LinkedInยทWednesday, 22 July 2026ยท22 Jul 2026
Is Agentic AI in production just a fancy game of roulette? ๐ฒ Right now, everyone is eager to deploy ๐๐ด๐ฒ๐ป๐๐ถ๐ฐ ๐๐ ๐๐ผ๐ฟ๐ธ๐ณ๐น๐ผ๐๐โฆ
Xavier Geerinck
CTO @ Scrydon | Sovereign Data & AI Layer | Agentic & Ontologies | ๐ก๏ธ
Is Agentic AI in production just a fancy game of roulette? ๐ฒ
Right now, everyone is eager to deploy ๐๐ด๐ฒ๐ป๐๐ถ๐ฐ ๐๐ ๐๐ผ๐ฟ๐ธ๐ณ๐น๐ผ๐๐ - AI models that make their own step-by-step decisions to complete complex tasks.
๐๐ฒ๐ฟ๐ฒ'๐ ๐๐ต๐ฒ ๐ฐ๐ฎ๐๐ฐ๐ต:
Traditional code is ๐ฑ๐ฒ๐๐ฒ๐ฟ๐บ๐ถ๐ป๐ถ๐๐๐ถ๐ฐ (Input A always yields Output B).
Agentic AI relies on ๐ฝ๐ฟ๐ผ๐ฏ๐ฎ๐ฏ๐ถ๐น๐ถ๐๐๐ถ๐ฐ ๐ฟ๐ฒ๐ฎ๐๐ผ๐ป๐ถ๐ป๐ด (it guesses the best next step based on probabilities).
If your AI agent makes 5 sequential decisions to complete a workflow, and each step has a 95% success rate:
โข Step 1: 95%
โข Step 2: 90%
โข Step 3: 85%
โข Step 4: 81%
โข ๐ข๐๐ฒ๐ฟ๐ฎ๐น๐น ๐ฝ๐ฟ๐ผ๐ฐ๐ฒ๐๐ ๐๐๐ฐ๐ฐ๐ฒ๐๐ ๐ฟ๐ฎ๐๐ฒ: ~๐ณ๐ณ%
That means ๐ญ ๐ผ๐๐ ๐ผ๐ณ ๐ฒ๐๐ฒ๐ฟ๐ ๐ฐ ๐ฟ๐๐ป๐ ๐ฐ๐ผ๐๐น๐ฑ ๐ฟ๐ฎ๐ป๐ฑ๐ผ๐บ๐น๐ ๐ณ๐ฎ๐ถ๐น just by chance - even if nothing in your environment changed.
Are we genuinely ready to put systems that "roll the dice" on every step into core production environments? Or are we setting ourselves up for impossible-to-debug failures on Day 4?
๐'๐ฑ ๐น๐ผ๐๐ฒ ๐๐ผ ๐ต๐ฒ๐ฎ๐ฟ ๐ณ๐ฟ๐ผ๐บ ๐ฑ๐ฒ๐๐ฒ๐น๐ผ๐ฝ๐ฒ๐ฟ๐ ๐ฎ๐ป๐ฑ ๐๐ ๐น๐ฒ๐ฎ๐ฑ๐ฒ๐ฟ๐:
โข How are you handling non-deterministic failures in your AI agents?
โข Are guardrails, human-in-the-loop, or deterministic fallbacks enough?
โข Or should core business logic remain strictly traditional code for now?
Drop your thoughts below! ๐
๐ฌ 2
View on LinkedIn Cross-referenced
Related on the wire
AI this, AI that, but now finally proven in production! It resulted in -60% Memory for us and -50% CPU! How? I've been reading andโฆ
What if you could OWN complex AI problem solving on your own premises / airgapped / hybrid cloud environment? That's what we askedโฆ
A ๐๐ฎ๐ป๐ฑ๐ฏ๐ผ๐
๐๐ต๐ฎ๐ ๐ผ๐ป๐น๐ ๐ต๐ผ๐น๐ฑ๐ ๐๐ต๐ฒ๐ป ๐๐ต๐ฒ ๐บ๐ผ๐ฑ๐ฒ๐น ๐ฏ๐ฒ๐ต๐ฎ๐๐ฒ๐ isnโt a sandboxโitโs a request. We watched frontierโฆ
Big move in the AI agent space: ๐จ๐ฆ ๐ฃ๐ฎ๐๐ฒ๐ป๐ ๐ญ๐ฎ,๐ฒ๐ณ๐ฌ,๐ฌ๐ฐ๐ฑ ๐๐ญ ๐๐ฎ๐ ๐ผ๐ณ๐ณ๐ถ๐ฐ๐ถ๐ฎ๐น๐น๐ ๐ฝ๐๐ฏ๐น๐ถ๐๐ต๐ฒ๐ฑ. The patentโฆ
๐ก๐ฉ๐๐๐๐ is amazing, but their biggest ๐ฎ๐ฑ๐๐ฎ๐ป๐๐ฎ๐ด๐ฒ isn't just hardware, ๐ถ๐'๐ ๐ฎ๐ฐ๐๐๐ฎ๐น๐น๐ ๐๐จ๐๐. It's the reasonโฆ
A crucial reality check from the FLI report! ๐ ๐ถ๐๐๐ฟ๐ฎ๐น'๐ ๐ฟ๐ฎ๐ป๐ธ๐ถ๐ป๐ด ๐๐ต๐ผ๐๐ ๐๐ต๐ฎ๐ ๐ฟ๐ฒ๐ด๐๐น๐ฎ๐๐ถ๐ผ๐ป ๐ฎ๐น๐ผ๐ป๐ฒโฆ