AI Readiness Assessment for Alberta Businesses: 12 Questions Before You Automate

Most businesses do not have an AI problem. They have a workflow problem wearing an AI hat.

The owner sees a good demo and pictures an assistant that answers customers, updates the CRM, and prepares quotes. Then the project reaches the real business. Details arrive through four channels, the CRM is optional, and the person who knows the exceptions is too busy to explain them.

That does not mean the business should avoid AI. It means the first step is an honest AI readiness assessment.

I use readiness to answer a practical question: can we improve one important workflow without creating a more expensive mess? You need a process worth fixing, a person who owns it, reliable source information, and sensible boundaries.

The 12 questions below will help you find out whether your Alberta business is ready to automate, what needs attention first, and how to choose a pilot with a real chance of working.

How to score your AI readiness assessment

Score each question from zero to two:

• 0 means no, unknown, or inconsistent
• 1 means partly, sometimes, or only for certain staff
• 2 means yes, clearly and consistently

Add the scores at the end. A low score on ownership or customer data can sink a project even if everything else looks tidy.

Use one specific workflow while answering. “Our sales process” is too broad. “Responding to website quote requests” is useful. So is “turning completed work orders into billing-ready files.”

If you need ideas, my guide to five AI workflows Alberta businesses should automate first covers good starting points in sales, scheduling, documents, reporting, and follow-up.

1. Does the workflow solve a costly problem?

Start with the business case, not the tool.

What happens when this workflow runs poorly? Maybe a lead waits until the next morning. A quote never gets followed up. A technician sends incomplete job notes. An invoice sits for a week because one document is missing. Staff spend Friday afternoon rebuilding a report from several systems.

Write down the consequence in plain language. Give yourself two points if the problem is frequent and the cost is visible. Give yourself one if the pain is real but poorly measured. Give yourself zero if AI merely seems interesting.

Interesting technology is a weak reason to change how people work.

2. Does the process happen often enough to matter?

Automation earns its keep through repetition.

A task that happens once a year may be annoying, but it is rarely a good first build. A task that happens daily or weekly gives you more chances to save time, collect feedback, and catch mistakes before expanding.

Frequency also makes testing easier. If you process 40 inquiries during a pilot, you can see patterns. If you process two, one unusual customer can distort the result.

Score two when the workflow happens every week and has enough volume to test. Score one when it is monthly or seasonal but still important. Score zero when it is rare, unpredictable, or different every time.

Seasonal businesses should judge frequency during the period when the work actually happens.

3. Can someone explain the current process?

If the team cannot explain the process, software cannot reliably follow it.

Ask the person doing the work to walk through the last five real examples. Where did each request begin? What information did they need? Which system did they check? What made them stop and ask a manager? What happened when information was missing?

Do not settle for the official process if nobody follows it. The useful map is what happens on a normal Tuesday when phones are ringing and somebody is away.

Score two if the steps, decisions, and common exceptions are understood. Score one if the process mostly lives in one employee’s head. Score zero if several people describe completely different workflows.

When the score is low, document and simplify before adding AI. This is not wasted prep. It often fixes part of the problem by itself.

4. Is there one clear trigger and one clear finish?

A narrow workflow needs a recognizable start and end.

A website form submission can trigger lead intake. A signed work order can trigger billing preparation. A completed appointment can trigger a review request. The finish might be a booked call, an approved draft, a complete customer record, or a file ready for human review.

Vague boundaries create sprawling projects. “Help with customer service” is not one workflow.

Score two when the trigger and desired output are obvious. Score one when one side is clear but the other needs work. Score zero when the proposed system is expected to help with everything.

My comparison of AI chatbots and AI agents can help if you are unsure whether the job needs a conversation tool, a workflow that takes action, or a simpler automation.

5. Is the source information reliable?

AI can summarize, classify, extract, and draft. It cannot make missing business information true.

Look at the source material. Are prices current? Are service areas documented? Are customer records complete? Do job notes identify the right customer and project? Can staff find the approved policy, or are they relying on an old PDF and memory?

Score two when the required information is current, accessible, and reasonably consistent. Score one when it is usable but needs cleanup. Score zero when records are missing, duplicated, or regularly contradicted.

A small set of trusted documents is more useful than thousands of files nobody wants to own. If the project depends on internal knowledge, read my guide to RAG and business data before loading a shared drive into a new system.

6. Are the rules stable enough to automate?

Some decisions follow a repeatable policy. Others depend on judgment, changing conditions, or experience that is hard to write down.

Sending an appointment reminder at an approved time is stable. Choosing whether an unusual job is safe, profitable, or legally compliant may not be. Drafting a quote from a current price table can be reasonable. Inventing a price when the scope is incomplete is not.

Score two when the rules are documented and change infrequently. Score one when most cases are predictable but exceptions need review. Score zero when every case depends on a manager’s judgment.

A score of one is often fine. The workflow simply needs a clear handoff to a person instead of pretending the exception does not exist.

