Skip to content
Home → Prepare for AI — Q10 AI Readiness

Prepare for AI — Q10 AI Readiness

AI readiness is not model shopping. Before a person, family, business, builder or public organization adds AI, it needs a bounded outcome, qualified data, explicit authority, evidence, evaluation, security, economics, human operating rules and a recovery path.

TEN FOUNDATIONS

What must exist before AI can work reliably

Use these as a readiness gate. A missing foundation does not always block experimentation, but it should limit autonomy and scale.

1

Outcome & scope

Define the decision or workflow, accountable owner, success metric, prohibited uses and stop conditions.

2

Data & rights

Know source, provenance, quality, sensitivity, freshness, retention, usage rights and data owner.

3

Identity & authority

Know every human and synthetic actor and what each may read, write, send, buy, delete, share or change.

4

Model / tool fit

Benchmark the minimum sufficient model, automation or conventional software against the real job.

5

Evidence & provenance

Record what supported an output or action, the version used, what changed and what remains uncertain.

6

Evaluation & TSLP

Test normal, edge, adversarial, change and recovery scenarios and measure false-clear behavior.

7

Security & privacy

Threat-model prompts, tools, agents, credentials, data flows, memory and third parties.

8

Economics & resource budget

Measure cost per accepted task, human intervention, latency, compute and resource consumption.

9

Human operating model

Define who approves, overrides, escalates, trains, audits and owns the outcome when AI is wrong.

10

Recovery & exit

Design kill, rollback, restore, export and provider-exit paths before critical dependency develops.

COME AS YOU ARE

Different groups need different preparation

People & families

AI literacy, privacy boundaries, verification habits, age/role limits, trusted contacts and a clear human decision owner.

Small business

One measurable workflow, approved tools, data rules, SOPs, cost controls and a monthly quality/ROI review.

Enterprise

Central inventory, decision rights, lifecycle gates, vendor controls, evidence, cost allocation, incident response and portfolio governance.

Builders & AI teams

Representative evals, permissions, observability, change revalidation, receipts, tool security and rollback.

Government & civic

Public accountability, records, accessible explanations, procurement evidence, continuity, correction and durable human authority.

Regulated / critical

Risk classification, separation of duties, validation, audit trail, fail-safe modes, DR, domain experts and independent review where required.

INTERACTION DESIGN

Human↔synthetic and synthetic↔synthetic are not the same problem

Human ↔ Synthetic

Needs understandable boundaries, consent, evidence, uncertainty, escalation, accessibility, correction and override.

human authorityverificationexplanationappeal / correction

Synthetic ↔ Synthetic

Adds machine-verifiable identity, scoped delegation, lineage, TTL/depth/budget limits, tool permissions, loop termination, handoff receipts and a route back to accountable humans.

identitydelegationTTLbudgetrollback
Minimum viable readiness: select one bounded workflow → establish baseline → qualify data → define authority → choose the minimum sufficient method → run TSLP → measure economics → rehearse rollback → issue an evidence-backed go/no-go decision.

External anchors: NIST AI RMF · ISO/IEC 42001 · Q10 Research Observatory