Output: A personal experiment plan with checks and boundaries.
Before this: No technical background required. Start at the beginning.
Work through the fictional examples and feedback. Short notes are enough.
LecXie suggested study allocations, not provider runtimes or verified learning time. Optional deeper work is outside the total.
0 / 6
Stops marked explored
Off by default. Notes stay in this tab while navigating Learning; reload or close loses them unless exported. Nothing is sent to a learner database. Avoid personal or sensitive information.
Stop 1 of 6
Name the kind of AI
Use precise language about a system before choosing it.
Journey fit: Start with the prerequisite concepts above. The selected resource supplies context; the local case is where you use it.
Selected resource · AI, Machine Learning, Deep Learning and Generative AI Explained
IBM Technology
AI, Machine Learning, Deep Learning and Generative AI Explained
What you will learn: Distinguish the broad field of AI from machine learning, deep learning and generative systems.
Why this is worth the time: Introduces the vocabulary before models, prompts and evaluation.
Before this
None; introductory
Study time
Suggested 15-minute study activity; exact video runtime unverified
Access
Free public video; platform conditions apply
Currency / limits
Published 5 Aug 2024; introductory terminology, not current performance evidence
Loads a third-party player only when selected. No autoplay. If the provider blocks playback, use the original link below.
Self-reported and shared across routes. Source date: 2024-08-05 · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Original local activity · Fictional case
Use precise language about a system before choosing it.
A supplier calls a fixed rule that routes expense claims “generative AI”. What would establish that description?
Apply it to a working example
Use a hypothetical or approved non-sensitive example. Describe one system’s input, output and mechanism without using the word intelligent.
A self-report of exploration, not verified proficiency. Your work is not sent to LecXie.
Stop 2 of 6
Prediction and verification
Separate fluent language from evidence that a statement is true.
Journey fit: Build on Name the kind of AI before this activity. The selected resource supplies context; the local case is where you use it.
Selected resource · Large Language Models explained briefly
3Blue1Brown · Grant Sanderson
Large Language Models explained briefly
What you will learn: Explain how language generation differs from knowing that a claim is true.
Why this is worth the time: A visual starting point for prediction and verification. Use the OpenLearn technical sections instead if you prefer text.
Before this
None; no coding needed
Study time
Suggested 15-minute study activity; video runtime not asserted
Access
Free public visual explainer; no account needed to read
Currency / limits
20 Nov 2024; mechanism primer, not a current model comparison
Loads a third-party player only when selected. No autoplay. If the provider blocks playback, use the original link below.
Self-reported and shared across routes. Source date: 2024-11-20 · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Original local activity · Fictional case
Separate fluent language from evidence that a statement is true.
An assistant produces a plausible reference that you cannot locate. What is the useful next action?
Apply it to a working example
Use a hypothetical or approved non-sensitive example. Name a claim and the originating source you would use to check it.
A self-report of exploration, not verified proficiency. Your work is not sent to LecXie.
Stop 3 of 6
Test on a separate example
Understand why a system can fit examples without generalising.
Journey fit: Build on Prediction and verification before this activity. The selected resource supplies context; the local case is where you use it.
Selected resource · Navigating the Jagged Technological Frontier
Harvard Business School AI Institute
Navigating the Jagged Technological Frontier
What you will learn: Ask which task and comparison a productivity claim actually tested.
Why this is worth the time: Introduces uneven task performance before you judge evidence for your own work.
Before this
Understand the basic idea of an AI model
Study time
15 min selected reading
Access
Free public research summary; no account needed
Currency / limits
21 Sep 2023; historical consultant study, not a 2026 benchmark
Self-reported and shared across routes. Source date: 2023-09-21 · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Original local activity · Fictional case
Understand why a system can fit examples without generalising.
A classifier matches every training example. Does that establish performance on unseen cases?
Apply it to a working example
Use a hypothetical or approved non-sensitive example. Record what changed in training versus test loss and one limitation of this toy.
A self-report of exploration, not verified proficiency. Your work is not sent to LecXie.
Stop 4 of 6
Choose safe information
Recognise when to stop and seek approval rather than paste data into a tool.
Journey fit: Build on Test on a separate example before this activity. The selected resource supplies context; the local case is where you use it.
Selected resource · AI Risk Management Framework
NIST
AI Risk Management Framework
What you will learn: Turn a vague risk concern into ownership, context, measurement and controls.
Why this is worth the time: Provides a general risk lens. It does not replace local controls or legal advice.
