I wholeheartedly agree with this part of your article. In fact, I would expand it further to include a few other skills that are about to become extremely valuable:
- Business acumen and knowledge about how your business runs - SOP expertise
- Business modeling, data modeling, business logic, spreadsheet expertise (Coda stuff!)
- Knowledge curation - managing the key documents about policies, procedures, & how-to
- AI Agentic tools - building AI “experts” that do business processing better than people
And these skills are mostly vested in middle-layer (middle-aged?) managers who don’t know how valuable these skills are about to become.
My exec-coaching business has been swamped with new opportunities this month!
Business managers are afraid for their jobs, their careers, their teams, and their line of business.
They see the incoming “AI Tsunami” sweeping away all the safety, certainty, and stability they have been living with throughout their careers. All the media hype and hoopla do not help.
Corporations are issuing “AI First” mandates that require employees to improve productivity through AI or face poor reviews (or worse), but without a lot of guidance about how to do this. (And admitting your ignorance, uncertainty, and fears is not encouraged, alas).
But my biggest clients are facing a huge explosion in their IT departments’ “Technical Debt” because the amount of code being produced has exploded, with almost no control over formal testing, certification, deployment safety, etc. A ‘shadow IT’ plague (as they see it).
Placing vibe-coding tools in the hands of ‘citizen developers’ who cannot understand the code being generated, are not trained in the art of writing clear requirement specifications, regression testing, or source code management, is like handing out crates of hand grenades and bottles of tequila at the xmas party!
Managers who use AI to improve their personal productivity are doing well. Those who naively use AI to produce complex applications in python or js to solve business problems - not so much!
So my own stance is that vibe-coding (for the masses) is not yet ready. But vibe-(no)coding (‘vibe-making’?) using nocode or spreadsheet tools is far more reliable and safe. Especially since the resulting workflows are understandable by the users who make them.
Expressing the “problem model” clearly as a pile of big hairy english prompts is not as good as using more formal process and algorithm analysis tools.
Getting your AI response as a great heap of procedural code you cannot understand is not as good as getting an AI response in the form of a Notion, Coda, Excel, or N8N solution, which is far more useful (and safer for the business).
Coda MCP has the potential to be a significant part of that approach. But for now, it is not ready for prime time yet:
- it is only available to a limited set of beta testers
- it is undergoing rapid evolution in response to that testing
- it is only available to users of copilot, antigravity, kaggle etc
- ie: people who know how to marshal and deploy the resulting agents
- not people who rely on Coda to build their business workflows
I applaud you, @Bill_French, for your work on mcpOS, which addresses many of my concerns above.
Your recent articles have given me much to think about, and I will respond here with a more considered and detailed reaction over the weekend.
Respect
➤𝖒𝖆𝖝