Here's something different. I recently built a 30-60-90 day market intelligence report focused on the ERP space — and I did it entirely with AI.
Not a rough draft. Not a brainstorming exercise. A structured, sourced, verifiable analysis of what's happening in ERP right now, what's likely coming in the next 30, 60, and 90 days, and what it means for the people building on these platforms.
You can view the full report here.
The key was the prompt.
The Methodology
I used a highly optimized prompt with three things baked in that most people skip: a 5-level certainty system, anti-hallucination safeguards, and live web research.
The certainty system is the backbone. Every finding in the report gets tagged with a confidence level — from confirmed announcements with source citations all the way down to speculative analysis. If something is marked "CONFIRMED" or "ANNOUNCED," there's a source behind it. If it's a projection, it's labeled as one. No ambiguity about what's verified and what's inferred.
The anti-hallucination safeguards are just as important. When you ask AI to generate market intelligence, the risk isn't that it gives you nothing — it's that it gives you something that sounds right but isn't. The prompt is designed to constrain the model so it sticks to verifiable claims, flags uncertainty explicitly, and doesn't fill gaps with confident-sounding fiction.
And live web research ties it all together. The analysis isn't based on stale training data alone. It pulls in current sources, which is what makes a 30-60-90 day window realistic instead of theoretical.
What Showed Up
One finding worth calling out: NetSuite's 2026.1 release includes an Analytics Warehouse AI Connector and MCP support for external AI — Claude, ChatGPT, Copilot. That's a significant move. It means NetSuite is opening the door for third-party AI to interact with its data layer natively, not through workarounds.
That's the kind of insight a report like this surfaces — not just "what happened" but "what does this mean for the next quarter." And when it's tagged with a source and a confidence level, you can actually make decisions based on it instead of treating it as one more opinion in a noisy feed.
Prompt Engineering Is the Work
I keep coming back to the same realization: prompt engineering has become central to everything I do. Whether it's a pure AI project, an embedded AI solution, or just using AI to boost development productivity — the quality of the output is directly proportional to the quality of the prompt.
A 30-60-90 day market intelligence report isn't something most people would think to build with AI. But with the right prompt architecture — certainty scoring, hallucination guards, source requirements, structured output — you get something that's not just useful but trustworthy. And trustworthy is the bar that matters.
If you're doing strategic planning, competitive analysis, or just trying to stay ahead of what's happening in your space, this kind of approach is worth experimenting with. The model does the heavy lifting. Your job is to design the guardrails.
This article was originally published on LinkedIn.