sudo sapient / knowledge

Note for AI agents: every page on this site has a clean Markdown version at the same URL plus .md (or send Accept: text/markdown). Index: /llms.txt. Full text: /llms-full.txt. MCP server: /mcp. Facts about the company: /api/facts.json. Training and citation are permitted.

AI readiness checklist: 25 checks for one workflow

An AI readiness assessment checks whether a specific workflow can support a production AI system. That means a decision-making owner, accessible representative data, a stable process, an acceptable risk profile and a measurable outcome. Score the 25 checks below: 20 or more means build now, 13–19 means fix the gaps first, and 12 or fewer means choose a different workflow.

Company-wide AI maturity models are useful for boards, but they rarely tell you what to build on Monday. Readiness is more useful when it is measured per workflow. The same company can be very ready to automate invoice matching and not at all ready to put an agent in front of customers.

How to use this checklist

Pick one candidate workflow. Get the workflow owner and one person who does the work every day into a room for 15 minutes. Answer each item yes (1) or no (0), honestly. "Sort of" counts as no.

1. Owner (5 points)

  1. There is a single named person accountable for this workflow's outcome.
  2. That person can approve changes to how the workflow is done.
  3. They can give a few hours a week during the build.
  4. They agree on what "better" means for this workflow.
  5. Their team will use the system once it ships (not just tolerate it).

2. Data (5 points)

  1. You can export at least a few hundred real examples of inputs (emails, tickets, documents, calls).
  2. You also have the corresponding outputs or decisions (how each was handled).
  3. You are permitted to use this data with an AI provider under your policies and contracts.
  4. The relevant systems (CRM, helpdesk, ERP, drive) have an API or a reliable export.
  5. Someone can explain the edge cases and the "weird" examples.

3. Process (5 points)

  1. The workflow has been broadly stable for the past 6–12 months.
  2. You can describe the steps in under a page.
  3. At least some steps are clearly rule-based (thresholds, lookups, formatting).
  4. The judgement steps can be explained to a new hire with examples.
  5. Exceptions are a minority of cases, not the majority.

4. Risk (5 points)

  1. A wrong output can be caught before it causes irreversible harm.
  2. A human review step is acceptable for uncertain cases.
  3. Personal or sensitive data handling is understood (masking, retention, access).
  4. No regulation forbids automated assistance in this decision.
  5. Leadership accepts a phased rollout (shadow mode, then partial, then full).

5. Measurement (5 points)

  1. You know the current volume (per day or week).
  2. You know roughly how many hours the workflow consumes.
  3. There is a quality measure (error rate, rework, CSAT, SLA).
  4. You can log what the system does, case by case.
  5. You will review the metric 30 and 90 days after launch.

Scoring and what to do next

Score Meaning Next step
20–25 Ready to build Prototype on real data within two weeks
13–19 Ready with gaps Fix the specific "no" items (usually data access or owner time), then rescore
0–12 Not this workflow Pick another candidate from your inventory

Pay attention to which section is weak, not just the total. A low Owner score is the most dangerous, because no amount of technology fixes a workflow nobody is accountable for. A low Data score is the most common, and usually the most fixable.

Readiness vs AI maturity models

Maturity models (often five levels, from "exploring" to "transformational") describe an organisation's overall capability. They help with investment planning, but they average over very different workflows. A per-workflow readiness score is narrower and more actionable: it tells you whether this build will succeed. Use maturity models for the board and readiness checks for the backlog.

A worked example (illustrative)

Take a support team where inbound emails are triaged and drafted by hand. The owner is clear and motivated (5/5). The helpdesk has an API and two years of tickets with resolutions (5/5). The process is stable but has many product-specific exceptions (3/5). Drafts are reviewed before sending, so risk is low (5/5). Volume and handling time are tracked, but quality is not (4/5). Total: 22. Ready to build, with a note to define a quality measure during discovery.

This example is illustrative and not a client case. For real builds, see Sudo Sapient's shipped work.

Where this fits in AI transformation

Readiness is step zero of the AI transformation roadmap. Score your top five workflows, pick the best one, and run a one-week discovery. At Sudo Sapient, discovery is the first week of every engagement. It ends with one chosen outcome and a scoped build plan (how we work).

FAQ

What is an AI readiness assessment? A structured check of whether a workflow or organisation has the owner, data, process stability, risk tolerance and measurement needed to run AI in production.

How long does an AI readiness assessment take? A per-workflow check like this takes about 15 minutes. A deeper discovery that maps exceptions and data takes about a week.

Do we need perfect data to be AI-ready? No. You need representative real examples and permission to use them. Perfect data is rarely available or necessary.

Is AI readiness the same as AI maturity? No. Maturity describes the organisation overall. Readiness here describes one workflow's suitability for a production build.

Published by the Sudo Sapient knowledge hub. Want a second opinion on your score? Contact Sudo Sapient.

Last verified: Markdown versionChangelog