Webinar recap
Presented by Chris Obdam, CEO at Betty Blocks
On April 28th, 2026, Chris Obdam hosted a strategic session on one of the most pressing problems in enterprise technology: why AI initiatives keep failing to reach production - and what the only viable path forward looks like. The session opened with a striking reality check. Almost 70% of businesses are currently using AI, yet 80% report no measurable impact on either productivity or employment.
Chris shared a real-world example from a large enterprise where 4,000 AI initiatives had been started - and only 5 made it to production. The core argument: the prompt is not the problem. The problem is the artefact it produces. Vibe coding tools generate raw, ungoverned code that IT departments cannot approve, cannot audit, and cannot safely connect to enterprise systems. The solution is not to slow down - it is to put a governed layer between the prompt and the output. That layer is Prompt-to-Meta-to-Code, and it is what Betty Blocks has been building for 20 years.
Frequently Asked Questions
Questions from the webinar
Can a senior developer read the exported code without knowing Betty Blocks?
The goal is that the code should look as good as and ideally better than what a senior developer would have written themselves. It is standard React and WebAssembly. You do not need to learn Betty Blocks to use it. Download it, send it to your architect, have them check it. That is exactly what we encourage.
How can Betty Blocks help IT trust the code that comes out of the platform?
Betty Blocks does not use AI to generate code. That is the key distinction. When you explain to an IT department how the meta model works - that there is no dark code, no AI-generated vulnerabilities - it opens a very different conversation. What we have found is that senior architects and IT managers genuinely like what they see when they understand the architecture. It is a good conversation to have.
If our ERP data structure changes, does the app break?
The blueprint always adapts. It is a visual representation of your app structure that updates automatically when the application changes. The same applies to requirements - with AI, those stay in sync rather than going in a drawer and becoming irrelevant. If ERP data changes, we version it and surface it through the model.
What's a realistic timeline from first prompt to production?
The SaaS subscription tracker built live during the session could have gone to production the same day. The quality is there from the first build. The variable is IT approval and that is where the governance layer does its work, because IT can see exactly what was built, who built it, and what the code looks like. Getting to production is a matter of approval, not rework.
What does IT handover look like when a citizen developer builds something?
It is more of a functional handover than a technical one, because IT always has visibility. Through the Software Portfolio in My Betty Blocks, IT managers can see every application being generated - who built it, what it does, and whether it is still being used. The handover is a matter of transferring ownership in the system, not re-explaining what was built.
Why do most AI initiatives never reach production?
Most AI tools generate raw, ungoverned code that IT cannot approve, audit, or safely connect to enterprise systems. As Chris Obdam showed, one large enterprise started 4,000 AI initiatives - only 5 reached production. The problem isn't the prompt. It's the artefact it produces. Betty Blocks solves this with Prompt-to-Meta-to-Code: capturing intent as structured metadata and compiling trusted code from that model - giving business users AI speed and IT the governance they need.