Customer
I was charged Indonesian tax.
Please refund it.
About Axtor
Built from years of watching enterprise automation evolve.
AI is moving from assisting to acting. The question is whether those actions still reflect the human intent, evidence, policy, and authority that should govern execution.
01Assist
02Reason
03Act
Same potential.
Higher stakes.
The origin story
The systems got smarter. The control problem didn't disappear.
Enterprise systems are more capable than ever, with AI agents taking on real work across the business.
Technical capability does not prove that a proposed action is the right one. As autonomy increases, the harder question is which actions should execute, under what scope, evidence, policy, and authority.
CRM
ERP
Support
Finance
Other systems
AI Agent
Proposed Action
Axtor Control EngineALLOWBLOCKCLARIFYAPPROVAL
A concrete moment
A refund request looks simple. Until it isn't.
A customer asks for a refund. The AI agent can access the account, see the history, and propose an action.
The runtime decision can depend on the intended charge, available evidence, refund policy, authority, external constraints, and risk.
AI agent
I can process a refund for $75.
Shall I proceed?
Runtime control context
Intent✓ Clear
Evidence✓ Invoice available
Policy! Refund path limited
Authority? Not established
External constraint! Cross-border tax
Risk! Tax correction
Why Axtor exists
From permission to runtime control.
Axtor evaluates whether a proposed AI action is sufficiently supported by scope, intent, evidence, policy, authority, and risk before execution.
Scope
Intent
Evidence
Policy
Authority
Risk
What we believe
Intent should be evidenced.
Actions should be grounded in relevant evidence, not assumed.
Uncertainty is information.
When evidence conflicts or intent is unclear, surface uncertainty rather than manufacture confidence.
Probabilistic and deterministic controls have different jobs.
Semantic assessment informs the decision. Existing controls still enforce permissions, limits and approvals.
Consequential actions deserve an explanation.
Enterprises should be able to understand which context and controls shaped a runtime decision.
Built from enterprise reality
Yogi Wiputra
Founder
Nearly two decades in technology, working at the intersection of enterprise customer experience, integration, automation, and AI systems. That experience shaped the view that agentic systems need runtime controls that go beyond technical permission.
Our direction
Building runtime control for the agentic enterprise.
More capable AI agents need clearer control before execution.