The Intentional Learning Framework
Specification Papers
The independent standard for good learning design - built for your team, your system, and your AI to work from.
Seven papers. More than 500 pages.
A decade of hands-on research, encoded so the people who build and choose learning technology - and the AI inside it - can design learning that works on purpose, not by accident.
What the 'ILF Specification Papers' are (the quick answer)
The full, independent standard for good learning design, enrolment to completion - seven papers, 500+ pages, written for your team and the AI in your product to build from directly.
If you build learning technology → Benchmark your platform against the standard, design the gaps out, and ship AI that reasons from real expertise instead of imitating its surface (proof of the expertise behind your product).
If you run education → An independent standard to hold any tool or AI against before it reaches your learners, and to steer your own builds so good practice is the default, not a lottery.
Your system was never built to help people learn.
Now AI is filling that gap.
Think about what a learning management system does by default. It enrols, tracks, records and reports. Every one of those things is done to a learner, or for them - none is the same as helping someone actually learn. That gap has always been there, and good learning design quietly falls through it.
Not every educator is a learning designer, yet most are asked to do design work that was never their craft, with no time to do it - so they fill the gaps with guesswork, and good practice fails to make it into the room. And learning designers say they are often hampered by system constraints.
Now AI fills the same gap, faster, and at scale, behind a polished surface that hides what's missing. Point it at "build the course" and it won't tell you where its expertise ends. Used poorly, it scales the weaknesses we already had.
The cost isn't abstract. Learners disengage and quietly leave, often before the teaching even starts. Educators carry the remediation. And an organisation running a tool built on none of this inherits the weakness at scale, with its name on it. In my own doctoral research, this gap was universal: every interviewed educator reported a gap between the practice they believe in and the practice they manage to deliver. The barrier is rarely will - it's time, workload, and a craft that shifts with every context.
Everything behind the framework, made buildable
The Intentional Learning Framework is the map. These Specification Papers are everything behind it: a decade of hands-on research with real educators and adult learners across vocational education, higher education and workplace training, turned into the tested strategies, exemplars, pitfalls and expert-grounded prompts that actually move outcomes.
Seven papers, more than 500 pages written so your build team and your AI can work from them directly - not a stack of academic reading to interpret, but the research and the why beneath it.
The point is to reason from good practice, not imitate it. A system that imitates reproduces the surface: it looks right, it sounds like a good educator - but point it at a situation nobody scripted and it has nothing underneath to draw on. A system that reasons from good practice can make the right call in a moment the manual never named. That distinction is the whole product.
And good design doesn't mean keeping everyone on the LMS screen. Sometimes the best thing a well-designed system can do is send the learner away from the LMS itself - trigger the pen-and-paper retrieval task, prompt the practice that has to happen offline. These papers specify good practice wherever it belongs, in the system and out of it. A tool that measures success by screen time has already misunderstood the job.
Built to be built from
Every paper follows the same spine, so your team always knows where to look:
-
What the research shows, and the why beneath it, so a decision can be reasoned, not guessed.
-
The tested strategies, the concrete changes that move the numbers.
-
Common pitfalls, named directly, including the ones AI systems reliably fall into.
-
What good looks like, written to the team doing the building, and their AI.
-
Exemplars and templates, fresh and genericised, ready to apply.
-
Expert-grounded AI prompts and diagnostics built as components to run inside your product under human oversight - not text for a chat window.
-
A dual checklist: a quick test your team can apply, and discrete rules your system can be built on.
That last item is one artifact serving two readers at once. The same checklist is a build spec for the team making a system, and an evaluation test for the organisation choosing one.
Not the soft option. The measurable one.
"Better learning design" can sound soft - something nice to have when there's time. So here are some numbers. Every strategy in the Framework goes through trials, control groups and qualitative validation before it earns its place.
-
One strategy raised first-unit completion from 53% to 84%.
-
One authentic rewrite of a single assessment question lifted correct responses from 17% to 88%, same learners, same day. In focus groups, learners rated the rewrite as harder, not easier.
-
One reframe of a single discussion task took participation from around 5% to 91%, same educator, same cohort, same topic.
-
One strategy drove assessment resubmissions down an average of 72% across nine sites.
Every one of those numbers is also a cost. A discussion nobody joins is a learning activity you paid to build and nobody used. Resubmissions are staff hours. Non-completions are lost outcomes, lost learners, and lost income. A learner who can't show what they know is a qualification that means less than it should.
Sit with the second result, because it's the thing the quick-fix tools get backwards. Good design is not the same as 'easy'. The rewritten question was reported by learners as more demanding, yet it unlocked capability the original wording had suppressed. That's productive friction: the right difficulty in the right place. A tool that chases engagement by stripping out effort is optimising for the wrong thing.
These are examples, not a scoreboard. Good learning design isn't the soft option. It's the measurable one, and it's exactly the judgement the Specification Papers encode.
