agent workflows
architecture judged 89% likely a named technique or architecture
See how it has spread across media tiers
Mentions over time
Table view
| Week | Articles | Sources |
|---|---|---|
| 2024-W04 | 1 | 1 |
| 2024-W20 | 1 | 1 |
| 2024-W25 | 1 | 1 |
| 2024-W53 | 1 | 1 |
| 2025-W31 | 1 | 1 |
| 2025-W36 | 1 | 1 |
| 2025-W47 | 1 | 1 |
| 2026-W02 | 1 | 1 |
| 2026-W03 | 1 | 1 |
| 2026-W05 | 2 | 1 |
| 2026-W09 | 1 | 1 |
| 2026-W10 | 1 | 1 |
| 2026-W11 | 6 | 1 |
| 2026-W14 | 1 | 1 |
| 2026-W15 | 1 | 1 |
| 2026-W16 | 1 | 1 |
| 2026-W17 | 1 | 1 |
| 2026-W18 | 2 | 2 |
| 2026-W24 | 3 | 2 |
| 2026-W25 | 1 | 1 |
| 2026-W29 | 1 | 1 |
| 2026-W31 | 1 | 1 |
| 2026-W32 | 1 | 1 |
| 2026-W34 | 1 | 1 |
| 2026-W36 | 1 | 1 |
| 2026-W38 | 1 | 1 |
| 2026-W39 | 3 | 3 |
| 2026-W40 | 14 | 5 |
| 2026-W41 | 2 | 2 |
In their words
The Orchestration Problem: Why Cron Breaks for Agents Cron is fine for “run this script at 02:00.” Agent workflows are not that shape.
Model usage is only one cost component of an agent workflow.
Benchmark: https://www.kaggle.com/benchmarks/karmendra8386/cost-per-correct Tooling: tokenecon — my open-source cost accounting for agent workflows · Preprint: https://doi.org/10.5281/zenodo.23195511 A single model call has a predictable price.
In agent workflows, input tokens can add up quickly as more context is carried across multiple model calls.
While developing Susan AI, I get to explore: AI API integration Prompt engineering AI UX Web application architecture Authentication concepts State management API design Error handling AI response streaming File processing Agent workflows Deployment Performance Product design The project is therefore both a product experiment and a learning laboratory .
The goal is to reduce redundant API traffic and lower the risk of hitting provider rate limits as agentic workflows scale.
The danger begins when you apply the same logic to dynamic applications: agent workflows, user‑driven tools, real‑time document processing, and multi‑turn conversations with shifting context.
But when you start building agent workflows across several applications, you end up dealing with many of the same problems repeatedly: permissions, application state, references, dependencies, artifact transfers, and checking what actually happened.