Plan the stack

An AI system readiness plan should not start with tools. It should start with the work your team repeats, the data that proves what happened, and the decision an AI system is supposed to improve.

What AI system readiness really means

AI system readiness is not a software inventory. A team can own modern tools and still be unready if the workflow is vague, the CRM fields are empty, or nobody reviews the output.

NIST's AI RMF Core organizes AI risk work around govern, map, measure, and manage. That is a useful lens for marketing and AI systems too: know the context, define the controls, measure the behavior, and manage what happens after launch.

The plan question

The best plan question is not "are we ready for AI?" It is: which repeated workflow is ready for an AI system?

That keeps the work practical. You are not scoring the whole company. You are checking whether one path can support an agent, report, routing rule, content review, or customer handoff.

What to inspect

1. The repeated job

Name the work the team repeats every week. Examples: qualify a lead, respond to a common question, prepare a client update, review a campaign, produce a report, or route a support issue.

2. The input evidence

List the fields, notes, forms, documents, calls, pages, tickets, and records the system would need to read. If the evidence is missing or inconsistent, the plan should say so plainly.

3. The system boundary

Name the tools involved. A useful AI system may need website forms, CRM records, analytics, email, documents, task tools, or product data. The question is whether the system can read and write safely where the workflow actually happens.

4. The owner and review rule

AI output needs an owner. Decide who reviews the work, what they are allowed to approve, and which outputs must stop for human judgment.

5. The risk boundary

Some outputs should never be shipped without review: pricing exceptions, legal claims, medical claims, financial advice, regulated promises, sensitive customer messages, and anything based on uncertain source material.

6. The measurement path

Do not measure readiness by tool count. Measure accepted outputs, edits, rejects, missing fields, owner response, state movement, and the number of exceptions that needed human review.

A 60-minute plan flow

• Minutes 1-10: pick one workflow and one business result.
• Minutes 11-25: inspect recent examples and source fields.
• Minutes 26-35: map the handoff, owner, and current failure points.
• Minutes 36-45: define the AI output and review rule.
• Minutes 46-55: list integrations, permissions, and stop rules.
• Minutes 56-60: decide whether to build, clean up, or wait.

What AI can run after the plan

If the plan passes, the next system should be narrow.

• Sales Agent: account research, fit summary, next-note draft, and CRM handoff prep.
• Marketing Agent: campaign review, content checks, source-approved drafts, and intent classification.
• Client Agent: client update prep, account context, unresolved-item review, and follow-up tasks.
• Report Agent: weekly reporting, attribution summary, exception list, and next-action brief.

When to wait

Wait when nobody owns the workflow, source fields are unreliable, the team cannot agree on the output, or the AI would need to make sensitive promises without review. Waiting is not failure. It is a good decision when the path is not ready.

How Conversion System uses the plan

AI Strategy turns the plan into a build recommendation. AI Agents builds the first bounded agent when the path is ready. Custom AI Systems handles deeper integration when the workflow spans several tools.

Conversion Skills supports the operating layer with repeatable skills for plans, research, content checks, reporting, and workflow review.

FAQ

What is an AI system readiness plan?

An AI system readiness plan checks whether a specific workflow has the data, tools, owner, review rule, risk boundary, and measurement path needed for an AI system.

How long does an AI system readiness plan take?

A first pass can take about an hour if the team focuses on one workflow. A deeper build plan takes longer because it needs source records, system access, stakeholder review, and integration planning.

What are the most common readiness problems?

The common problems are missing source fields, unclear ownership, disconnected tools, vague definitions, unreviewed AI output, and no measurement path.

What data is needed for AI marketing or growth systems?

The system needs examples of the work: inputs, source material, owner actions, outcomes, edits, rejects, and the state change that proves the workflow improved.

What should we do if the plan fails?

Clean up the smallest missing piece first: source fields, CRM state, handoff owner, approved source material, or review rule. Then rerun the plan on the same workflow.