-heavy tasks are often better supported by text than audio

This is modality as a design principle — not a learner trait.

Unfortunately, modality was often simplified and misinterpreted as learning style. For example, dual channels - the idea that we process words and visuals through different mental pathways - became “visual vs auditory learners”. A design insight turned into a personal label.

That kind of leap was never supported by the research.

The self-diagnosis problem

There is another, deeper issue with learning styles — one that Ruth Clark addresses particularly clearly in her book Building Expertise.

Learning styles rely heavily on learners diagnosing their own needs and preferences. The problem with that? Research consistently shows that learners are very poor at self-diagnosis.

Across multiple studies, learners:

overestimate how well they have learned
confuse fluency with understanding
prefer easier, more comfortable instruction even when it produces worse outcomes
struggle to identify which instructional approaches actually help them perform better

In other words, people are not reliable judges of how they learn best. So, a model that encourages learners to self-diagnose and request a specific, instructional “prescription” rests on a very shaky foundation.

Why none of this matters when application is required

The massively important design insights of Mayer, Clark and others tell us lots about knowledge processing and recall but very little about what happens next.

When someone is back at work, under time pressure, trying to complete a task they:

don’t choose their preferred style
don’t reflect on presentation format
do what the task demands

In performance mode, preference disappears. What matters instead is:

Clarity
Relevance
Confidence
access to the right support at the right moment

Learning styles have little or nothing useful to say about any of that.

A more productive shift

If learning styles were an early attempt to personalise learning, the good news is that we now have far better, evidence-based ways of doing this.

Instead of asking, “What type of learner is this?” we get much further by asking:

What does good performance actually look like?
Where do people struggle when they try to apply this?
What decisions, judgements, or actions really matter?
What support would help most at that moment?

That shift moves L&D away from categorising learners and toward designing for performance.

Reframing the legacy of learning styles

Learning styles weren’t foolish. They were a signal. A signal that L&D wanted to move away from content dumping and toward something more thoughtful, more learner-centred and more effective.

But focusing too heavily on cognitive preferences distracted us from the harder — and more important — question, “How do we help people apply their learning better when it actually matters?”

Busting the Myth of Learning Styles

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 16 Dec 2025 14:00:01 GMT

Few ideas in Learning & Development have been as popular (or as persistent) as learning styles. Most of us are familiar with the basic idea: people have different styles of learning (‘visual’ or ‘auditory’ are couple of examples). Identify someone’s style, match the training to it, and ‘hey presto’ learning will be more effective.

You can understand the attraction. It’s an idea that seems deeply intuitive. That feels totally learner-centred. That is, in fact, irresistible to many learners and L&D professionals.

Which goes a long way to explaining why, for years, learning styles have been taught, assessed, and embedded into L&D practice. But there’s a big problem with this idea. Because when you look closely at the evidence, the idea simply doesn’t hold up.

Let’s be precise about what’s being challenged

The specific claim that research has examined and failed to support, is what psychologists call the meshing hypothesis. The idea that learners learn better when instruction is matched to their preferred learning style (e.g., visual learners taught visually, auditory learners taught verbally).

Now it’s important to note, that this is different from saying:

people don’t have preferences (they do),
or that variety is bad (it isn’t),
or that accessibility doesn’t matter (it absolutely does).

But back to that central question: does matching instruction to a learner’s declared style improve learning outcomes?

Fans of the learning styles idea are going to be disappointed. Because after more than two decades of research, the answer is still: no compelling evidence.

What the research actually says

One of the most widely cited reviews was published by Harold Pashler and colleagues in 2008. They set out very clearly what evidence would be required to support learning styles — and then examined the available studies.

Their conclusion was blunt: there was no credible evidence showing that matching instruction to learning styles improves learning.

Other reviews have reached the same conclusion. A very major UK review by Coffield et al. examined dozens of learning style models and found serious issues with reliability, validity, and practical usefulness.

As a result, learning styles are now commonly described in the research literature as a “neuromyth” — a belief about the brain or learning that sounds scientific but isn’t supported by robust evidence.

So much so that for years there has even been an open “Learning Styles Challenge”, offering a cash prize to anyone who can demonstrate real-world benefits from designing instruction around learning styles. To date, the prize remains unclaimed.

