George Hepworth's Blog

Former longtime Access MVP, current President of the Access Pacific User Group, and friend of the blog, George Hepworth, has been posting new content to his site recently:

Blog — Grover Park Consulting

Grover Park Consulting

Incorporating AI into the Access Dev Workflow

George's website includes the following summary of his current interests under the banner, "NOW EXPLORING":

George is actively incorporating AI assistants into the Access development workflow — primarily Claude and Claude Code — and will be sharing findings and techniques here as that work develops.

I'm looking forward to learning and incorporating ideas from George's AI workflow into my own.

Watch for Future Articles in my Week in Review

I've been remiss in not including his content in my Week in Review series. I will remedy that moving forward.

In the meantime, here is a list of his most recent articles with one-line descriptions (via Claude Opus 5) to get you caught up on what you (and I) may have missed:

• Wanna say something? Go right ahead and say it. I'm listening.: Announces that the GPC blog now supports reader comments — built "with Claude Code's help" — and invites readers to start a dialog rather than treat the site as a billboard.
• Open Scaffold Templates: A Midstream Correction: Explains why he scrapped his rule-heavy approach to AI database templates after concluding that "an AI agent is not a compiler," pivoting to specify outcomes instead of procedures.
• My AI assistant called a "prefix" a "marker": Traces why Claude Code kept swapping database terminology in drafted instructions back to his own plain-language rule, illustrating that inexplicable AI output almost always has a findable upstream cause.
• What building an "AI wizard" cost me — and it wasn't in tokens: Shares what building an AI-assisted template library taught him about user-facing prompts: strip the jargon and rationale from the question and tuck the reasoning behind an optional "Tell me more."
• The Benefit of AI and the Risk of AI Are the Same Thing: Argues that AI's real value is lowering the cost of exploring beyond what you already know — and that the same leverage erodes the judgment you need to catch its errors.
• Why I Don't Mind That AI Makes Mistakes: Compares three decades of learning Access by trial and error to working with AI agents, making the case that corrections are the feedback loop that forces reasoning into the open.
• Why I Track AI Costs I Don't Pay: Explains why token accounting matters even on a flat subscription, after measurement revealed roughly 94% of tokens go to re-reading conversation history rather than producing new work.
• Design first, build second: Open Template Scaffolds is now public: Introduces a free, open-source library of AI-readable templates for Access and SQL Server that lets you review and approve a database design before the AI builds anything.
• The $265 Query Fix: What Two Tools Couldn't Show Me About My Own Work: Recounts how his own DevPulse tracker surfaced a $265 session cost misattributed to a single trivial task — an accounting error Azure DevOps and Linear both showed as plausible.
• Is Your AI Assistant a Consultant or an Agent?: Contrasts the episodic "ask a chatbot" model with agentic AI that reads your schema and runs your code, and lays out the version control, scoped permissions, and context docs that make the latter safe.

The Parable of the Beautiful Tree

This article is a bit older (published September 8, 2024), but anyone who's spent any time answering reader questions on an Access forum will immediately appreciate the metaphor:

The Parable of the Beautiful Tree — Grover Park Consulting

Grover Park Consulting

^{Special thanks to Access MVP Colin Riddington for pointing me at George's resurrected blog!}