Runs on PeopleThe AI era
AI makes output cheap. People make it count.
AI makes output cheap. Code, drafts, analysis and plans now take minutes. That doesn't make people less important. It moves where they matter: judgement, trust, deciding what's worth doing, and owning the result.
What changes in each dimension
The nine dimensions still hold. AI shifts what each one asks of you.
- TrustBe open about where AI is used, especially in anything about people. Hidden AI erodes trust faster than bad AI.
- CapabilityJuniors used to learn by doing the routine work AI now does. Design new ways to build judgement, or you'll have no seniors in five years.
- EnergyFaster tools raise expectations. Watch for the pace creeping up because 'AI can do it'.
- ClarityWhen output is cheap, deciding what's worth doing is the scarce skill. Priorities matter more, not less.
- CadenceShorter cycles become possible. Keep the rhythms that make people think together: planning, review, retro.
- AgreementsAgree as a team how you use AI: what for, what not, what gets checked, and who signs off.
- OutcomesOutput is no longer the measure. Ask what changed for the customer, not how much was produced.
- ReliabilityAI makes confident mistakes. Build in review and testing proportional to the stakes.
- MeasuresOld productivity measures break. Lines of code, documents written and tickets closed all inflate. Measure outcomes and quality.
The AI frameworks, in plain English
What the big frameworks from AI companies, standards bodies and regulators actually say, and what each means for a manager.
Anthropic
Anthropic: building effective agents
Practical guidance for building with AI: start with the simplest thing that works, prefer predictable workflows over open-ended agents, keep the steps visible, and put people at the checkpoints.
What it means for youTreat AI like a new team member on probation: small, well-defined jobs first, visible work, and a human signing off what matters.
Anthropic
Anthropic: Responsible Scaling Policy
Anthropic's commitment to match safety measures to how capable a model is, using AI Safety Levels, with stronger safeguards required before more capable models are used.
What it means for youThe bigger the risk, the more checks. The same principle as change control and on-call: scale the safeguards to the stakes.
Amazon Web Services
AWS Responsible AI
Core dimensions for responsible AI: fairness, explainability, privacy and security, safety, controllability, veracity and robustness, governance, and transparency.
What it means for youA checklist for any AI use in your team. Most of the dimensions are people questions: who decides, who can see why, and who's accountable.
Amazon Web Services
AWS Well-Architected Generative AI Lens
Best practice for designing, running and improving generative AI workloads, built on the Well-Architected pillars such as operational excellence, security, reliability and cost.
What it means for youAI features need the same disciplines as any other system: ownership, monitoring, incident response and cost control.
Amazon Web Services
AWS Cloud Adoption Framework for AI (CAF-AI)
A guide to how organisations adopt AI, covering business, people, governance, platform, security and operations.
What it means for youAdopting AI is a change programme. The people perspective (skills, roles, culture) is where most adoption stalls.
US National Institute of Standards and Technology
NIST AI Risk Management Framework
A voluntary framework with four functions: Govern, Map, Measure and Manage AI risk.
What it means for youGovern is culture and accountability; Map is understanding the context; Measure and Manage are the Inspect and Adapt loop for AI.
International standards bodies (ISO and IEC)
ISO/IEC 42001
A certifiable standard for an AI management system: policies, roles, risk assessment and continual improvement.
What it means for youAI governance written down and owned. Process as people's agreements, applied to AI.
European Union
EU AI Act
Law that sorts AI uses by risk: unacceptable (banned), high-risk (strict duties), limited risk (transparency duties) and minimal risk.
What it means for youIf AI is used to make decisions about people, such as hiring, performance or access, expect it to count as high-risk. Managers need to know where that line is.
Microsoft
Microsoft Responsible AI Standard
Six principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.
What it means for youAccountability stays with a person. AI can recommend; someone has to own the decision.
What it means for your delivery framework
| Framework | What changes |
|---|---|
| SAFe | PI objectives and WSJF still work, but cost of delay and job size shift when AI shrinks the work. Lean-Agile leaders must redesign roles, not just cut headcount. Respect for people applies to how AI changes their jobs. |
| Scrum | Add AI review to the Definition of Done. Sprints may deliver more; the Sprint Review must still check it's the right thing. |
| Shape Up | Appetite gets more interesting: what's possible in six weeks grows. Shaping, and choosing what's worth betting on, matters more. |
| Kanban | Watch WIP: AI makes it easy to start more work. Finishing and reviewing becomes the bottleneck. |
| DORA | Throughput may rise; watch change failure rate and time to restore. Speed without stability isn't progress. |
| OKRs | Key results built on output (features shipped, content produced) become meaningless. Rewrite them as outcomes. |
Both sides of the desk
If you manage people
Redesign, don't just replace
- Redesign roles around judgement, not just tasks. Ask what each person should stop doing, and what they should now do more of.
- Make AI use a team agreement, not a secret.
- Protect how juniors learn. Pair them with people, not just tools.
- Keep the human moments human: feedback, hard conversations, recognition.
If you have a manager
Be the judgement, not the relay
- Use AI to do more of the work only you can do: judgement, relationships, ideas.
- Be the person who checks the output, not the one who forwards it.
- Make your thinking visible. 'AI wrote it' is not a reason; 'here's why I'm confident' is.
- Your reputation still runs on people. The people you know, know.
This page summarises public frameworks in our own words. Check the originals before relying on them: Anthropic (anthropic.com), AWS (aws.amazon.com), NIST (nist.gov), ISO (iso.org), the EU AI Act (eur-lex.europa.eu) and Microsoft (microsoft.com). Regulation changes; this isn't legal advice.