August 21, 2026
August 21, 2026
Is Your Data Ready for AI? A Practical Readiness Test Before You Automate
Before you automate, use this seven-question test to see whether your data is usable, owned, authorized and measurable for one specific workflow.
Before you automate, use this seven-question test to see whether your data is usable, owned, authorized and measurable for one specific workflow.
A leadership team asks its new AI assistant a simple question: Which customer accounts need attention this week? The answer looks polished. It is also wrong. Two important accounts are missing because the CRM is incomplete. One appears twice because a spreadsheet uses a different company name. The latest client concern lives only in an account manager’s inbox. The model did not create that confusion. It exposed it. Your data is ready for AI when it is fit for one defined workflow and decision—not when every record in the company is perfect.
AI data readiness is specific, not universal
“Is our data ready for AI?” sounds like a company-wide question. It usually is not.
Your data may be ready to summarize support tickets but not ready to predict customer churn. It may be ready to draft a project update but not to approve a payment. Different workflows require different sources, freshness, permissions, accuracy and human review.
That is why the first step is not “clean everything.” The first step is to name:
the decision or action the AI will support;
the information required;
the acceptable cost of a wrong answer;
the person who remains accountable.
Then you can test readiness with evidence.
The seven-question AI data readiness test
Score each question:
2 — Yes: clear, documented and working;
1 — Partly: usable, but gaps or manual work remain;
0 — No: unknown, inconsistent or unsafe.
This is a practical Myappics screening tool, not an industry benchmark.
1. Can you name the exact workflow and decision?
“Use AI in sales” is too broad.
“Summarize every new inquiry, classify its service need and route it to the correct owner” is specific enough to evaluate.
If the desired action is unclear, no amount of data cleanup will make the project ready.
2. Is there an authoritative source for each critical fact?
A system of record does not mean every piece of information must live in one platform.
It means the organization knows which source wins when records disagree.
For example:
the CRM owns contact and opportunity status;
the accounting system owns invoices and payments;
the project system owns delivery milestones;
an approved knowledge base owns current policies.
If nobody can say which source is authoritative, the AI will inherit the argument.
3. Do the fields mean the same thing to everyone?
A field called “active,” “qualified” or “complete” can hide five different definitions.
AI needs business context, not just columns and documents. Critical terms should have a plain-language definition, an owner and a rule for when they change.
You do not need a massive data dictionary. Start with the terms that change the workflow’s decision.
4. Is the data complete and current enough for the risk?
Perfect data is not the goal. Fit-for-purpose data is.
A weekly internal summary may tolerate a missing note that a payment approval cannot. The higher the consequence of an error, the stronger the quality check and human review must be.
Ask:
How often is the source updated?
Which fields are commonly missing?
What happens when two records conflict?
How will the system recognize stale information?
5. Can the right systems and people access it safely?
Accessible does not mean open to everyone.
Confirm that the workflow can reach the required information without bypassing privacy rules, contracts, role permissions or security controls. Sensitive data should have a clear boundary before it enters a model, automation or third-party service.
If the access rule is “we will figure it out during the pilot,” the pilot is not ready.
6. Is there a named owner and an exception path?
Someone must own the data used by the workflow. Someone must also decide what happens when the AI is uncertain, the source is unavailable or a recommendation looks wrong.
A reliable workflow answers four questions:
Who reviews exceptions?
What requires approval?
When does the system stop?
Where is the correction recorded?
Human review is part of the system design, not evidence that the AI failed.
7. Can you measure whether the result improved?
Before launch, define the baseline and the outcome.
Depending on the workflow, that might include:
cycle time;
rework;
missed handoffs;
response time;
error rate;
adoption;
cost per completed process;
revenue or retention only when attribution is defensible.
If the team cannot explain what “better” means, it cannot distinguish improvement from novelty.
What your score means
Add the seven answers for a maximum of 14.
12–14: Ready for a bounded pilot
The workflow has enough clarity to test. Keep the scope narrow, preserve human review and define a go/no-go decision before launch.
8–11: Repair the named gaps first
Do not start a company-wide data program. Fix the missing owner, definition, permission, source or measure that blocks this workflow.
0–7: Do not automate this yet
Map the workflow and its information flow before selecting technology. A fast build on an unclear foundation will only automate disagreement.
Again, these thresholds are a practical screen—not a guarantee of technical, legal or commercial readiness.
Three traps that make readiness harder
Cleaning every dataset before choosing a use case
This creates a long project with no decision at the end. Choose the workflow first. Prepare only the information it actually needs.
Buying a new platform to avoid ownership decisions
Technology can connect records. It cannot decide which definition is correct or who is accountable when the answer is wrong.
Treating readiness as a one-time event
Sources change. Employees change how they enter information. Permissions drift. Business rules move.
AI-ready data must be monitored and maintained after launch, just like the workflow itself.
The smallest useful next step
Choose one recurring workflow that consumes meaningful time or creates costly mistakes.
Put the seven questions in front of the business owner, the person closest to the work and whoever controls the relevant systems. Score the workflow together. Do not average away disagreement; investigate it.
The goal is not to prove that the company is “AI ready.” The goal is to identify the smallest reliable next move.
That might be a pilot. It might be a data repair. It might be a clearer owner. All three are progress when the decision is based on evidence.
Next step with Myappics
The free 2-minute AI Reality Check helps you see where your organization currently stands with AI and where time or money may be leaking.
