Appendix: Scrapped Ideas
A graveyard of brainstormed AI use cases that were evaluated and intentionally discarded.
During the initial brainstorming phase for VisiCore's AI strategy, several ideas were proposed that sounded great on paper but failed to survive critical analysis. They have been permanently scrapped.
We document them here to prevent re-hashing bad ideas in the future.
Prompt Firewall
What it was: A proxy/gateway that sits between customer users and external LLMs (like OpenAI/Anthropic). When an engineer types a prompt, the "Firewall" inspects it before it leaves the company boundary to strip out PII, secrets, or block malicious prompt injections.
Why it was proposed: Companies are terrified of engineers accidentally pasting API keys or customer data into ChatGPT.
Why we killed it: Cloudflare (via AI Gateway), Microsoft, and Palo Alto Networks are building this natively into enterprise network gear. We cannot and should not try to compete with trillion-dollar cybersecurity vendors on building a firewall product.
GoatRoute
What it was: Using an LLM to automatically generate the complex JavaScript/Regex routing logic inside Cribl (e.g., "Route all Palo Alto firewall logs containing 'deny' to Splunk index A, and drop everything else").
Why it was proposed: Writing Cribl routing rules by hand is tedious, requires specialized knowledge, and is highly prone to syntax errors.
Why we killed it: This is a "Runtime-AI" feature that belongs natively inside the platform. Cribl Copilot has essentially already built this natively into their product (or will very soon). If we built it as an external tool, Cribl would commoditize it immediately, making our product obsolete.
Alert Fatigue Assassin
What it was: SOC and NOC noise reduction. The agent analyzes historical alert data to safely deduplicate, correlate, and suppress noise without missing true-positives.
Why it was proposed: Analyst burnout is real, and missing a critical breach because it was buried in 10,000 false-positive alerts costs millions.
Why we killed it: Vince called out that Splunk AI Assistant 2.0 with Agent Mode is now agentic and is tackling this natively (announced Cisco Live 2026). It is built directly into the platform (Splunk Enterprise Security). Trying to build an external tool for this is redundant and cannot compete with native vendor integration.