Why the idea refuses to die

If the evidence is so weak, why does the idea persist? Partly because learning styles play into something everyone is a fan of: personalisation. They give learners language to describe themselves and give designers a sense that they are tailoring learning rather than mass-producing it.

They’re also easy to explain, easy to assess, and easy to sell. But ease and intuitiveness are not the same thing as effectiveness.

What actually works better

Ironically, many people who believe in learning styles are already doing things that do work.

Research consistently shows that learning is more effective when:

the mode of instruction matches the nature of the content (for example, diagrams for spatial information or audio for pronunciation),
learners get practice, not just content exposure,
retrieval and feedback are built into practice
complexity is managed carefully,
learners are supported in applying learning in context.

In other words, the question isn’t, “What type of learner are you?” It’s, “What kind of learning does this task require?” That shift — from labelling learners to analysing tasks — is what helps L&D become more effective.

Preferences still matter, BUt...

None of this means learner preferences should be ignored. Preferences matter for all sorts of reasons like:

motivation
engagement,
inclusion and accessibility.

But preference is not the same as learning effectiveness. Giving learners choice in how they engage with content can be valuable. Designing everything around a fixed learning style is not.

A more helpful way to think about it?

If we really want to personalise learning, there are better levers to pull than learning styles:

design for the task, not the learning style label
design for application and performance, not just exposure to or consumption of content
design for context, not abstraction

Learning styles promised personalisation. The evidence suggests we can do better — by focusing less on categories and labels and more on what people actually need to do more effectively in the workplace.

Neuroscience Sounds Scary: Music's Impact on Focus Isn't

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 02 Dec 2025 21:30:01 GMT

If you want to make something sound serious and inaccessible, here’s a simple trick: talk about neuroscience.

I’m serious. Just roll the word around in your head for a minute. Neuroscience.

See what I mean? It feels scary. Impersonal. Something conducted in hermetically sealed labs where nobody smiles and every sentence (uttered by scientists in crisply-laundered white coats) must contain at least three syllables.

It feels like the kind of thing that can’t possibly relate to regular people or everyday jobs — let alone the practical world of Learning & Development.

(And just to be clear, I’m poking fun at the discipline of neuroscience here, not at the very real cognitive differences that many people live with.)

But then you open John Medina’s book Brain Rules, and suddenly your perception of neuroscience is completely flipped. His writing is energetic, engaging, and fun. Medina is that rare person that can bring even the driest-sounding of subjects or topics to life. And if you haven’t ever read his book, I’d strongly encourage you to.

One of the ideas that stayed with me long after reading Brain Rules was his discussion of music. Not in a “Playing Mozart to babies makes them genius-level smart” way, but in a grounded, evidence-based way.

And more recently, I happened across some research from MindLab, a UK-based cognitive science group, that reminded me just how powerful (and misunderstood) music can be when it comes to helping us concentrate.

So, if you've ever wondered why some playlists help get you into the flow while others knock you off track, this article explores the brain-friendly principles behind that difference.

So, let’s dive into understanding what music does — and doesn’t — do when you’re trying to get work done.

First: Music does not make you smarter

Medina is unambiguous about this.

No, Mozart won’t raise your IQ.
No, playing classical music to babies doesn’t give them a head start.
No, music is not a cognitive shortcut.

So, if you’ve been blaming your lack of genius on a childhood without Chopin… sorry.

What music can do is change your mental state

But what Medina does explore is how music can:

alter your mood and your energy and stress levels
shift your focus by making tasks feel more enjoyable
increase alertness if you’re drifting
create a rhythm that helps with repetitive tasks

In other words, music doesn’t make you smarter but it certainly puts you in the right frame of mind to think more smartly.

From the point of view of cognitive load, this is important. After, all, anything that reduces stress or stabilises your frame of mind frees up more bandwidth for meaningful work.

The MindLab research adds a useful layer

The MindLab research reminded me of all this. It ran a series of task-based studies involving music. It’s not deep neuroscience — but it is practical psychology, and it very much aligns with Medina’s findings.

