Whitepaper

The AI Readiness Playbook: how to assess your organization before you invest

25 October 2026 · 15 min read · Hidden Node

Executive summary

Most organizations do not fail at AI because the technology is not ready. They fail because they pick the wrong problems, underestimate data and integration work, or launch systems people do not trust. This playbook gives leaders a structured way to check readiness before investing, across six dimensions:

1. Strategy and outcomes
2. Data
3. Processes
4. People and skills
5. Technology and integration
6. Governance, risk and compliance

For each dimension we describe what good looks like, the questions to ask and a simple 1–4 scoring guide. The playbook ends with a method to prioritise use cases and a 90-day plan template.

How to use this playbook

Run it as a short workshop series with a small group: a business sponsor, an operations lead, someone from IT or data, and someone from risk or compliance. Score each dimension honestly. Low scores are not failures; they show where to invest first and which use cases to avoid for now.

Dimension 1: Strategy and outcomes

What good looks like: leaders can name the business outcomes AI should improve (cost per case, turnaround time, revenue per customer, error rates), there is a sponsor with budget and decision rights, and AI projects are judged on those outcomes rather than on demos.

Questions to ask:

• Which two or three business metrics matter most this year?
• Who owns each metric, and would they sponsor an AI project?
• How will we decide whether a project succeeded?

Scoring guide:

1. No defined outcomes; interest is driven by curiosity or vendor pitches.
2. General ambition exists, but no owner or metric.
3. Clear outcomes and a sponsor for at least one area.
4. A portfolio of outcomes with owners, budgets and review cadence.

Dimension 2: Data

What good looks like: the data needed for priority workflows can be accessed, its quality is understood, and there are rules about who may use it and for what.

Questions to ask:

• Where does the information for this workflow live: systems, documents, email, people's heads?
• Can we get access to historical examples to test with?
• Is quality good enough to act on, and how do we know?
• Is any of it personal or sensitive, and under what consent or legal basis do we hold it?

Scoring guide:

1. Data is scattered, access is unclear, quality unknown.
2. Data exists but access requires major effort; quality issues known but unmeasured.
3. Data for priority workflows is accessible, with known quality and owners.
4. Well-governed data platform with catalogued, accessible, quality-monitored data.

Note that "messy" data is not a blocker for every use case. Many valuable projects work directly on documents and email as they are today.

Dimension 3: Processes

What good looks like: priority workflows are mapped, their volumes and handling times are known, and the decision rules are written down or can be.

Questions to ask:

• What triggers the workflow, what steps follow, and which systems are touched?
• Where do people spend time reading, checking or re-typing?
• What are the rules for a correct outcome, and where is judgement needed?
• What happens to exceptions today?

Scoring guide:

1. Processes vary by person; little documentation.
2. Main steps known, but volumes and times are not measured.
3. Priority processes mapped with baseline metrics.
4. Processes standardised, measured and continuously improved.

Dimension 4: People and skills

What good looks like: the people who do the work are involved early, there is a plan for training and changing roles, and someone internally can own AI systems after launch.

Questions to ask:

• How do the teams affected feel about automation, and what are they worried about?
• Who will review AI outputs, handle exceptions and own the system?
• What training will each role need?

Scoring guide:

1. No involvement of affected teams; skills gap unaddressed.
2. Some awareness; no structured plan.
3. Affected teams involved; training and ownership planned.
4. Established capability: trained users, internal owners and a feedback culture.

Dimension 5: Technology and integration

What good looks like: core systems have APIs or other safe ways to read and write data, there is an approved cloud environment, and security teams know how AI services will be assessed.

Questions to ask:

• Can our core systems be integrated safely, and who owns them?
• Do we have an approved cloud environment where AI workloads can run?
• What are the security requirements for new services and vendors?

Scoring guide:

1. Legacy systems with no integration options; no approved environment.
2. Some integration options; security process unclear for AI.
3. Approved environment and integration paths for priority systems.
4. Mature platform with reusable integration, monitoring and deployment practices.

Dimension 6: Governance, risk and compliance

What good looks like: there are clear rules on acceptable AI use, data protection obligations (such as India's DPDP Act or the GDPR) are understood, high-risk decisions keep a person accountable, and systems are tested before and monitored after launch.

Questions to ask:

• Which decisions must always be made or approved by a person?
• What are our obligations for the personal data involved?
• How will we test, monitor and audit AI systems?
• Who approves launch, and who responds if something goes wrong?

Scoring guide:

1. No policy; AI use is unmanaged.
2. Informal guidance; no testing or audit standard.
3. Policy, approval process and testing standard in place for new systems.
4. Embedded governance with monitoring, audits and regular review.

Reading your scores

• Mostly 1s and 2s: start with strategy and one carefully chosen, low-risk use case where people review all outputs. Invest in data access and governance alongside.
• Mostly 2s and 3s: you are ready for a production use case with clear metrics. Use it to build reusable foundations.
• Mostly 3s and 4s: move to a portfolio approach, running several use cases on shared foundations.

Prioritising use cases

List candidate workflows and score each from 1 to 3 on:

• Value: volume × time or cost per case, plus quality and speed benefits.
• Feasibility: data access, clear rules and integration effort.
• Risk: consequences of an error and how easily errors are caught.
• Readiness: sponsor, team engagement and measurable baseline.

Choose a first project that scores well on value and feasibility, low on risk, and has an engaged sponsor. Avoid starting with the most complex or most visible problem.

The 90-day plan template

Days 1–15: Define

• Confirm the use case, sponsor and success metrics.
• Measure the baseline.
• Agree data access, security and compliance requirements.
• Build a test set of real past cases.

Days 16–60: Build and test

• Build the workflow end to end, including integrations and the review interface.
• Evaluate against the test set and agree thresholds.
• Run in shadow mode alongside the team.
• Train reviewers and update procedures.

Days 61–90: Launch and measure

• Launch for a limited group with people reviewing outputs.
• Monitor quality, escalations and the business metric.
• Report results against the baseline.
• Decide: expand, adjust or stop.

Conclusion

AI readiness is not a single yes-or-no question. Every organization has areas where it is ready today and areas that need work. Assessing honestly, choosing a first project that fits your readiness, and building foundations as you go is the fastest reliable path to results.

Hidden Node runs this assessment free of charge over two weeks, ending with a ranked list of use cases and a 90-day plan. Contact us to book one.