What to prepare before asking for AI help
A better adviser conversation starts with concrete inputs, not a general request to "make us more efficient with AI".
1. One recurring workflow worth assessing
Write the workflow in one sentence. For example: "prepare a weekly list of stale opportunities for sales review" is clearer than "automate sales". "Draft suggested updates for outdated website pages" is clearer than "use AI for marketing".
2. Current tools, systems, and data sources
List the tools the workflow already uses: CRM, website CMS, spreadsheets, email, booking calendar, proposal tool, project board, analytics exports, knowledge base, or shared drive. Then mark which source is trusted when records disagree. The AI Data Readiness Checklist for NZ Businesses gives a deeper review.
3. Examples of delays, rework, or manual admin
Bring three to five real examples: a missed follow-up, stale CRM record, manually assembled report, outdated page, memory-based support answer, or task that depends on the owner remembering every step.
4. Actions that must stay under human approval
Name the stop signs before anyone designs an agent. For most small businesses, AI should not independently send sales emails, publish public pages, change pricing, make customer promises, approve discounts, update sensitive records, or handle privacy-sensitive decisions. Use AI Approval Gates for Business Automation to map boundaries.
5. A simple measure of whether AI helped
Choose one practical measure: cleaner records, fewer missed follow-ups, faster draft preparation, more reliable reporting, fewer manual checks, or clearer owner visibility. Avoid measuring success only by output volume.