Here’s what they found:

Different tasks benefit from different types of music

Classical is good for accuracy-based tasks and detail-heavy work
Ambient/electronic music is better for repetitive tasks, data entry
Dance music boosts speed, but sometimes reduces accuracy
Pop is okay for conceptual work (if lyrics aren’t distracting)

Music reduces stress, which improves performance

Participants consistently showed lower cortisol when listening to calming music. If your stress levels are lower, you attention levels are generally higher. Better attention levels generally lead to better results or output.

Lyrics are the enemy of language-heavy tasks

If your task requires reading, writing, or generating language…music with lyrics will probably slow you down. (This is why your “productive Spotify playlist” works brilliantly—until it doesn’t.)

A simple, practical takeaway

Most people in L&D will have moments in their working day when they have important jobs that just need to get done. Defining some objectives, designing a course outline, working up an e-learning interaction, to name but a few. So, can music help?

Well, I think the real takeaway here isn’t about genre, although clearly that can be a factor. It’s more about fit. Making sure there’s a good fit between:

task type
emotional state
personal musical preference

And really, that’s Medina in a nutshell: know how your brain works, and design for it. And sometimes, working out how your brain works can begin with something as simple as recognising why a particular playlist helps you slip into flow — and why another playlist derails your concentration completely.

A couple of final thoughts

Reading Medina’s work makes you realise that learning about neuroscience doesn’t have to be intimidating or abstract. Quite the opposite.

And in case you’re wondering? Yes, I wrote this while listening to music. Did it make me smarter? Sadly, not. But it did make the words you’ve just been reading flow more readily.

Brain Rules: 12 Principles for Surviving and Thriving at Work, Home and School
John J.Medina
Pear Press 2008
Brain rules.net

Designing at the Jagged Edge: What AI’s Uneven Abilities Mean for L&D

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 25 Nov 2025 10:00:01 GMT

In last week’s post, I wrote about the jagged edge — the uneven frontier between what AI is exceptionally good at and where it is still surprisingly weak.

That insight came into focus for me while standing on the patio of the Stahl House in the Hollywood Hills, looking down at the sharp, uneven ridges, set against the flat, orderly grid of Los Angeles below.

That contrast — smooth here, jagged there — feels like the perfect metaphor for AI right now. Not evenly distributed. Not predictable. Not consistent.

The takeaway from last week’s post was pretty clear: success with AI isn’t about mastering the technology. It’s about recognising where it shines and where it stumbles.

And that raises the next logical question.

What Does the Jagged Edge Look Like for L&D?

This is the part that matters. The practical, day-to-day reality of using AI in real L&D work.

Because if you work in L&D, you are very likely navigating that jagged edge — whether through generative content requests, experiments with course outlines, or colleagues asking, “Can we use this AI thing to help with…?”

Understanding where AI is strong (and where it is unreliable) is becoming a genuinely useful professional skill. So, here are some thoughts on what this looks like as we head towards the end of 2025

Where AI Really Helps L&D (The “Smooth” Side)

This is where the ground is smooth and firm and flat. Where AI consistently behaves well and adds value without fuss.

Summarising long, dense content

Policies, procedures, reports, transcripts — AI is excellent at turning them into:

short summaries
digestible bullet points
suggested learning objectives
concise recaps for learners

This isn’t just speculation. If you’ve used AI for this kind of task already you’ll know it’s pretty reliable.

Cleaning up messy SME input

We all love our SMEs; but we also know that a massive, unstructured ‘brain dump’ is often their default mode. However, if you give AI:

rambling emails
half-formed notes
slides covered in dense bullet points

it will make a good first pass at organising that content more clearly and logically. The output won’t be perfect. It will still need some extra human brain power to get it over the finish line; but it will get you to the end of the process much more quickly.

Drafting outlines, ideas and examples

AI is consistently strong at:

drafting module outlines
drafting simple scenarios
creating examples and non-examples
rewriting content at different reading levels

It certainly doesn’t replace creativity and original thinking; but it can really help unblock those creative juices when you are struggling to come up with good ideas.

Re-framing content for clarity

“Explain this as if…”

“Give me examples from…”

“Rewrite this in plain English.”

This is an area of real strength. In the right circumstances, AI can be just as great at giving you the last 10% (refining and polishing) as it at getting you 90% of the way there.

Where AI Still Struggles (The Jagged Edge)

These are the uneven ridges — the places where footing is uncertain and relying on AI is definitely risky.

