AT0M by Pienomial

// use cases · working samples

See AT0M decide.

Every example here is a working sample. Pick one or paste your own text, and AT0M answers all of its questions in a single pass, with a probability for every possible answer. Turn on the comparison to put a chatbot beside it.

Yes/no a probability Choice pick one option Score rate 1–5

// browse by pattern, industry or function

Five things AT0M does, across every industry.

Almost every use case is one of five patterns. Pick a pattern to see it across industries, or narrow by industry and function.

// AT0M + LLMs

AT0M alongside your LLM and your systems.

AT0M doesn't write, and it doesn't replace your language model or your business systems. It makes the fast, typed decisions between them — whether a request should reach the LLM, whether a draft can go out, and where every record should go next.

Guardrails on both sides of your language model

Two typed checks: one before a request reaches the LLM, one before its draft reaches a person.

  1. Request arrivesFrom a customer, patient, resident or employee.
  2. AT0M checks the requestPersonal data? Injection attempt? Does it need the LLM at all?
  3. Your LLM draftsOnly requests that pass, with sensitive data removed.
  4. AT0M checks the draftUnsupported claims, advice, off-label or policy statements.
  5. Send, revise or blockYour code acts on the verdict; a person takes over when needed.

1Before: check the request

Runs before anything is sent to the language model, so personal data and injection attempts never leave your network, and requests with a standard answer skip the LLM entirely.

Between the two checks, your LLM writes the draft. It can be any model, hosted or local. The samples in step 2 are drafts an LLM might produce.

2After: check the draft

Runs on every draft before a person sees it. A verdict of pass, revise or block tells your code what to do next.

One step in your workflow, not the whole workflow

With or without an LLM, AT0M sits in the middle of a larger system. Every use case on this page shows what comes in, what AT0M decides, what your systems do and who stays responsible.

  1. A record arrivesA message, form, report or sensor summary comes in from a system you already run.
  2. AT0M decidesIt answers your typed questions in one pass, with a calibrated probability on every answer.
  3. Your systems actYour code applies thresholds: act automatically, queue for review or escalate, and log every probability.
  4. People decide what mattersDecisions about a person's rights, money, health or job stay with accountable people.

Ready to build your own?

Write and run your own questions in the builder, then follow the steps from first question to deployment inside your own perimeter.