Level 3 Certificate in Advanced AI
Build, buy and lead AI responsibly for a team or organisation.
What it’s for
For people who build, buy or lead the use of AI for others. You learn how models are made and why they fail, how to test AI before you trust it, how automation and agents work and fail, and how to handle data protection, security, governance and adoption across a team.
Who it’s for
- You lead a team or a project that uses AI
- You choose, buy or set up AI tools for other people
- You build automations or apps with AI
- You’ve finished AI in Practice, or already know what it covers
Before you start: Regular, confident use of AI. AI in Practice (Level 2) covers what you need.
What you’ll be able to do
- Explain how large language models are built and why they fail
- Design and run an evaluation before trusting an AI tool
- Plan automations and agents with the right controls
- Apply UK data protection and security principles to AI
- Set up governance, and lead a team through adoption
The units
Each unit lists what you’ll learn (its learning outcomes) and what you’ll be able to do to show it (its assessment criteria). Every marked question is tied to one of them.
1Inside the modelsPretraining, fine-tuning and human feedback; tokens, context and reasoning; open and closed models; and why models fail.13 min
Outcome 1Understand how large language models are built
- Describe pretraining, fine-tuning and learning from human feedback
- Explain tokens, context windows and reasoning modes
- Compare open-weight and closed models
Outcome 2Understand why models fail
- Explain hallucination, sycophancy and bias
- Explain retrieval-augmented generation and when it helps
2Evaluating AI: test before you trustDefining success, building a test set, comparing fairly, and keeping watch when models and prompts change.13 min
Outcome 1Design an evaluation
- Define what success looks like for a task
- Build a test set that reflects real use, including hard cases
- Use rubrics and blind comparisons
Outcome 2Run evaluations and act on them
- Measure accuracy and consistency
- Analyse failures to find their causes
- Test again when models or prompts change
3Automation and agentsChains, workflows and agents; tools and connectors; where people approve; and the risks, from prompt injection to runaway actions.14 min
Outcome 1Understand AI workflows and agents
- Distinguish prompt chains, fixed workflows and agents
- Describe how tools and connectors give AI access to other systems
- Decide where a person must approve an action
Outcome 2Manage the risks of automation
- Explain prompt injection
- Apply least privilege to AI tools
- Plan for failures, monitoring and a way to stop
4Data protection and securityApplying UK GDPR to AI, checking suppliers properly, and closing the common ways data leaks.14 min
Outcome 1Apply UK data protection principles to AI
- Identify personal data in an AI use
- Explain lawful basis, data minimisation and retention
- Know when a data protection impact assessment is needed
Outcome 2Assess AI suppliers and security
- Check a supplier’s training defaults, retention and processing terms
- Require access controls such as single sign-on and audit logs
- Reduce the risk of data leaking through AI tools
5Governance, risk and the lawAn AI policy and register that people use, proportionate risk assessment, and the legal landscape in the UK and EU.13 min
Outcome 1Set up AI governance
- Write an AI policy people can follow
- Keep a register of AI uses, each with an owner
- Assess risk in proportion to the use
Outcome 2Understand the legal landscape
- Describe the UK’s approach and the role of regulators
- Describe the EU AI Act’s risk-based approach and when it can reach UK organisations
- Recognise copyright and equality risks
6Leading AI adoptionChoosing use cases, running pilots that prove something, buying or building, and bringing people with you.13 min
Outcome 1Choose and prioritise use cases
- Score use cases by value and risk
- Run a pilot with clear measures
- Decide whether to buy or build
Outcome 2Lead people through change
- Plan training and support
- Talk openly about AI and jobs
- Track costs, benefits and environmental impact
How it’s assessed
Each unit ends with a 5-question unit check: get 4 right to complete the unit, and retake it as often as you need with fresh questions. When all six are complete, the final assessment opens: 24 multiple-choice questions, 4 from each unit, in 40 minutes. Your grade comes from your final assessment mark.
Your certificate
Your name, the level and your grade, with its own number and a QR code linking to a public page where anyone can check it. A second page records the units and learning outcomes you’ve met. It arrives as a PDF by email, you can add it to LinkedIn in one tap, and we’ll post you a printed copy for £10 if you’d like one.
It’s an Unjargoned certificate of achievement, not a regulated qualification.