Anything requiring organisational context

AI doesn’t know:

your policies
your culture
the shortcuts your learners actually use
the unofficial steps that matter
the messy real-world constraints

Its answers can sound right in theory, but can be very wrong in practical terms.

Emotional nuance, judgement or interpersonal dynamics

Handling conflict. Managing a difficult conversation. AI can mimic empathy but it often misses:

tone
boundaries
what not to say

Use it in these kinds of contexts with extreme care.

Multi-step reasoning

AI still struggles with:

multi-stage logic
conditional pathways
decisions that depend on context
keeping its own reasoning consistent

It can excel at steps 1–3 then completely misfire on step 4.

Highly specialised expertise

In specialised domains like medicine, engineering, law and compliance, plausible-sounding nonsense is especially dangerous.

Explaining the “why” behind a rule

AI is good at restating a rule. Much weaker at explaining the reasoning behind it. Learners need the “why” for authentic understanding — and AI often can’t supply it.

Navigating the Terrain

As noted in last week’s post, the jagged edge isn’t a barrier — it’s a map. In an L&D context, it really helps us understand where:

AI can accelerate design work, and
humans must remain firmly in the loop.

The real work right now is not avoiding the jagged edge but learning to navigate along it with confidence.

The Jagged Edge of AI: What It Means for Learning and Performance

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 18 Nov 2025 09:45:00 GMT

A few weeks ago, during a trip to the US, I was lucky enough to find myself standing on the patio of the Stahl House in the Hollywood Hills — a glass-and-steel architectural icon perched high above Los Angeles.

The view from the patio is breath-taking; and as a lover of mid-century minimalist architecture, I found the house equally stunning.

From that patio vantage point, the other striking aspect was the contrast between the hills immediately below the patio and the distant cityscape. Way below, the city stretched out in a perfect grid: flat, predictable, orderly. But right in front of me, the landscape fell away in uneven ridges — sharp, fractured, unpredictable.

I was reminded of the sharp, uneven edges of those hills the other day when I came across the term jagged edge - to describe what the world of AI looks like at the moment.

So, let’s explore the meaning of that term and what it might mean for L&D.

Often, we talk about AI as though it’s spreading evenly across everything; but it isn’t. It’s advancing in fits and starts — powerful in one moment, incredibly clumsy in the next. Researchers have named this unevenness the jagged edge of AI.

Understanding the Jagged Edge

You may well have experienced this jagged edge first hand, when AI is strong and accurate in some areas but weak and unreliable in others.

It can draft a pretty good course outline in seconds. It can generate accurate code that works. Yet still produce nonsense when asked to evaluate a moral dilemma.

This unevenness exists because AI doesn’t understand; it predicts. It excels where data is abundant and patterns are clear — structured writing, number crunching, summarising, and retrieval. It falls short where context, empathy or tacit knowledge contribute significantly to the outcome.

Evidence from the Research

This metaphor for uneven progress is backed by evidence. In 2023, researchers from Harvard Business School and Boston Consulting Group ran a field experiment exploring how AI affected knowledge workers. They called their study Navigating the Jagged Technological Frontier.

They found that workers using AI completed 12% more tasks and worked 25% faster when those tasks fell within AI’s frontier — areas where AI already performed well.

But when tasks sat just beyond that frontier, the results flipped: performance declined.
In other words, AI’s progress isn’t smooth or universal. It’s jagged. Full of peaks and troughs that shift depending on the nature of the work.

The Human–AI Division of Labour

At first glance, that jaggedness can seem like a flaw. But if you think of it more like a map or a bunch of helpful road signs — it usefully highlights where humans and machines each add the most value.

AI excels at automating structure: generating first drafts, summarising information, categorising, and identifying patterns.
We excel at interpreting complexity: spotting anomalies, understanding tone, making ethical calls, and connecting dots in ambiguous situations.

When we understand where the jagged edge runs, we can design smarter systems and workflows that let each party play to its strengths.

What This Means for L&D

For L&D professionals, that insight is particularly useful. It tells us the real opportunity isn’t just using AI generically. It’s about recognising where AI’s edge lies in our own jobs — what can be safely automated, and what still needs human input and decision-making.

