AI Readiness Assessment Scorecard

How to use this tool

1. 1. Answer each axis honestly.
2. 2. Review your total score and weakest axis.
3. 3. Use the top actions as a pre build checklist.
4. 4. Share the result with leadership or delivery teams.

What agentic AI readiness actually means

Agentic AI readiness is the ability to ship and operate an autonomous workflow that takes action, calls tools, and produces evidence. It is a stricter bar than running a chatbot. Readiness is the combination of clean data, observable operations, owned governance, present skills, and a product surface that can tolerate occasional model failure without harming users.

Many teams confuse interest in AI with readiness for AI. Interest is a starting point. Readiness is measured against the production conditions that catch unready systems: bad data, missing escalation, undefined risk owners, unstaffed evals, and product flows that punish small mistakes.

The five readiness dimensions explained

Data measures whether the workflow has access to clean, labelled, and current information. Operations measures observability, escalation paths, and on call coverage when an agent misbehaves. Governance measures whether risk owners, retention rules, and audit logs are defined and assigned. Skills measure whether the team can debug a prompt, write an eval, and reason about a failure mode without outside help.

Product surface is the fifth and most overlooked dimension. It measures whether the user interface, business process, and downstream systems can tolerate occasional uncertainty. A high stakes flow with no review path is unsafe to ship even if every other axis is strong.

How your score is interpreted

A score above 75 usually means ready to pilot with production discipline. A score between 50 and 75 means scoped remediation is needed before customer facing work. A score below 50 means the team should build foundations such as data quality, governance, or evals before committing to an agentic build.

The weakest axis matters more than the average. A system fails through its weakest link, so a strong data score with a missing governance owner is still a launch risk worth fixing first.

Enterprise AI readiness checklist

An enterprise level checklist adds two questions to each axis. For data, ask whether sensitive fields are tagged and access controlled. For operations, ask whether traces are retained long enough for incident review. For governance, ask whether a named owner can stop a launch. For skills, ask whether the team can run an adversarial eval. For product surface, ask whether a human can intervene before harm.

The scorecard converts these into a single readiness band. Pair it with a short architecture review when the band lands in the medium range, since most teams cluster there and need targeted intervention rather than a full rebuild.

How the scoring works

The score uses weighted axes: data 25 percent, operations 20 percent, governance 20 percent, skills 20 percent, and product surface 15 percent. Data receives the highest weight because weak data breaks retrieval, evals, and operational trust. Governance is scored separately because policy decisions should not be hidden inside an engineering process.

Weights are visible in the formula panel so you can argue with them. Adjust them in the code if your industry has different risk priorities.

Assumptions and methodology

This tool uses transparent browser-side calculations and curated assumptions rather than LLM-generated recommendations. Outputs are planning estimates. They should be validated against provider pricing, production traces, engineering quotes, or domain review before money, compliance, safety, or hiring decisions are made.

Numerical defaults are dated and surfaced on the page. The methodology favours explicit assumptions over false precision: every estimate is meant to expose the variable that drives the result, not to pretend that early planning data is exact.

Frequently asked questions

How do I know if my company is ready for agentic AI?

Run an AI readiness assessment across data, operations, governance, skills, and product surface. Companies that are ready usually have clean data with access controls, observable operations with named on call, explicit governance owners, an in house team that can debug prompts, and a product flow that allows human intervention before harm.

What is an AI maturity model?

An AI maturity model is a tiered description of an organisation's ability to ship and operate AI systems. Common tiers run from experimental to repeatable to scaled to optimised. This scorecard maps to a maturity model: scores below 50 sit at experimental, 50 to 75 at repeatable, and above 75 at scaled with production discipline in place.

How long does the readiness assessment take?

The scorecard takes about five minutes for a first pass. A leadership team can answer quickly, then revisit the weakest axes with evidence from docs, logs, support queues, and governance artifacts. A full evidence backed assessment usually takes two to four hours and benefits from an outside reviewer.

What does each readiness axis measure?

Data measures quality, currency, and access control. Operations measures observability and escalation. Governance measures policy and risk ownership. Skills measure team ability to debug and improve the system. Product surface measures whether the workflow can tolerate uncertainty and human review without harming users.

Is my data stored when I take the assessment?

No. The scorecard runs in your browser. Answers are kept in browser state and an optional URL hash that you can share. Nothing is sent to a server. The scorecard would only persist data if a future opt in benchmark is explicitly added with consent UI.

What is a good AI readiness score?

A score above 75 usually means ready to pilot with production discipline. Scores from 50 to 75 need scoped remediation, typically on governance or evals. Scores below 50 should build foundations such as data quality, escalation paths, and a named risk owner before committing to a customer facing agent.

What should I do if my score is below 50?

Start with the weakest axis. Usually that means cleaning data, defining escalation, writing governance rules, or assigning an owner for evals and observability before building the agent itself. A small foundational project for four to eight weeks often raises the weakest axis enough to unblock a real pilot.