If the result raises a useful question, you can schedule a free 30-minute conversation with Myappics. We will discuss your needs, clarify the priorities and decide together whether we are the right fit.
No forced roadmap. No invented ROI. Just a clearer next decision.
Sources and further reading
A leadership team asks its new AI assistant a simple question: Which customer accounts need attention this week? The answer looks polished. It is also wrong. Two important accounts are missing because the CRM is incomplete. One appears twice because a spreadsheet uses a different company name. The latest client concern lives only in an account manager’s inbox. The model did not create that confusion. It exposed it. Your data is ready for AI when it is fit for one defined workflow and decision—not when every record in the company is perfect.
AI data readiness is specific, not universal
“Is our data ready for AI?” sounds like a company-wide question. It usually is not.
Your data may be ready to summarize support tickets but not ready to predict customer churn. It may be ready to draft a project update but not to approve a payment. Different workflows require different sources, freshness, permissions, accuracy and human review.
That is why the first step is not “clean everything.” The first step is to name:
the decision or action the AI will support;
the information required;
the acceptable cost of a wrong answer;
the person who remains accountable.
Then you can test readiness with evidence.
The seven-question AI data readiness test
Score each question:
2 — Yes: clear, documented and working;
1 — Partly: usable, but gaps or manual work remain;
0 — No: unknown, inconsistent or unsafe.
This is a practical Myappics screening tool, not an industry benchmark.
1. Can you name the exact workflow and decision?
“Use AI in sales” is too broad.
“Summarize every new inquiry, classify its service need and route it to the correct owner” is specific enough to evaluate.
If the desired action is unclear, no amount of data cleanup will make the project ready.
2. Is there an authoritative source for each critical fact?
A system of record does not mean every piece of information must live in one platform.
It means the organization knows which source wins when records disagree.
For example:
the CRM owns contact and opportunity status;
the accounting system owns invoices and payments;
the project system owns delivery milestones;
an approved knowledge base owns current policies.
If nobody can say which source is authoritative, the AI will inherit the argument.
3. Do the fields mean the same thing to everyone?
A field called “active,” “qualified” or “complete” can hide five different definitions.
AI needs business context, not just columns and documents. Critical terms should have a plain-language definition, an owner and a rule for when they change.
You do not need a massive data dictionary. Start with the terms that change the workflow’s decision.
4. Is the data complete and current enough for the risk?
Perfect data is not the goal. Fit-for-purpose data is.
A weekly internal summary may tolerate a missing note that a payment approval cannot. The higher the consequence of an error, the stronger the quality check and human review must be.
Ask:
How often is the source updated?
Which fields are commonly missing?
What happens when two records conflict?
How will the system recognize stale information?
5. Can the right systems and people access it safely?
Accessible does not mean open to everyone.
Confirm that the workflow can reach the required information without bypassing privacy rules, contracts, role permissions or security controls. Sensitive data should have a clear boundary before it enters a model, automation or third-party service.
If the access rule is “we will figure it out during the pilot,” the pilot is not ready.
6. Is there a named owner and an exception path?
Someone must own the data used by the workflow. Someone must also decide what happens when the AI is uncertain, the source is unavailable or a recommendation looks wrong.
A reliable workflow answers four questions:
Who reviews exceptions?
What requires approval?
When does the system stop?
Where is the correction recorded?
Human review is part of the system design, not evidence that the AI failed.
7. Can you measure whether the result improved?
Before launch, define the baseline and the outcome.
Depending on the workflow, that might include:
cycle time;
rework;
missed handoffs;
response time;
error rate;
adoption;
cost per completed process;
revenue or retention only when attribution is defensible.
If the team cannot explain what “better” means, it cannot distinguish improvement from novelty.
What your score means
Add the seven answers for a maximum of 14.
12–14: Ready for a bounded pilot
The workflow has enough clarity to test. Keep the scope narrow, preserve human review and define a go/no-go decision before launch.
8–11: Repair the named gaps first
Do not start a company-wide data program. Fix the missing owner, definition, permission, source or measure that blocks this workflow.
0–7: Do not automate this yet
Map the workflow and its information flow before selecting technology. A fast build on an unclear foundation will only automate disagreement.
Again, these thresholds are a practical screen—not a guarantee of technical, legal or commercial readiness.
Three traps that make readiness harder
Cleaning every dataset before choosing a use case
This creates a long project with no decision at the end. Choose the workflow first. Prepare only the information it actually needs.
Buying a new platform to avoid ownership decisions
Technology can connect records. It cannot decide which definition is correct or who is accountable when the answer is wrong.
Treating readiness as a one-time event
Sources change. Employees change how they enter information. Permissions drift. Business rules move.
AI-ready data must be monitored and maintained after launch, just like the workflow itself.
The smallest useful next step
Choose one recurring workflow that consumes meaningful time or creates costly mistakes.
Put the seven questions in front of the business owner, the person closest to the work and whoever controls the relevant systems. Score the workflow together. Do not average away disagreement; investigate it.
The goal is not to prove that the company is “AI ready.” The goal is to identify the smallest reliable next move.
That might be a pilot. It might be a data repair. It might be a clearer owner. All three are progress when the decision is based on evidence.
Next step with Myappics
The free 2-minute AI Reality Check helps you see where your organization currently stands with AI and where time or money may be leaking.
If the result raises a useful question, you can schedule a free 30-minute conversation with Myappics. We will discuss your needs, clarify the priorities and decide together whether we are the right fit.
No forced roadmap. No invented ROI. Just a clearer next decision.
Sources and further reading