For example:

A learning designer can use AI to structure a skills framework (below the edge), but must still apply contextual awareness to fit it to their own organisation (beyond the edge).
A subject matter expert might use AI to generate examples or case studies (below the edge), but they’ll need to refine and validate accuracy, richness and complexity of those examples (beyond the edge).

It’s going to take us all a while to get used to this ‘new frontier’; and we’ll need to be ready for the jagged edge to shift as AI technology evolves. But recognising and navigating the jagged edge confidently, within our own job roles, could massively transform our productivity and effectiveness.

Reference:
Dell’Acqua, F., Kalliamvakou, E., Kirov, S., Ransbotham, S., & Rock, D. (2023). Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality. Harvard Business School Working Paper 24-013.

Designing Training That Builds Performance Support Habits

andrew@pacificblue.co.uk (Andrew Jackson) — Tue, 11 Nov 2025 09:30:01 GMT

I’ve written a lot over the last few months about the significance of shifting from a training-focused paradigm to a performance-support focused one. It’s a big shift. Understandably, plenty of people feel like this is me saying, ‘do away with your training events.’

Why training still matters — but needs to evolve

Definitely not the case. But it does raise a big question that people often ask:

"What would training look like if it were designed with performance support in mind from the start?"

Let’s dig into answering that, to see how you might refine what you do during a training event, when designing for a more performance-support focused paradigm.

It’s easy to see performance support as a nice ‘add-on’. Something to highlight to the learners towards the end of the event, as a rather perfunctory ‘transition to the workplace’ segment. I confess, I’ve run my fair share of such segments over the years.

Unfortunately, this approach will pretty much guarantee that your performance support solution gets little use or has low perceived value.

From event thinking to performance thinking

So, if you are providing performance support tools or materials, then they need to be introduced and used as an integral part of the training – not as an afterthought.

The question remains: 'How to do that?'

In our impact and instructional design programme, we introduce attendees to a very simple, very flexible but nevertheless very robust design flow that works very well during in-person and live online events.

It’s a three part design flow. It starts with presenting to learners, moves to providing structured practice, and finished with much less structured extended practice. The idea being that as you go through this sequence, you are helping the learners’ capability and self-direction to increase

From a learning design point of view, this is about providing multiple points of practice that is highly interactivity. Learners gradually become more autonomous and increasingly confident in their ability to apply what they are learning.

Ideally, by the time you reach the extended practice phase of a session, the trainer will be able to fully step back, observe the learners and only provide feedback (where needed) during a group de-brief after the activity.

The good news is that from a performance support perspective, this is a very effective design flow, too. The difference comes in how you think about and design your practice activities.

What a performance-support-focused training session looks like

So, from a learning design perspective, a good flow of practice activities will help the learners to process and embed what they have learnt; the only ‘external’ assistance comes from the trainer’s input during the structured practice stage and any feedback or de-brief after the extended practice stage.

From a performance support perspective, the structured and extended practice activity stages of the flow are the perfect moments to enhance that trainer input and support by introducing and phasing in use of the performance support tools or materials.

This familiarises learners with the tools and how to use them; and positions them as the natural post-training ‘go-to’ which will help to get the job done.

Bringing performance support into the classroom

For example, imagine a training event that is focused on helping learners to plan and prepare annual performance reviews with their team members.

The course design analysis phase highlighted the benefit of providing a simple performance support tool to guide learners through the planning and preparation.

Having learners use the planning tool as part of any practice activity makes the practice much more authentic; and, ensures the learners see the tool as part and parcel of how to apply what they are learning.

Training as the gateway, not the goal

All this highlights clearly that selecting and designing performance support tools or materials must be an integral part of your existing impact and instructional design process. Not an afterthought. The training is vital but only the first stage of a journey to the really important goal: improved workplace outcomes.

Achieving this level of performance support integration requires L&D to get close to the business. To have a good understanding of what desired workplace performance looks like. To be clear about how knowledge and skills covered during a training event are applied in the learners’ workflow, once they are back on the job.

If that sounds a bit challenging, it shouldn't. In reality, performance-support-focused training isn’t a different discipline — it’s simply the next logical step in doing what great L&D has always aimed to do: help people do their jobs more effectively.