OpenClaw Field Notes – Issue 003 — WordPress Payments, Oracle E-Business Suite Payroll, and Infrastructure Lessons from the Field

OPENCLAW FIELD NOTES — ISSUE 003

Date Range: Feb 24 – Mar 1
Platforms: Oracle E-Business Suite, WordPress, Square Payments, Microsoft 365, HAProxy, Linux
Thesis: AI expands what small engineering teams can realistically build, analyze, and support.


TL;DR

This week we built a custom WordPress + Square payment integration, investigated a year-end accounting anomaly in Oracle E-Business Suite Payroll, stabilized an Oracle EBS SSL cutover using HAProxy, and tightened how our assistant behaves inside Microsoft Teams and Microsoft Planner workflows.

A clear theme emerged across all of it:

AI doesn’t replace the work. It expands what a small consulting operation can realistically build, analyze, and support.

Some work moved faster.
Some work became possible that normally wouldn’t be worth building.

And a few automation ideas ran into a simple truth:

If the underlying platform doesn’t expose the right primitives, automation stops.


Platforms Involved This Week

The projects below touched a mix of systems commonly used by our clients:

  • Oracle E-Business Suite (Payroll / ECC / infrastructure)
  • WordPress
  • Square Payments
  • Microsoft 365 (Teams + Planner)
  • HAProxy / Linux infrastructure
  • SELinux

These field notes reflect real operational work across those platforms.


What We Shipped This Week

  • Custom WordPress + Square deposit checkout system with verified webhooks
  • Root-cause explanation for Oracle E-Business Suite payroll → GL year-end anomalies
  • Stable SSL termination architecture for Oracle EBS using HAProxy
  • Microsoft Teams assistant behavior tightened for external meeting contexts

Capability Expansion

Work That Normally Wouldn’t Exist

Some projects only happen because modern tooling dramatically reduces the time needed to experiment and build.

This week’s best example was a custom WordPress payment workflow built on Square.

WordPress + Square Deposit Checkout

Instead of relying on another paid WooCommerce plugin, we built a site-specific WordPress plugin to control the checkout lifecycle.

The goal was simple:

Create a deposit system that is auditable, predictable, and independent of fragile plugin ecosystems.

Deposit Logic Built Into Checkout

Product selections generate Square checkout links while enforcing rules such as:

  • deposit percentages
  • minimum deposit amounts
  • consistent pricing logic

Sandbox Testing

A site-wide TEST MODE banner ensures sandbox payments cannot be confused with real transactions.

Production-Grade Webhooks

A lightweight MU-plugin webhook handler was implemented to:

  • bypass heavy security plugin overhead
  • verify Square webhook signatures
  • process events safely and quickly

Order Lifecycle Tracking

A small order-tracking table records the checkout lifecycle:

checkout_creating
→ checkout_created
→ deposit_paid

Each request receives an order_id so webhook events can be correlated later.

The result is a controlled payment workflow inside WordPress without locking the site into a paid plugin dependency.


Efficiency Multipliers

Work That Still Had to Happen — But Much Faster

Most consulting work still requires deep domain knowledge.

What AI changes is how quickly analysis and iteration can happen.

Oracle E-Business Suite Payroll Rules

Paid leave logic in Oracle E-Business Suite Payroll often behaves less like a simple formula and more like a rules engine embedded inside payroll processing.

This case involved:

  • weekly thresholds
  • bi-weekly payroll periods
  • date-earned logic
  • conditional reductions to regular hours

The Rule Clarified

For each paid leave entry:

  1. Determine which work week the date earned falls into.
  2. Calculate weekly total:
qualifying_hours + paid_leave_hours
  1. Compare the result against scheduled hours.

If:

total ≥ scheduled_hours

Regular hours must be reduced.

Otherwise they remain unchanged.

AI accelerated SQL queries, Fast Formula drafting, scenario modeling, and debugging — but the final implementation relied on deep domain expertise in the Oracle payroll engine.


