CTO of DutyGraph. We map how work actually happens before it's handed to AI, with a named person accountable for every agent. Looking for 5-10 pilot teams.
@brocmayo agreed, a person on every fix turns into rubber stamping fast. low risk stuff can auto apply, anything touching money or customer data gets a person
Say your sales team builds an app in a chatbot. They hit Publish and "Anyone with the link" is already ticked.
Researchers just found 35,578 open Grok-built apps. 24 made by X and SpaceXSI staff.
Say your AI agent gets into a government health portal in June. You catch it in August. You tell them in September, through a public email inbox.
That was OpenAI's agent and Australia's Medicare stats portal.
All of these share one gap: nobody wrote down whose agent it was.
DutyGraph maps people, duties, tasks and software. Every agent gets a named owner, written scope and human checkpoints.
Cohort #1 has only a few spots left before year end. Apply here: https://t.co/aPWOFgtoOf
Ops tools are catching up. Mirantis Agent Platform went GA with spend caps down to each agent, a one-click network cutoff, and a person approving fixes before agents apply them.
Controls still need an owner. DutyGraph puts a name on each one.
https://t.co/Ey7CiwODM6
Say your coding agent is halfway through a job at 3pm and hits the daily AI cap. Who decides if it keeps going?
Apple engineers just had that cap cut from $300 to $200 a day.