This week, a developer tested an AI triage assistant against a real VS Code Issue (#339955) — the AI independently produced accurate classifications, but the project's design principles deliberately left a safeguard: AI doesn't make the final call for maintainers.The triage assistant does four things: reads Issue content, classifies the issue type (Bug, Feature Request, Inquiry, etc.), suggests priority and labels, and drafts a reply. The model runs DeepSeek-V3.2 on the Lanyun YuanShengDai MaaS platform, accessed via an OpenAI-compatible API.The MCP referenced in the test is a protocol that lets AI models invoke local tools. The Issue describes an MCP service disappearing from the workspace after a user upgrade — the AI independently pinpointed the root cause.
Industry view
We've noticed that this 'AI first-pass triage + human final decision' division of labor is becoming a de facto standard across multiple domains. Support tickets, bug reports, internal approvals — all isomorphic scenarios: high information volume but each item is easy, requiring classification before handling. AI eats 80% of the repetitive work; humans focus on the 20% that requires judgment.But there are dissenting views. Members of the developer community point out: for ambiguous or contested Issues, AI tends to produce 'plausible-sounding but off-target' recommendations, which actually increases review overhead; when Issue descriptions are incomplete, AI's priority judgment may misalign with project strategy — priority is about resource trade-offs, not technical analysis.The more pragmatic view: the value of these tools isn't how smart the AI is, but compressing 'read 100 Issues' into 'read 20 AI summaries'. What gets saved is time, not judgment.
Impact on regular people
For traditional enterprise IT: this is a template for AI adoption — not replacing any role, but consuming the most time-consuming triage step in the workflow, lightening the load for downstream human work.For individual careers: your future workflow will very likely include a step where 'the AI assistant goes through it first'. It won't make the call for you, but it'll pull you out of the manual grind of 'read 100 then decide'.For the consumer market: similar 'AI triage' will permeate customer service, food delivery complaints, e-commerce after-sales. Consumers will encounter AI first, then humans — this is already happening.
