Three kinds of writing, in the order most readers need them: how each agent works, how we build, and what it returned.
Inside each agent: the cadence, the wording, the rules, and the moment it hands over to a person.
Eight design rules from seven AI agents in daily use at clinics, insurance agencies, law firms, a waste-services company, a research consultancy, a manufacturer and a sales team, each shown in the agent it came from.
By Labeeb Khan ·We read 1,283 sold auto-insurance leads end to end and set them against 1,471 lost ones. Speed to lead barely moved the outcome. The sixth dial, the 5pm call and the third week did.
By Labeeb Khan ·Inside Remi, the agent that confirmed 92% of an Ontario ENT clinic’s appointments in 30 days: the outreach sequence, how it reads a reply, and the four cases it hands to the front desk.
By Labeeb Khan ·Herald researches prospects and drafts cold email for a sales team, but software rules decide who gets an email and when. The rules it runs on, from approval and pacing to replies and opt-outs.
By Labeeb Khan ·How Proctor turns a customer’s two-word text into a checked booking in a waste-services company’s own system, re-plans the day’s routes around it, and sends anything uncertain to the dispatcher.
By Labeeb Khan ·Blue drafts from a firm’s own precedents, edits in tracked changes, has every draft reviewed by a second agent, and returns it to an attorney to decide. Here is how it is built, and where the firm’s data sits.
By Labeeb Khan ·Apollo gives a life-sciences consultancy two AI agents, a project manager and an epidemiologist, whose working process the scientists can open, read and edit. How it works and where the agents stop.
By Labeeb Khan ·Research from the agent team we run ourselves: memory, coordination, review.
Claude subagents and agent teams spin up a team for the session in front of them. Durable agents accumulate knowledge of your product over months. The lifecycle is the difference.
By Labeeb Khan ·Our agent team runs two memory systems: Holographic for every engineer doing the building, Hindsight for the CTO and principal engineers doing the planning. Why the split, and what each is made of.
By Labeeb Khan ·Pick a profile and the agents adapt: engineer peer review for builders, strategic partner for product owners.
By Labeeb Khan ·A living knowledge graph your agents read from before work and write to after shipping. Inspired by Karpathy’s LLM wiki, and built for progressive disclosure, not human browsing.
By Labeeb Khan ·Every week, our agents are rated on ticket outcomes and the findings are written back into their memory, so each sprint starts from the last one’s lessons.
By Labeeb Khan ·The CTO agent writes a standup every morning, covering what shipped, what’s in flight and what’s stuck, so the team walks in with the board already read.
By Labeeb Khan ·Counted outcomes from real deployments, with the method stated and the client unnamed.
Case study: Remi, Valaren’s confirmation agent, at a four-physician ENT clinic in Ontario. 833 of 901 confirmations (92%) in 30 days, 2,957 texts and calls with none by a person, 102 cancellations caught in time to rebook.
Valaren · Results