The fourth single source of truth
Power BI already contains much of the business meaning AI needs. The hard part is deciding which meaning deserves to be reused.
Read the pieceAI, ERP, data, and the work behind them
Peter Nicholson.
I write about AI, ERP and data, and what they change for people doing the work.
I'm an IT and data nerd with years in manufacturing sales. Both sides shape what I build, question and write about.
From the notebook
Latest writing
ERP decisions, useful data, and AI that has to work beyond a demo. Get occasional notes by email
Latest piece · ERP
What a Sage migration exposes before the new ERP goes live
Acumatica published a useful comparison for businesses replacing Sage 50 or Sage 300 this week. It describes two broad routes. A finance-first move may suit a business whose operational systems already work.
Read the piece
Should you downgrade your ERP?
Eighty-odd episodes of this podcast and almost every one has pointed the same direction. Outgrow the spreadsheets. Get off the paper. Upgrade the system. Migrate
Read the piece
We put an AI on the podcast. Then we talked about what actually breaks.
We opened this week's episode by handing the microphone to an AI. Not a clip, not a pre-recorded bit — a live voice model,
Read the piece
Your second ERP migration is a different project
The thing keeping you on a system everyone hates isn't the system. It's the scar tissue from the first migration. And round two is a different project — easier, better understood, and mostly a matter of not squandering everything you learned the first time.
Read the piece
Nobody models the Fabric bill until it lands
Fabric pricing guides focus on SKUs and thresholds, but the real cost emerges after deployment. In Microsoft Fabric, shared capacity means every refresh, model, and pipeline draws from one pool—so ongoing spend is driven less by licensing choice and more by architecture and governance.
Read the piece
Semantic model sprawl is a symptom, not the disease
Semantic model sprawl isn't caused by having too many Power BI models. It's caused by having too many competing definitions of the business. Fix the numbers everyone trusts, and the model count largely takes care of itself.
Read the piece
Keeping your ERP upgrades boring
ERP upgrades shouldn't be exciting. If yours are, you're probably paying for decisions you made years ago. Here's why boring upgrades are the mark of a healthy ERP.
Read the piece
What to extract before you replace your ERP
When you replace an ERP, you're ending a ten-year conversation between people and software. That conversation is messy, half-finished, and full of
Read the pieceSelected AI writing
From ERP systems to small experiments
I build and test things, then write about where AI helps and what still needs a human decision.
ERP systems
Teaching an LLM to use Acumatica
Connecting an AI assistant to a live ERP tenant, and testing what it can answer.
Read the piece Business dataThe semantic model finally has its moment
Why an LLM needs business definitions, not just another pile of documents.
Read the piece ExperimentsFour apps and a Mac Mini under the stairs
The small apps that survived an experiment with always-on AI agents.
Read the pieceGet in touch
Have something to share?
A different take on something I've written, a podcast guest I should know, or an invitation to speak. Send me a short note.
Speaking and podcast conversations