Predictika
A deterministic engine behind a probabilistic model
A language model is fluent, and probabilistic. It will give you a plausible answer, and sometimes a plausible answer that is wrong. Predictika puts a deterministic engine behind it. The model proposes; the engine checks the proposal against your rules; the verdict is the same every time.
That is what makes an AI system safe to put in front of money, code, tickets and records – the part that has to be right is not left to a guess.
The model talks. The engine decides.
Two halves: the agent proposes, the engine decides
Logic Validation Agent (LVA)
Sits between the model and the action. It reads what the agent produced – structured output or free text – pulls out the choices that matter, hands them to the engine, and, when a choice is rejected, turns the rejection into a corrected re-prompt the model can act on. It is the translator and the loop.
Logic Validation Engine (LVE)
The deterministic core. Your policies, catalogs, price lists and limits are loaded as constraints, and the engine reasons over all of them together, in any order the choices arrive. It returns allow, needs approval or block, with the rule that decided and the reason. One engine runs product and pricing logic, finance and audit rules and agent-action policies; the validation modules differ, the engine does not.
Reasons, not guesses
Reasoned together
Every rule is evaluated jointly, so a combination that breaks a floor is caught even when each rule passes on its own.
Order-independent
A buyer, or an agent, can state the facts in any sequence and get the same result.
Explained
Every verdict carries the rule that produced it, written for three readers at once – the agent, the audit log and the person. A soft finding holds the action for approval; a hard rule failure rejects it.
Built to change
Rules are modeled declaratively in LVML, where sub-entities inherit both parameters and constraints from the structures that contain them, so a rule is edited in one place. One audit model took six rounds of change in under twelve hours.
Patented, and built on the same idea for thirty years
- US Patent 11,544,475 B2
The deterministic logic engine integrated with an AI system to validate business logic.
View Patent → - US Patent 11,914,970 B2
The adaptive layer that turns a rejection into a corrective prompt and refines the model's output.
View Patent → - Patent pending
Conversational autonomous AI agents guided by a logic engine for accurate and reliable reasoning.
The lineage runs from Xerox PARC through Selectica's enterprise CPQ engine to Predictika – the same constraint-based reasoning, now behind LLM-driven agents.
See it on your own rules
Tell us the agent action or quote you worry about most. We build a first validator from your public catalog, policies or price list and send a private link in under two days.