David Norton Consulting, Inc.

07/07/2026

We wanted more than a slide deck.

So we built a live Oracle EBS Vision + Enterprise Command Center environment on OCI — then used OCI CLI and our AI CXO system to validate the environment, package the story, and turn it into a working client walkthrough.

If your Oracle EBS team wants better operational visibility without starting from a custom BI build, ECC is worth a closer look.

Read the story here:
https://www.davidnortonconsulting.com/oracle-ecc-vision-demo-oci-ai-cxo-system/

07/03/2026

Four pigs. One processor date. A lot of decisions.

For Tournesol Farms, building a pork pre-order system started long before there was a form on a website.

First came the operational work:

- planning the cut strategy
- organizing inventory
- creating processor cut sheets
- building customer order documents
- setting up a WordPress pre-order workflow
- making sure order notifications actually landed

That’s the kind of practical AI work we care about at David Norton Consulting — using AI to support real planning, documentation, and implementation, while keeping human judgment at the center.

The result: an online pre-order system built around how Tournesol Farms actually operates.

Read the case study in the comments.

Built by David Norton Consulting for Tournesol Farms.

06/28/2026

Got chicks?

We’re excited to share a new tool we developed for Tournesol Farms

*ChickID — an AI chick breed identifier.*

Upload a photo of a day-old chick, add a few optional details like egg color or parent breeds, and ChickID generates a breed guess, s*x estimate, key observations, and a downloadable report.

It’s built for backyard flock owners, homesteaders, hatcheries, feed stores, and anyone staring at a brooder thinking, “Okay… what exactly did we bring home?”

Try it here:
https://tournesolfarms.org/chickid/

Proud to help Tournesol Farms bring practical, farm-friendly AI tools to the community.

Tractor Supply Co.
Blue Seal Feeds
Agway Inc

05/25/2026

ERP customers do not need AI stories. They need working demos, credible options, and work they can actually review.

This week’s OpenClaw Field Notes covers real customer-facing ERP delivery work:
- stabilizing the Vision demo surface
- advancing EBS to O365 integration planning
- shipping ADP interface fixes and benefits configuration work as deployable artifacts

That is the pattern that matters: practical delivery, visible progress, and workflows that support real operators in real environments.

Read Issue 015 here: >>>>

05/05/2026

OpenClaw Field Notes — Issue 011 (Apr 20–Apr 26)

AI doesn’t fail at thinking first. It fails at the interfaces: **access, delivery, permissions, and trust**.

This week we hit real friction (the kind that stalls ex*****on) and hardened the paths:
- Repeated “email already in use” dead-ends + invisible collisions → stabilized a shared mailbox workflow
- Approvals got a deterministic home (one thread per persona)
- The portal looked connected… until someone relied on it → fixed the disconnect loop so operators can trust the surface

Principle: **Permission the path, then automate it.**
If the interfaces don’t work, nothing else matters.

Read Issue 011: Link in Comments

04/14/2026

🤖 We just published something worth reading.

Six days ago, I handed an AI agent a real infrastructure project: deploy Oracle E-Business Suite on Oracle Cloud — full HTTPS, WAF, zero manual maintenance.

It was capable. It did a lot right. And it still got stuck.

Not because it lacked ability. Because ex*****on alone isn't delivery. The agent would hit an ambiguity and stop to ask. Every time. Progress → blocker → pause → wait → repeat.

The fix wasn't a better AI. It was a second AI — with a different job.

One agent executes. One agent orchestrates. That combination finished in hours what the solo agent had been circling for days.

Total cost: ~$250 in AI compute and cloud infrastructure. Traditional Oracle consulting equivalent: $5,000–12,000+.
We wrote up the full story — what broke, what changed, and the pattern we're now using on every complex deployment.

Link in Comments>>>>

04/14/2026

We ran a real experiment: gave an AI agent the full job of deploying **Oracle E‑Business Suite on OCI** — end-to-end, with **HTTPS, WAF, production routing, and zero manual maintenance**.

It executed well. And it still took days.

Not because the AI was “bad,” but because real infrastructure work is full of ordinary blockers (IAM gaps, load balancer health checks, cert automation stalls, unstable intermediate hosts, base image defects). A senior engineer navigates those quickly because they know when to act vs. when to ask.

Our takeaway: **the operating model matters more than the agent.**

What changed everything for us was splitting the work into two roles:
- **Worker Agent** = executes the plan
- **Orchestrator Agent** = sets the tempo, breaks deadlocks, changes strategy, keeps humans out of low-level ambiguity

If you’re trying to use AI for real delivery (not demos), this pattern is worth a look:

Two AIs Walk Into a Data Center A real deployment story about Oracle E-Business Suite on OCI — and why ex*****on alone was not enough. We expected this to take a couple of hours. I’ve deployed an Oracle E-Business Suite Vision environment in that time before — but without a load balancer, HTTP...

04/14/2026

🤖 We just published something worth reading.

Six days ago, I handed an AI agent a real infrastructure project: deploy Oracle E-Business Suite on Oracle Cloud — full HTTPS, WAF, zero manual maintenance.

It was capable. It did a lot right. And it still got stuck.

Not because it lacked ability. Because ex*****on alone isn't delivery. The agent would hit an ambiguity and stop to ask. Every time. Progress → blocker → pause → wait → repeat.

The fix wasn't a better AI. It was a second AI — with a different job.

One agent executes. One agent orchestrates. That combination finished in hours what the solo agent had been circling for days.

Total cost: ~$250 in AI compute and cloud infrastructure. Traditional Oracle consulting equivalent: $5,000–12,000+.
We wrote up the full story — what broke, what changed, and the pattern we're now using on every complex deployment.

👉 Read it here: https://www.davidnortonconsulting.com/two-ais-walk-into-a-data-center/

---

04/09/2026

Paul Pipeline just published **OpenClaw Field Notes — Issue 008** (Mar 30–Apr 5).

Most support delays aren’t caused by hard problems—they’re caused by broken access paths, missing context, and tooling that only works when everything goes perfectly.

This week’s unlock: **agents execute directly against client environments under consultant‑authorized access**—no more “tell me what to run / paste results back” loops.

We also tightened control-plane reliability (fewer silent failures, fewer wrong turns) and kept moving the assistant closer to the operator surface (SQL Developer) where context-switching usually introduces mistakes.

Read Issue 008 here:
https://www.davidnortonconsulting.com/openclaw-field-notes-issue-008-if-access-isnt-deterministic-support-doesnt-scale/

If your Oracle EBS support team is fighting access rituals, attachment overload, or context gaps—reach out.

04/08/2026

We just published OpenClaw Field Notes — Issue 007.

This week’s focus: building *repeatable systems* that work in real operations—not impressive demos.

We shipped a control plane for multi-agent delivery (tools + session visibility + work-in-progress in one place), and we continued advancing support workflows designed for real-world volume: tickets, attachments, approvals, and fragmented context.

If you support complex systems—especially Oracle EBS—this is the kind of ex*****on layer that reduces reconstruction time and increases throughput.

Read it here: >>>>>Link in comments