Understand
Start with the key ideas, limits and evidence behind the decision.
Evidence-led AI guidance
Use practical guides to evaluate tools, protect information, verify output and build workflows that remain accountable to people.
Read, apply, review
Start with the key ideas, limits and evidence behind the decision.
Use the interactive checklist to translate the guidance into your own workflow.
Follow related tools, recipes and learning only when the next step is justified.
Guide library
Search by question or filter by the kind of decision you need to make.
20 guides shown
A practical view of where AI assists useful work, where it creates new review work and where people must remain responsible.
Read the guide
Choose a general-purpose assistant by workflow fit, data controls, ecosystem and review needs rather than popularity alone.
Read the guide
Understand data minimisation, provider controls, retention, access and approval before entering business information into an AI service.
Read the guide
Use a repeatable verification process for claims, calculations, quotations, links and recommendations.
Read the guide
Create clearer AI instructions by defining the outcome, context, constraints, evidence and review format.
Read the guide
Start with customer evidence and brand judgment so AI assists distinctive marketing rather than multiplying empty content.
Read the guide
Assign each tool a clear role, preserve source truth and avoid compounding errors across an uncontrolled chain.
Read the guide
Recognise work where AI adds unacceptable risk, weakens trust or creates more review effort than value.
Read the guide
Use AI to accelerate structured exploration while validating the customer problem through real-world evidence.
Read the guide
Compare subscription price with adoption, integration, review effort, duplicated tools and the value of the workflow improved.
Read the guide
Build a usable AI policy covering approved tools, prohibited data, human accountability, review, incidents and regular updates.
Read the guide
Assess an AI product using representative tasks, current evidence, data controls, administration, integration and total cost.
Read the guide
Understand confident but unsupported AI output and design workflows that expose uncertainty, preserve sources and require verification.
Read the guide
Use AI to assist accessible work without relying on automation to replace standards, testing or disabled people’s experience.
Read the guide
Review source rights, provider terms, output provenance, brand permissions and human contribution before publishing AI-assisted content.
Read the guide
Design AI-assisted automation with bounded authority, validation, approvals, monitoring, audit records and a tested fallback.
Read the guide
Improve service workflows while preserving accurate policy, privacy, transparent escalation and access to accountable people.
Read the guide
Improve discoverability across traditional and AI-assisted search through useful content, technical clarity, entities, evidence and measurement.
Read the guide
Measure AI value through workflow quality, cycle time, review effort, errors, adoption, risk and total cost rather than usage alone.
Read the guide
Prepare people, workflows and controls for AI through role-based learning, protected experiments, support and evidence-led change.
Read the guideContinue with context