Relational infrastructure for AI-shaped decisions
AI is now inside the documents and decisions you have to defend.
Decisions that Converse, not Coerce.
I read them the way a hostile reviewer will and hand you a memo on exactly what to harden, before a lender or a court finds the gap for you.
I · The practice
An adversarial read of what your team has to stand behind.
For risk, ESG, compliance and safeguards teams. Begin with one fixed-scope read. Five working days to a memo, no retainer.
Always/Never Worksheet1 of 3 survive a hostile read
- We always keep a human in the loop.Cut or rewrite
- We never use customer data to train our models.Hedge it
- Every automated decision can be appealed to a person.Defensible
II · The publication
Field notes on how accountability fails, traced in arcs.
283 episodes across 40 arcs, with the library, the book, and editions in other languages.
“She works the social and technical sides of a system with equal fluency, and reads governance for where accountability actually fails rather than the obvious places.”
- Field-tested in
- ESG, safeguards, grievance, audit, social performance
- Proof, labelled
- Case examples marked verified, anonymised, composite or illustrative.
Questions teams ask
- What does Sociable Systems do?
- Sociable Systems gives risk, compliance, ESG, and safeguards teams an independent, adversarial read of the AI-shaped documents and decisions they have to defend, and hands back a written memo on exactly what to harden before a lender, regulator, or court finds the gap. The method borrows from industrial safety: named hazards, stop-work authority, evidence trails, and refusal architecture.
- Who is it for?
- Risk, ESG, compliance, audit, procurement, and safeguards teams who have to stand behind AI-assisted decisions in high-stakes public, industrial, and social-risk settings. It is not for general AI strategy, model building, or low-stakes experimentation.
- How much does an AI governance document review cost?
- Pricing is fixed and published rather than quoted case by case. Every fixed-scope read, artifact, and training track is listed with its current rate in the catalogue, and on its own page. GrieVoice runs as a scoped pilot. There is no retainer; you can begin with a single fixed-scope read.
- What do I get, and how fast?
- A written hardening memo, delivered as PDF and markdown within five working days of receiving the document and confirming scope. It tells you which claims hold, which are soft, who can enforce them, and a prioritised list of what to fix first.
- Is this AI compliance or AI safety consulting?
- It is adjacent but more specific. The work is human-in-the-loop accountability for high-stakes decisions: reviewing AI-assisted evidence, challenging confident automated summaries before they harden into the record, and interrogating AI vendors before procurement. It does not build models or write general AI policy from scratch.
- Do you offer AI vendor due diligence?
- Yes. The Vendor Interrogation Pack ($99) gives procurement, legal, ESG, and operations teams six pressure-test questions, answer patterns, and one forensic read of an AI scoring, screening, or triage vendor's reply before signing.