How to Prepare for an AI Readiness Assessment: A Practical Business Checklist

To prepare for an AI readiness assessment, document one real business workflow, gather a few representative examples, identify the systems and people involved, and be clear about the outcome you want to improve. You do not need a finished AI strategy, a preferred software platform, or a technical team before you begin.

A useful assessment is not a software demonstration. It is a structured conversation about how work happens now, where time or information is being lost, and whether AI, conventional automation, system integration, or a simpler process change is the appropriate next step.

This checklist is designed for owners, operations managers and team leads at small and mid-sized Ontario businesses who want a practical starting point rather than a list of disconnected tools.

What should an AI readiness assessment help you decide?

An AI readiness assessment should help you understand which opportunities deserve attention and what would need to be true before implementation. Depending on the business, that may include:

• the repetitive or information-heavy work that creates avoidable effort;
• the business outcome the change is meant to improve;
• the information and systems the workflow depends on;
• the risks, permissions and approval points that must be controlled;
• the people who will use, review and own the process; and
• the sensible first phase, pilot or improvement to consider next.

The goal is not to force AI into every task. A well-prepared assessment may show that a reliable form, a CRM connection, clearer ownership or rules-based automation should come before an AI component. That practical distinction is central to AI and digital transformation planning.

1. Choose one workflow or business problem

Start with a process that matters to the business and happens often enough to observe. “We want to use AI” is too broad to assess. A more useful starting point is a specific problem such as:

• customer inquiries arriving through several inboxes or forms;
• staff copying the same information into a CRM, spreadsheet or project system;
• documents or PDFs that must be reviewed and entered manually;
• recurring reports that take too long to prepare; or
• employees searching through internal procedures or product information.

For example, a transportation company might want to understand why quote requests wait in a shared inbox. A manufacturer might be dealing with repetitive document checks. A professional-services firm might want employees to find approved internal information more quickly. The assessment should examine the actual workflow, not an abstract industry trend.

Choose one process for the first conversation. You can bring a short list of additional ideas, but keep one opportunity as the main example so the assessment produces a concrete discussion.

2. Map how the work happens today

Write the current process in plain language from the first input to the final outcome. Do not try to make it look efficient. The points that feel frustrating, repetitive or unclear are useful assessment information.

Record:

• where the request, document or task begins;
• who receives it and decides what happens next;
• which systems, files, forms or inboxes are used;
• where information is copied, checked or re-entered;
• where the work waits for a handoff;
• which exceptions require experience or judgement; and
• what counts as complete.

A simple before-and-after description is often enough. For example:

1. A customer submits an inquiry or sends an email.
2. An employee reads the request and identifies what is needed.
3. Details are copied into another system or document.
4. A responsible person checks the information and decides the next step.
5. The business prepares a response, quote, task or follow-up.
6. The status and source information are stored for future reference.

Once the current flow is visible, it becomes easier to discuss where assistance could help. AI might summarize, classify, extract or prepare a draft. It should not quietly invent missing information or take responsibility for a decision that belongs to an employee.

3. Gather representative information

Bring a small, representative sample of the information the workflow uses. The point is not to assemble every historical record. The point is to show normal cases, unusual cases and the gaps that make the work difficult.

Useful examples might include:

• redacted customer inquiries or form submissions;
• anonymized documents, PDFs or spreadsheets;
• the current report or quote template;
• sample internal procedures or reference documents;
• examples of incomplete, conflicting or exceptional requests; and
• a description of where each source is stored.

Remove personal information and confidential details unless there is a clear, approved reason to discuss them. A sample should be safe to review. If the workflow depends on information that cannot be shared, describe its structure and sensitivity instead.

Also note what the business considers authoritative. If three spreadsheets contain different versions of a price, process or customer record, that is not a minor detail. It may be a data or ownership problem that needs attention before an AI workflow can be trusted.

4. List the systems, access and handoffs

Make a simple inventory of the tools involved. Include the systems employees use every day, even if they do not seem connected to the proposed AI idea.

• website and contact forms;
• shared or individual email accounts;
• CRM or customer database;
• accounting, inventory or estimating software;
• cloud storage and document libraries;
• spreadsheets and internal databases;
• project management or scheduling tools; and
• reporting and analytics platforms.

For each system, identify who can access it, what information it owns, and where the next person expects to find the result. Pay attention to duplicate entry and informal workarounds. A proposed automation may save time in one step while creating a new tracking gap elsewhere if the systems are not mapped together.

If the workflow involves inquiries, ownership, status or follow-up, review it alongside the business’s CRM and lead-management workflow. The question is not only whether AI can prepare information. It is where the customer record and next action should live.

5. Involve the people who do and approve the work

The person who sponsors an assessment may not be the person who performs the process every day. Include both perspectives where possible.

Ask the people involved:

• Which part of the process takes the most time?
• Where do requests most often wait or get lost?
• Which information is difficult to find or verify?
• What errors create the most rework?
• Which exceptions require experienced judgement?
• What would make the proposed change harder to use?
• Who should review a draft or exception?

