I was invited to speak to a group of Broadband Operators yesterday on AI yesterday. I am not an expert. I have spent most of 2025 taking classes, watching demos, selling AI food, and was involved in a few really cool AI projects (thanks to Strolid, Inc. & Thomas McCarthy-Howe, a certified genius!).
Here are 3 lessons:
- You better have an AI Plan that starts now.
- Most people are measuring the wrong things in AI POC’s. That’s why the dissatisfaction.
- Every AI project is really a Data Project. Without good data, useless AI.
Agentic AI has over 2000 providers. It is a mess for 2 reasons: (1) it is too easy to be an Agentic AI provider and (2) 2000 of these providers lack Data, an integral part of the Agent being effective.
On the customer side, these projects misfire because most businesses are SMB (small business) and small businesses do not have data that is easily accessible. It is usually stored in some proprietary silo, like ConnectWise or Stripe or Quickbooks or customer software. Without data the AI Agent is a generic robot that might as well be the paperclip from Microsoft.
These businesses normally do not have written procedures and documented workflows. How can you add AI to a process if you do not know what that process or workflow is?
Here are six pieces for your AI Plan:
- Let some of your employees play with AI and meet with them regularly to see what they like and don’t like.
- Plan to upskill your workforce.
- Map your workflows. Have everyone start documenting their processes.
- Collect data – PDFs, FAQs, documents, workbooks, manuals, knowledge bases, everything.
- Learn to Prompt!
- Measure and then realize you are probably using the wrong yardstick.
Need help? Be happy to spend 20 minutes chatting with you. Make an appointment to speak with me: HERE
If you want a copy of the slides, please email me at peter (at) rad-info dot net.
Big thanks to Gapingvoid for always producing great cartoons, drawings, books and insights.