The enterprise AI reference, built by people who ship it.
An AI company that turns data into decisions. We publish the architecture, testing and governance material we use on our own systems, and the tools that run on it. Free, no sign-up.
Answered from what we have published, every passage attributed. No account, and a person will follow up only if you ask.
A layer over your stack, not a replacement for it
Keep Microsoft, Oracle, AWS, SAP, ServiceNow. What is badly served is the work between them — deciding what to buy, on what evidence, and finding out afterwards whether it worked. Here is that chain, and which parts of it we actually do today.
Discover · Compare · Decide · Procure. Built and in use. 184 topics mapped, vendor scoring that stress-tests its own weighting, a maturity assessment set by your weakest dimension rather than your average, and an RFP scaffold carrying the cost lines proposals leave out. Every tool runs in your browser and keeps nothing.
Implement · Operate. We do not do these, and saying so is not modesty. Our own governance documents cap every system we touch at observation only, with no clause permitting an automatic action anywhere. Software that could act on your estate while its vendor has never written down what it may not do is the exact thing our specifications exist to describe.
Measure. The schema is published and holds nothing. The Enterprise Outcome Record defines how a result is recorded against a baseline captured before the decision. We hold zero such records. That number sits on our own scoreboard and changes the day it is true, which is a different promise from the one usually made here.
We run the company on an agent workforce
Not a chatbot with a job title. A declared organisation where every seat is an agent, work is routed down a chain of command, and no agent may be given a task before it has been trained. The control room records what was refused as carefully as what was done.
The line that makes it safe is the same one our specifications describe: agents prepare, people commit. An agent researches, drafts, recommends and reports. Approving, committing, spending and deploying are reserved to a person, and every governance document we operate under sits at observation only.
The standards, and the tool to check them
Nine specifications for measuring and governing enterprise AI, under CC BY 4.0 so you can implement them without asking. Each one states what we do not operate.
BEIS
Eleven metrics for an enterprise intelligence system, each naming how it gets gamed. Three are scored not implemented and kept in, because silently dropping a metric you would score zero on is how a vendor specification lies.
The Autonomy Ladder
Six levels, defined by what a customer can audit rather than what a vendor can demonstrate. Seven actions that must never become autonomous.
The Enterprise Constitution
A policy envelope in a form a validator can reject. Run against a governance document written the way most are, it returns thirteen named failures.
Run it against us
The conformance harness is one file, MIT-licensed, and sends us nothing. Bring your own questions. Our own score has been published going down.
We hold no certification, publish no customer case study, and issue no “BvLogic Certified” badge. Everything above is checkable, and each page says where its evidence runs out.
Knowledge centers
Each center is a complete topic cluster: theory, practice, checklists, templates and working examples.
AI Testing Center
How to prove a non-deterministic system behaves. Every topic carries theory, practice, examples, checklists, templates, sample reports and tools.
PMO Knowledge Center
Project and delivery governance: the documents, ceremonies and controls, with templates for each.
AI Company Framework
How an AI-operated organisation is structured, department by department: responsibilities, KPIs, agents, SOPs, templates and workflows.
Get an answer in under five minutes
No account, no upload to a server. These run entirely in your browser.
Systems running in production
We don't have customer case studies to show yet. We do have products you can go and look at, which is the more honest proof.
Enterprise Services
Strategy, implementation, governance and independent assurance for AI systems. Scoped engagements with written deliverables.
See what we doNewsletter
One email. What we published, what we learned, and what broke and why. Sent when there is something worth sending rather than on a schedule. No cadence promises we can't keep.
Subscribe