Include the person who owns the final customer, financial, operational or safety decision. An AI-assisted workflow should make that responsibility clearer, not hide it behind a tool.

Early involvement also helps distinguish a useful improvement from a technically interesting project that employees cannot fit into their day. Adoption, training and ownership belong in the initial assessment.

6. Identify privacy, security and approval requirements

Before discussing an AI tool, identify what could go wrong if the information is exposed, misread or sent to the wrong place. Customer emails, employee records, pricing information, drawings, financial documents and internal policies may all require careful handling.

Prepare to discuss:

• what information the workflow contains;
• which information is confidential or personal;
• where the information is currently stored;
• which users should be able to access it;
• what an AI service or connected platform would need to see;
• how long records should be retained; and
• how access changes when an employee changes roles.

Then mark the human approval points. A system may be allowed to prepare a summary, flag a missing field or suggest a next step. A responsible employee may still need to approve a price, customer commitment, financial record, external message or exception.

A strong first project often automates preparation, organization or routing before it automates a consequential action. The assessment should make that boundary explicit.

7. Define the outcome and a baseline

Decide what improvement would matter before you decide what technology to use. Depending on the workflow, the useful measure might be:

• time spent preparing or reviewing a request;
• number of manual handoffs or duplicate entries;
• amount of rework caused by missing information;
• time between an inquiry and the next assigned action;
• percentage of records with required fields completed; or
• how easily employees can locate the final record.

Use measures that reflect the actual business problem. Website visits, time on site and other engagement signals are not the same as qualified leads or revenue. If marketing measurement is part of the discussion, keep the difference clear.

You do not need a perfect baseline. A reasonable estimate, a short observation period or a few reviewed examples can be enough to frame the opportunity. The important thing is to know what the project is supposed to improve and how you will decide whether to continue.

What to bring to the assessment

A one-page brief can keep the conversation focused. Include:

• Business problem: What is difficult, slow, repetitive or unreliable?
• Current workflow: What happens from the first input to the final outcome?
• People: Who performs, reviews and owns the work?
• Information: What documents, records or messages are involved?
• Systems: Where is the information stored and where should the result go?
• Exceptions: What cannot be handled safely without human judgement?
• Risks: What privacy, security, financial, customer or operational concerns apply?
• Outcome: What would a useful improvement look like?
• Constraints: What budget, timing, staffing or integration limitations matter?

If you have additional opportunities, list them separately and explain why they may matter. Avoid turning the first assessment into an unranked catalogue of every task in the organization.

What you do not need before an AI readiness assessment

You do not need to choose a platform

ChatGPT, Microsoft Copilot, Claude, Gemini, a CRM feature or a custom application may each be appropriate in different situations. Choosing a product before understanding the workflow can limit the assessment to the tool’s preferred use case.

You do not need a complete technical inventory

Bring what you know. The assessment can identify missing system, data or access details that need to be confirmed. Be honest about uncertainty rather than filling gaps with assumptions.

You do not need to clean every record

It is useful to show where data is inconsistent, incomplete or duplicated. You do not need to fix the entire business before discussing whether a focused improvement is worthwhile.

You do not need to promise that AI will replace a process

The right outcome may be a smaller amount of manual entry, faster preparation, better routing or more consistent information. The assessment should test that possibility without assuming that human involvement can disappear.

What happens after the assessment?

A useful result is a prioritized roadmap, not just a list of software names. It may recommend improving the current process, preparing data, connecting existing systems, testing a limited workflow, training employees or postponing an idea until a foundation is ready.

The roadmap should separate discovery, design, pilot, rollout and ongoing support. Each phase should have an owner, a clear boundary, a review point and a way to decide whether the next phase is justified.

Design Tactics’ AI Readiness Assessment service is built around understanding current operations, identifying realistic opportunities and creating a practical path toward implementation. The assessment can also show when traditional automation or better system integration is the more appropriate answer.

Frequently asked questions

Do I need technical expertise to prepare?

No. You need a clear understanding of how the business works and where the process creates friction. Technical details can be identified and confirmed during the assessment.

How much data should I bring?

Bring enough redacted or anonymized examples to show normal cases, exceptions and missing information. A small, representative sample is more useful than an unstructured export of every record.

Should we start with more than one workflow?

You can discuss several opportunities, but choose one main workflow for the first assessment. A focused example makes it easier to compare value, feasibility, risk, integration requirements and employee impact.

What if the assessment shows that AI is not the right answer?

That can be a useful result. A clearer process, better data, conventional automation or a system connection may solve the underlying problem with less complexity. The objective is a dependable business improvement, not an AI label.

The practical answer

Preparing for an AI readiness assessment means bringing the business context that a software demonstration cannot provide: one real workflow, the people and systems around it, representative information, the risks that matter, and a measurable outcome.

Start small, describe the current process honestly and keep human approval visible wherever the work affects customers, money, safety, privacy or operational commitments. That gives the assessment a useful foundation and makes any later implementation more deliberate.

If you want help identifying where AI, automation or integration could fit, contact Design Tactics with one workflow your team repeats and the problem you want to improve.