From enrolment to completion, not just the teaching moments, not just the learning content
Most learning design obsesses over the teaching and the assessment, as though those are the only parts that count. But by the time a learner reaches the teaching, you've often already lost the ones who couldn't find the course, never quite started, or drifted away in the first week. The drop-off happens long before the part everyone designs for.
When you hand that whole journey to a system - or the AI inside it - it needs to understand every stage, not just the two in the middle. Otherwise it optimises the teaching and quietly loses everyone who never got that far.
The Intentional Foundations paper sits underneath all of it: the principles every stage is built on, including how a system decides what to build, what to help an educator do well, and what stays human craft it can only point toward.
Then the six stages:
-
Explore +Enrol. They find you, weigh it up, and decide it's worth their time.
-
Welcome +Orient. They arrive, feel genuinely welcomed, and find their feet.
-
Roadmap +Expectations. They know where they're going, what's expected, and why it matters.
-
Active Learning. They do the work, build the skills, and stay engaged.
-
Assessment +Feedback. They show what they can do, and learn from how they went.
-
Closure +Next Steps. They finish well, and know what comes next.
Seven papers in all: the six stages, plus Foundations. Each stands on its own, and each is sharper for the ones around it.
For technology teams and AI developers
You can build a product. What you can't reverse-engineer is a decade of sector-specific learning research to deliver a system models real expertise.
The Specification Papers give you two ways to close that gap:
-
benchmark your platform against the standard and design the gaps out, so good practice becomes the default your users work in; and
-
build new AI-powered features that reason from best practice rather than imitate it.
Buyers are increasingly asking suppliers to show the expertise behind their outputs. A tool that looks useful but undermines learning is a risk you don't want surfacing during or after the sale.
Licensed for your build, the papers become your standard.
Be the product with the expertise behind the surface, not the shiny one that only looks like it.
For education providers
AI is already in your teaching and assessment, and the various systems you run were built to manage learners, not to help them learn. You're accountable for whether what you adopt is good for learning, and "it came from a supplier" was never a defence here.
Adopt the wrong tool and the harm lands on your learners (and your name).
The Specification Papers give you an independent standard for what good learning design looks like, so you can hold any tool or AI product against it before you let it near your learners. Empower your system to steer your educators toward good practice instead of leaving it to chance.
However your design capability is set up (no dedicated designers, too few, or a whole team not yet working from one playbook) the standard holds every build to the same bar, so good practice is consistent across every course, not just the ones a specialist had time to touch.
Durable because it's technology-independent
The papers specify what a system must do, and why, never how to build it in any particular product. They name capability and behaviour, not tools, vendors or features. That's why they age well as your stack evolves, and it's the same discipline the papers ask of you: decide what good learning looks like first, then choose the technology to serve it. Never the reverse. We practise what we specify.
AI is arriving in your system whether you plan for it or not. The day it does, it needs to know what 'good' looks like, or it will scale what doesn't. AI scales whatever you give it - so give it good practice. This is the layer you install before the AI, not after.
Access is perpetual. What you license, you keep.
Research you can build on
You're trusting these papers to shape what your product or your organisation does with learners, so it's fair to ask who stands behind them.
Dr Deniese Cox holds a PhD in online education and a Masters in learning and development, is an award-winning researcher, and is a top-rated presenter with an independently verified rating of 4.8 out of 5 across hundreds of participant reviews.
She has also been a university lecturer (teaching online pedagogy at bachelors level, and strategic learning design at masters level), a VET trainer, and a workplace educator, so the research is grounded in also knowing first-hand what it's like to juggle those roles and still fight for the best outcomes for learners.
"I do the hands-on research, so your team doesn't have to".
The rest of the ILF family
The papers are the foundation of the ILF family - and the layer you build everything else from. If you'd rather not build by hand, the same standard comes closer to done-for-you:
-
ILF Intelligence Library (integration-ready) - pre-built, governed SKILL.md file packages derived from these papers, so your agents reason from and apply good practice without authoring the method from scratch. Available two ways: purchase outright (ZIP files, self-hosted, ideal if you have in-house technical capability), or MCP Access via subscription (hosted, ready to go, no development team needed).
These papers make AI good for learning.
There's a companion for making it defensible.
The ILF Specification Papers are one half of adopting AI you can stand behind: the half that makes it genuinely good for learning. The other half is making it defensible inside a heavily regulated sector - and that's the work of the wider Expert-Grounded AI initiative (EGAI), run alongside Phill Bevan.
Through EGAI you can also reach:
-
The AI Adoption Specification Paper (130+ pages, updated as regulations evolve)
-
The AI Adoption Toolkit with 30+ templates, tools, and resources to assist with effective AI adoption in your organisation.
-
The AI Adoption Intelligence Library - SKILL.md packs that translate the specification paper into structured instructions, prompts, decision rules and quality checks for AI agents, applications and internal build teams.
Informing the human. Powering the AI.
Find the right fit
The papers are licensed by organisation type and scale, with separate tiers for technology providers and education providers. Tell us where you sit and what you're building or running, and we'll find the right fit.
Build the system your learners needed all along: one that helps them learn, not just one that manages them.