AI agent projects should start by asking whether they need an agent at all
kashifmanzoor · x · 2026-07-21
The AI agent decision framework
The post argues that many agent projects should start by asking a simpler question: does this task really need an AI agent at all?
It proposes four tests:
- Variability test — if the task can be mapped as a flowchart, a workflow with LLM steps is usually cheaper, faster, and more reliable than an agent.
- Reversibility test — if the worst action is severe and irreversible, the agent needs a human gate, which can erase the expected ROI.
- Verification test — agents work best when success can be checked automatically, such as code compiling or records reconciling. If a human expert must verify the output, the system is closer to a drafting tool than an autonomous agent.
- Economics test — the real cost includes reliability engineering, evals, tracing, canaries, approval flows, and human ownership time. Token cost is usually the smallest line item.
The core message: many “agent” projects are better built as workflows, and teams should price reliability overhead before committing to autonomy.
More from coding & agent
- DeepMind paper: LLMs can't sample randomly on their own, only convert given seeds — rohanpaul_ai · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- BELL: Open-Source Stack Combines Browserbase, Eve, iMessage and Stripe Link for Personal Agents — jeff_weinstein · 2026-10-27
- Legora runs 2,000 parallel agent evals on Modal Sandboxes for legal AI — blelbach · 2026-10-27
- Two engineers rebuilt a FedEx supplier's billing core in 3.5 months vs an estimated 7-8 — alex_verem · 2026-10-27
- GPT-6.1 Terra scores 48 on AA Coding Agent Index at one-third the cost of GPT-5.6 Terra — ArtificialAnlys · 2026-10-27