Payroll → GL Year-End Anomalies

Another investigation examined why some payroll transactions posted with a GL date equal to the pay period end while others posted in the next year.

What the Data Showed

Most entries behaved normally.

The entries posting to the following year were associated with post-payroll corrections, including:

  • payment voids
  • reversals
  • re-QuickPays
  • reissued checks

The Explanation

Standard payroll costing posts to the pay period end date.

Corrective payroll actions can derive accounting dates from the later transaction date, pushing entries into the next accounting period.

Recommended Controls

  • define a cutoff policy for void/reissue activity
  • require finance approval for post-year-end corrections
  • run a pre-Transfer-to-GL validation report

Operational Compression

Meeting Recap → Microsoft Planner Tasks

We established conventions for automatically converting meeting discussions into tasks inside Microsoft Planner.

Plan: Administrative Operations
Bucket: Inbox / Intake

meeting discussion
→ recap
→ Planner task
→ execution

The goal isn’t full automation — it’s reducing friction between discussion and action.


Human Bandwidth Expansion

Microsoft Teams Assistant Behavior

We refined how our assistant introduces itself when external participants are present in Microsoft Teams.

“Jethro Shepherd, Chief of Staff (AI), David Norton Consulting.”

In external conversations the assistant now:

  • limits operational disclosure
  • defaults to minimal information sharing
  • defers to human approval when necessary

WhatsApp Watchdog Experiment

We attempted to deploy a message monitoring watchdog for WhatsApp.

The design was simple:

  • read recent messages
  • acknowledge or triage them
  • route follow-ups into operational workflows

The experiment failed for a simple reason.

The messaging API endpoint expected to retrieve message history actually creates polls.

pollQuestion required

The primitive required for message monitoring does not currently exist in the integration.

Automation requires both logic and platform capabilities.


Infrastructure Notes

Oracle E-Business Suite SSL Cutover

An HTTPS migration for an Oracle E-Business Suite environment was stabilized this week.

The initial plan used Oracle HTTP Server wallets with Let’s Encrypt certificates.

Certificate issuance worked, but OHS under OPMN would not reliably start with imported wallets.

The Pragmatic Solution

We implemented HAProxy TLS termination in front of the EBS stack.

/ecc/   → 127.0.0.1:7776
default → 127.0.0.1:8000

SELinux initially blocked backend connections.

haproxy_connect_any

Additional fixes included:

  • removing a cookie domain rewrite rule causing login loops
  • aligning EBS profile options to emit HTTPS URLs
  • correcting mixed-content image references

Sometimes the best architecture is simply the most stable one available.


Wins

  • WordPress + Square deposit system deployed with verified webhooks
  • Oracle E-Business Suite payroll accounting anomaly explained
  • Stable HTTPS architecture implemented for EBS using HAProxy
  • Microsoft Teams assistant behavior improved
  • Planner task routing conventions established

Principle of the Week

Infrastructure correctness beats theoretical purity.

The original SSL design should have worked inside Oracle HTTP Server.

But “should work” and “operates reliably in production” are two very different things.

The pragmatic solution — HAProxy terminating TLS in front of the stack — proved far more stable.

Reliability matters more than elegance.


Next Week Focus

  • Enable message-history retrieval for the WhatsApp watchdog experiment
  • Complete meeting recap → Planner task automation pipeline
  • Finish ECC authentication routing validation
  • Harden Square checkout correlation logic
  • Finalize payroll validation scenarios

About OpenClaw Field Notes

OpenClaw Field Notes is a weekly build log documenting how AI tools integrate with real consulting work across Oracle E-Business Suite, WordPress, Microsoft 365, and modern infrastructure stacks.

AI is used as a tool.

Humans remain accountable.

Next Field Notes: new systems, real debugging, and the operational realities of AI-assisted consulting.

Leave a Comment