AI readiness checklist

Workflow and business case

We have chosen one specific workflow to start with, rather than a broad goal such as using AI across the business.
We can describe the workflow's steps, inputs, outputs and handoffs as they happen today, not only as the procedure document says.
We know roughly how often the workflow runs and how long each instance takes.
We can state what the current process costs in hours, delays, errors or lost revenue.
We know which steps follow clear rules and which need reading, interpretation or judgment.
We have checked that a simpler fix, such as a process change or an unused feature in existing software, would not solve the problem.
We can name one owner accountable for the workflow's results.

Data

We know which systems, inboxes and document stores hold the information the workflow needs.
The data is accurate and complete enough that a new employee could do the task from it.
We have real historical examples, such as past emails, invoices or tickets, that a solution can be tested against.
We know which data is confidential, personal or regulated, and where it is allowed to be processed and stored.
Someone is responsible for keeping the key records up to date.
We have agreed who owns the data and any outputs a new system creates.

Systems and integration

We have listed every application the workflow touches, including spreadsheets and shared inboxes.
We have confirmed that the core systems, such as the ERP or CRM, offer APIs or other supported integration options.
Our software licenses permit API access and automated use at the volume we expect.
We know who administers each system and can grant access for integration work.
We have a test or sandbox environment, or another safe way to test without changing live records.
We know whether the solution should run in our own AWS or Azure account or be hosted by a partner.

Security and governance

We have a written policy on which AI tools employees may use and with what data.
We have decided which actions AI may take automatically and which require a person's approval.
Access for any new integration or AI agent will follow least privilege, with no shared administrator accounts.
We know our legal, contractual and regulatory obligations for the data involved.
Actions taken by the system will be logged so we can review what happened and why.
IT and security are involved from the start rather than asked to approve at the end.

People and adoption

The people who do the work today have been asked where it slows down.
End users will test the solution with real cases before it goes live.
We have a plan for how roles and responsibilities change once the workflow is automated.
Training time is budgeted for each role that will use or review the system.
A senior sponsor has agreed to back the change if adoption is slow.
We have explained to staff what the project is for, including what it is not intended to do.

Ownership and measurement

We have measured a baseline for the workflow before any change, such as cycle time, error rate or hours spent.
We have agreed a target and how it will be measured after launch.
We know who will own the solution day to day once it is live.
We have agreed who handles maintenance, monitoring and incidents, and under what agreement.
We have a budget for running costs, including cloud hosting and AI model usage, not just the build.
We have set a review date to decide whether to expand, adjust or stop.

If several sections have gaps, a discovery call with Kastling is a no-commitment way to talk through which ones matter for your first workflow. Bring the checklist and we can discuss where to start.

See results

AI readiness checklist

Workflow and business case

Data

Systems and integration

Security and governance

People and adoption

Ownership and measurement

Read the guide

See how an idea becomes a working solution.