7. Can the risky actions wait for human approval?

The first version should make work easier to review, not remove accountability.

An AI system can prepare a customer reply without sending it. It can extract invoice details without approving payment. It can recommend a lead category without deleting the record. It can flag a maintenance note without deciding whether equipment is safe to operate.

Score two if high-impact actions can stay behind an approval step. Score one if permissions need redesign. Score zero if the pilot requires AI to make irreversible financial, safety, employment, legal, or customer commitments.

If a vendor treats human review as an inconvenience, I would keep looking.

8. Do you know what data the system may access?

List the information the workflow needs. Then remove anything it does not.

A lead intake system may need contact details and the customer’s request. It probably does not need payroll files. A document assistant may need a specific folder, not the whole company drive. A scheduling tool may need approved availability without receiving unrestricted access to every calendar detail.

Score two when access can be limited by role, folder, record type, or system. Score one when permissions exist but are broad or poorly understood. Score zero when nobody knows what the tool can read, retain, or share.

Ask vendors where data is stored, how long it is retained, whether it trains their models, and how access is removed.

9. Is one person responsible for the workflow?

AI projects fail quietly when ownership is shared by everyone and held by nobody.

Choose an internal owner who understands the work and has enough authority to make small process decisions. That person does not need to build the technology. They need to review exceptions, collect staff feedback, keep source information current, and decide whether the result is useful.

Score two when one named person owns the workflow and its result. Score one when the likely owner exists but the responsibility is informal. Score zero when the project belongs to “the team.”

The consultant or software vendor should not be the only owner. They can build and support the system. Your business still owns the process.

10. Will the people doing the work help test it?

A workflow can pass a technical test and still make the job worse.

The people who handle the task know where the awkward cases live, which approvals take longer, and which alerts will be ignored after a week.

Score two if the staff doing the work can test real examples and report corrections. Score one if they can participate but have not been included yet. Score zero if the system will simply appear one Monday morning.

I am not interested in using AI to surprise employees. My post on what happens when businesses introduce AI to their teams explains why staff involvement produces better systems and less understandable resistance.

11. Can you measure one business result?

Pick a measure before the build starts.

For lead intake, track response time, qualified conversations, or booked calls. For documents, track handling time, missing fields, or correction rates. For billing preparation, track the time between completed work and a ready invoice package. For customer support, track resolution time and the number of conversations that need a person.

Score two if you have a baseline and one primary measure. Score one if the measure is clear but the baseline is rough. Score zero if success means that people liked the demo.

A rough baseline is enough to begin. Review a sample of recent work, time the current process for a week, or count the backlog. What matters is having something honest to compare.

12. Is there a safe way to stop?

Every pilot needs an off switch and a fallback process.

Know how to pause the automation, revoke access, recover the original records, and return to the current process. Keep logs so a person can see what the system read, created, changed, or sent. Decide who responds when something goes wrong.

Score two when rollback is documented and tested. Score one when stopping is possible but informal. Score zero when nobody knows how to disconnect the workflow.

A reversible pilot makes it easier to test useful ideas without betting the operation on version one.

What your score means

Use the total as a starting point:

• 19 to 24: ready for a narrow pilot, assuming no major privacy or safety concern is hiding inside a high score
• 13 to 18: promising, but fix the lowest-scoring areas before connecting systems or sending messages automatically
• 0 to 12: improve the process, ownership, records, and controls first

A total can hide a bad dependency. I would fix any zero on source information, data access, ownership, or rollback before automating.

Turn the assessment into a 30-day pilot

Once the workflow is ready, keep the first month tight.

During week one, map the process and collect real examples. In week two, define the fields, rules, permissions, approval points, and failure cases. In week three, run the system beside the current process without removing the human check. In week four, compare the result with your baseline and decide whether to improve, expand, or stop.

That decision should be boring and evidence-based. Did response get faster? Did staff correct fewer records? Did the billing package arrive sooner? Did the workflow save time after review was included?

For a longer rollout, use my 90-day AI growth pilot for Alberta small businesses as the next step.

Local context still changes the answer

A field service company working around Grande Prairie may need offline handoffs and after-hours intake. An Edmonton professional services firm may care more about document permissions and CRM discipline.

I work with businesses in Edmonton and Calgary, but I do not start by copying the same automation into each company. The useful work is understanding how information moves through that particular business, where judgment belongs, and which delay is worth fixing first.

Ready is better than impressive

A business is ready for AI when it can name one painful workflow, explain how it works, supply reliable information, keep a person accountable, and measure the result.

That is less exciting than promising an AI transformation. It is also how useful systems get built.

If your assessment exposes gaps, fix them. If it points to one clear workflow, test it with boundaries. Either result saves you from buying software before the business is ready to use it.

If you want a second set of eyes on your scorecard, book a consult with me. I will help you assess the workflow, identify the weak points, and design the smallest pilot that can prove whether AI has a business case.