Zishan Ali Khan
I build agentic AI systems for institutions where autonomous failure is a legal event, banks, insurers, retailers, and then publish the review: what held, what broke, and what it cost per case.
That discipline should scale down, too. Away from the day job, my passion is democratizing it: small businesses and non-profits, especially Canadian ones 🍁, deserve the same honest AI conversation the big institutions get. If that is you, reach out →
Enterprise AI GTM & Solutions · Camunda · Toronto / New York
Field guides
All guides →Reference-grade documents on the protocols and practices of production agentic AI. The kind you bookmark, written after building the thing.
AG-UI: Streaming the Interface, Not Just the Answer
A field guide to the agent-user interaction protocol, and what happens when a process engine drives the stream
AP2: The Agent Payments Protocol
What I learned making an agent that buys FIFA tickets while I sleep
Where Agents Actually Work: Field Notes from 50+ Enterprise Builds
Two years, seven industries, four workflow shapes. A map of where agentic AI earns its keep, and where plain code still wins.
The Control Plane: Governing Agentic AI at Scale
Agent frameworks own the inner loop. Someone has to own the outer one.
Evaluating Agentic Systems: The Four-Axis Method
Cost per case, per-field quality, tool-skip rates, and temperature: how to pick a model for an agent that lives inside a governed process
Tool Contracts: Why Agents Skip Your Tools
A tool spec is a contract between a deterministic system and a probabilistic caller. Most teams write it for the wrong reader.
Selected work
All work →AURUM · AG-UI wealth onboarding
A private-bank onboarding agent that streams the whole interface, not just text. Generative UI cards, risk gauges, and human approval desks over one SSE stream, orchestrated by a process engine. Runs without API keys in shadow mode.
AURUM Pays · AP2 agent payments
in buildOne of the first working implementations of Google's AP2 v0.2 outside their own samples, wired to an external governance layer. A shopping agent buys FIFA tickets overnight under a cryptographically signed mandate it cannot override. Repo goes public when EP 03 ships.
Mortgage concierge · multi-agent A2A
Multi-agent mortgage pre-qualification with a judge agent gating decisions at a confidence threshold, and a locally-hosted research agent behind a data-privacy boundary. Became the worked example in Camunda's official A2A blog series and a Gemini 3 hackathon listing.
Four-axis agent eval harness
Built when nobody was asking for agent evals, now it is the first question in every room. Measures cost per case, accuracy against a gold set, tool-skip rates, and temperature sensitivity for process-embedded agents. Field-validated on live enterprise workloads across 20+ models. The method is public; the code is not.
Writing
All writing →Recent talks
All talks →A newsletter is coming. Until then, new guides and episodes land on LinkedIn first and in the RSS feed.