E-commerce AI implementation checklist: before you start a project

A significant number of AI implementation projects fail not because the technology doesn't work, but because the project was scoped based on assumptions about the store's data, integrations, or team capacity that turned out to be wrong.

This checklist is for store owners and operations managers who are considering an AI project but haven't committed yet. Answering these questions honestly before you start a conversation with any vendor — including us — will save time and prevent frustration.

Twelve questions to work through

• 01
  What specific workflow are you trying to improve?

  "We want AI" is not a project. "We want to reduce the time our team spends answering order status queries, which currently takes about two hours a day" is a project. Start with a specific, measurable workflow.

• 02
  Can you describe the workflow in enough detail to build it?

  If you can't describe what happens step by step in the workflow you want to automate, that's a gap to fill before scoping. Document the current process first.

• 03
  Do you have the data the project requires?

  Support automation needs documented policies and order data access. Recommendations need product data and purchase history. Cart recovery needs order and email tool integration. Identify what's needed and whether you have it.

• 04
  Is that data accessible and reasonably clean?

  Having the data and being able to use it aren't the same thing. Policies in someone's head, inconsistent product tagging, and order history spread across a migrated platform all require cleanup before they can support automation.

• 05
  Who will grant API access to your platforms?

  Name the specific person who can provide API credentials or install a custom app on your Shopify or WooCommerce store. If that person is unavailable for the next month, factor that into your timeline.

• 06
  Who will own the agent after launch?

  Someone needs to monitor it, update it when policies change, and field edge cases. If nobody has capacity for this, either build that into a maintenance retainer or reconsider the timeline.

• 07
  What does your team know about the project?

  If you're building a customer support agent, your support team needs to know before it launches, not after. The same applies to any agent that changes how your team interacts with customers. Involve them early.

• 08
  What is your escalation model?

  For any agent that touches customers, define upfront: what happens when it can't answer? Who gets the escalation? Via which tool? What context does the escalation include? This needs to be decided before build, not after.

• 09
  Have you checked for GDPR implications?

  If the agent will process personal data — order information, email addresses, browsing behaviour — you need a data processing agreement with your vendor. This is a legal requirement under GDPR, not optional. Verify your vendor provides one.

• 10
  What does success look like, and is it measurable?

  "The agent works" is not a success metric. "Tickets requiring human response falls below X per week" or "first response time under 5 minutes" are measurable. Define this before you start so you can evaluate it after.

• 11
  What is your minimum viable scope?

  If the project hits complications, what's the smallest version that would still be useful? Starting narrower and expanding is usually better than scoping ambitiously and discovering mid-project that prerequisites weren't in place.

• 12
  What would make you cancel the project?

  Think through your exit conditions now. If a platform integration turns out to require three months of custom development work, is that still worth it? Knowing your limits in advance makes scope conversations faster and cleaner.

What to do with this list

Work through it before any vendor conversation, including a discovery call. You don't need complete answers to everything — gaps are useful to surface early. But the questions where you can't produce even a rough answer are the ones most likely to cause problems mid-project.

The discovery call is the right place to go through unresolved questions with someone who can help you think through the implications for your specific situation.

Ready to talk? Book a discovery call — it's free and takes about 30 minutes.