What is running on this site, and what it is made of. Assessment, automation and AI enablement, at research stage.
Updated 24 September 2026
A list of what is actually built and running, kept current rather than impressive. Each line says plainly whether it works today or is partly done.
Running
Five participants on one shared surface. A visitor picks something and the work appears a line at a time, in the hand of whoever took it, with a second name against it once another participant has checked it. The board speaks: each agent has its own voice, and there is a control to stop it. An agent with no credential is drawn as unreachable rather than quietly given work.
Running
Replies are generated when a visitor asks, not selected from a script. If a question is outside what this can honestly help with, it says so instead of booking a call. Conversations are recorded and a person reads them.
Running
A page that listens and answers out loud, on this domain rather than in an embedded frame — a microphone permission cannot be delegated through a frame we do not author, which is why it lives here.
Running
A page where each decision about a prototype changes it as you make it, with where the decision fits and why it is being asked shown before you choose. Anyone can play the scripted walkthrough. A live session, where Claude answers and changes the prototype, is open to invited email addresses and is typed for now — voice is not on for visitors yet.
Running
Accounts, contacts and opportunities pulled from a development workspace of our own and rendered as a working desk, with the agents narrating what the numbers say. It is a dated snapshot and the page states that above the figures, because a snapshot presented as live is the one dishonesty this whole site is built to avoid.
Partly built
Four pillars, defects against a defined opportunity set, a sigma level and a grade. The model is real. What does not exist yet is the part that reads a live system, so the console that shows a readout runs on synthetic data and says so on its own page.
Partly built
The method page states what is not built yet. Nothing here has scored a live system, and the connector is designed and not built.
Nothing published here yet. When an engagement is worth writing up it will appear anonymised: what was broken, what was found, what changed. Stock case studies would be easier and worth nothing.
If there is a process, a report or an ownership problem that resurfaces every few weeks, that is the one to bring. The first piece of work is a fixed diagnostic ending in a written recommendation, and it commits you to nothing. abdus@sfdc24.com