Before this
A candidate task and people affected by it
Study time
Suggested 25-minute exploration of Govern, Map, Measure and Manage; not a provider runtime
Access
Free public framework; no account needed
Currency / limits
AI RMF 1.0 (2023) is being revised; check current guidance before use
Self-reported and shared across routes. Source date: 2023-01-26 · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Selected resource · Guidance on AI and data protection
Information Commissioner’s Office · UK
Guidance on AI and data protection
What you will learn: Identify data and accountability questions to take to the responsible specialist.
Why this is worth the time: Adds a UK data-protection lens to the general risk framework. Relevance depends on jurisdiction.
Before this
A candidate task and its information flows
Study time
15 min selected guidance
Access
Free public UK guidance; no account needed
Currency / limits
Under review after the Data (Use and Access) Act; UK-specific
Self-reported and shared across routes. Source date: not established · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Original local activity · Fictional case
Recognise when to stop and seek approval rather than paste data into a tool.
You are trying an unapproved public AI tool. Which input is suitable for an initial learning exercise?
Apply it to a working example
Use a hypothetical or approved non-sensitive example. List what must stay out of the tool and who can approve a real use.
A self-report of exploration, not verified proficiency. Your work is not sent to LecXie.
Stop 5 of 6
Write a brief a person can check
Make an AI-assisted task specific enough to review.
Journey fit: Build on Choose safe information before this activity. The selected resource supplies context; the local case is where you use it.
Selected resource · Lesson 2B: The 4D Framework
Anthropic · Rick Dakan & Joseph Feller
Lesson 2B: The 4D Framework
What you will learn: Use delegation, description, discernment and diligence to plan a human–AI task.
Why this is worth the time: A practical bridge from understanding a model to describing a useful task and taking responsibility for the result.
Before this
Basic model limits
Study time
Suggested 15-minute reflection; runtime not verified
Access
Free video; no Claude account to watch; full-course access not verified
Currency / limits
Published 12 Jun 2025; AI-fluency framework, not an independent outcome evaluation
Loads a third-party player only when selected. No autoplay. If the provider blocks playback, use the original link below.
Self-reported and shared across routes. Source date: 2025-06-12 · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Selected resource · User Needs + Defining Success
Google PAIR
User Needs + Defining Success
What you will learn: Identify a user problem, desired outcome and whether AI adds value.
Why this is worth the time: Makes choosing a useful application more concrete; use it before choosing automation simply because it is possible.
Before this
A candidate task or user problem
Study time
Suggested 20-minute study activity
Access
Free public text; no account needed
Currency / limits
Foundational guidebook originating in 2019; page undated
Self-reported and shared across routes. Source date: not established · Last update: not established · Checked 2026-10-15. Part of this stop’s selected reading.
Original local activity · Fictional case
Make an AI-assisted task specific enough to review.
Turn “write something useful about AI” into a bounded brief. Which improvement gives a reviewer a useful test?
Apply it to a working example
Use a hypothetical or approved non-sensitive example. Write a short brief with audience, permitted inputs, output and two checks.
A self-report of exploration, not verified proficiency. Your work is not sent to LecXie.
Stop 6 of 6
Plan one bounded experiment
Use your earlier notes to propose a small, reversible test.
Journey fit: Build on Write a brief a person can check before this activity. The selected resource supplies context; the local case is where you use it.
Selected resource · Impact Evaluation of AI Interventions
HM Treasury · Evaluation Task Force
Impact Evaluation of AI Interventions
What you will learn: Specify a baseline, comparison and outcome measures before starting a pilot.
Why this is worth the time: Connects the earlier notes to a testable pilot rather than a demonstration.
Before this
Task, safeguards, workflow and capability notes from stops 1–5
Study time
20 min selected sections; full guidance takes longer
Access
Free public guidance; no account needed
Currency / limits
Updated 15 May 2026; UK public-sector guidance, adapt to context
Self-reported and shared across routes. Source date: not established · Last update: 2026-05-15 · Checked 2026-10-15. Part of this stop’s selected reading.
Original local activity · Fictional case
Use your earlier notes to propose a small, reversible test.
A demonstration looked impressive. What would make the next trial informative?
Apply it to a working example
Use a hypothetical or approved non-sensitive example. Propose one task, a baseline, comparison, responsible person and stop/review rule.
A self-report of exploration, not verified proficiency. Your work is not sent to LecXie.
A personal experiment plan with checks and boundaries.
Assembled from your working notes. Empty sections remain prompts. This is a draft for discussion with the responsible people.
Optional deeper branches
Separate from the core route. Alternatives cover a similar outcome; deeper work extends it. Choose deliberately rather than repeating everything.
Explore generalisation in a visual simulation. Alternative to the local training/test toy, not a compulsory extra. Before this: Test on a separate example.
