Enterprise questions
Questions enterprises ask before deploying Tajalli.
Direct answers for technical, security, risk and procurement teams. Where a claim has limits, the limits are stated.
01Enterprise questions
Architecture & positioning
Does Tajalli replace our existing optimization, AI or operational systems?
No. Tajalli is designed to operate between decision systems and consequential execution.
QISTAS can evaluate whether an operational commitment proposed by an existing optimizer or planning system is currently admissible. RADM can evaluate whether an action proposed by an AI agent is currently authorized.
The objective is generally to add a deterministic control layer without requiring the organization to replace its existing intelligence or optimization stack.
Is Tajalli an AI model?
No. Tajalli may operate alongside AI systems, but its core role is deterministic control.
RADM is specifically designed to mediate between probabilistic AI agents and consequential enterprise tools. QISTAS performs deterministic admissibility and certification functions for operational commitments.
What does “deterministic” mean in Tajalli?
It means that the operational decision is generated from explicit state, evidence and policy rather than from model confidence or free-form probabilistic interpretation.
Under the same canonical input state and the same governing policy version, Tajalli is designed to reproduce the same decision.
The claim concerns the Tajalli decision layer. It does not imply that the external world is deterministic.
Does deterministic control mean Tajalli predicts the future perfectly?
No. Tajalli is not based on claiming perfect future prediction.
Its architecture instead limits a decision to the evidence and conditions under which that decision remains valid. When a governing state crosses a relevant boundary, the existing certification or authorization can be withdrawn and evaluated again.
Is Tajalli already proven across every industry shown on this website?
No. This website distinguishes between validated use cases, controlled research or benchmark environments, and architectural applicability to other domains.
Potential applicability is not a deployment claim. The Assurance page states which capabilities are validated, designed or under evaluation.
02Enterprise questions
Tajalli QISTAS
Does QISTAS replace forecasting or optimization?
No. Forecasting estimates what may occur. Optimization determines a preferred decision. QISTAS determines what can currently be certified as admissible under the available evidence and governing constraints.
These functions are complementary.
What happens when the real-world state changes after QISTAS issues a certification?
The certification is not treated as permanently valid. QISTAS associates decisions with validity conditions and operational boundaries.
When a relevant boundary is crossed, the existing certification can be withdrawn and the new state evaluated again. This is central to the architecture.
What invalidates a QISTAS certification?
Each certification carries explicit validity conditions: the evidence it was derived from and the operational boundaries within which it holds.
A certification is invalidated when monitored evidence crosses one of those boundaries, when required evidence stops arriving, or when the governing policy version is replaced. It is then withdrawn and the current state is evaluated again.
What happens when evidence becomes stale?
Staleness is treated as a boundary in its own right. Evidence older than its permitted age no longer supports the certification built on it.
Depending on configuration, QISTAS withdraws the affected certification or reduces it to what remaining evidence still supports. It does not continue publishing a certification on expired evidence.
Can QISTAS operate in Shadow Mode?
Yes. In Shadow Mode, QISTAS consumes the same proposals and evidence as the production workflow and computes certifications in parallel, without gating any commitment.
Its decisions are then compared with what was actually committed. Shadow studies can begin from historical or file-based exports.
How is a QISTAS decision replayed?
Each certification is recorded with its input evidence, canonical state, policy version, engine version and validity boundaries.
Replay re-executes the certification from those recorded inputs. Because the decision layer is deterministic, the replayed certification must match the original; any difference is treated as a defect.
03Enterprise questions
Tajalli RADM
Is RADM another LLM guardrail?
Not in the conventional sense. Most model-level guardrails focus on generated content, prompts or responses.
RADM is designed around runtime authority and effect control. The important question is not only “What did the agent say?” but “What is the agent actually permitted to cause?”
Can RADM stop every possible AI attack or unsafe action?
No responsible system should make that claim.
RADM is designed to reduce specific classes of unauthorized or invalid execution by imposing an explicit deterministic authority boundary before consequential tools. Its protection depends on the integration boundary, the evidence available, policy coverage and the set of actions actually mediated by the system.
Controlled benchmark results must not be represented as universal security proof.
Can an agent bypass RADM and call the tool directly?
The strongest deployment architecture places consequential execution behind an exclusive mediated boundary.
If an agent retains an independent path directly to the underlying tool, no middleware layer can truthfully claim complete execution control over that bypass path.
The deployment architecture therefore matters as much as the runtime decision engine.
What happens when authority is ambiguous?
RADM does not resolve ambiguity in favor of execution.
If identity, authority, provenance or the consequence-relevant state cannot be established for an action, the action is blocked or routed to an explicit approval path, according to policy. Ambiguity is a reason to stop, not a reason to guess.
Does RADM need to inspect prompts?
Not as its primary mechanism. RADM decides on the proposed action: which tool, which operation, which arguments, from which sources, with which effect.
Prompt or conversation context may be recorded as provenance where useful, but authorization does not depend on interpreting the model's language.
Can humans remain in the approval loop?
Yes. Policy can require explicit human approval for specific effects, thresholds or tools.
Such actions are staged rather than executed, held for an attributable human decision, and then committed or aborted. The approval itself becomes part of the decision record.
04Enterprise questions
Deployment & governance
What happens when required evidence is missing or stale?
Tajalli does not convert missing evidence into assumed permission.
Depending on product configuration and operating mode, the system can refuse certification, withdraw an existing decision or require another decision path.
The governing principle: missing evidence does not become invented authority.
Can Tajalli be introduced without controlling production systems?
Yes. Enterprise adoption is designed to be staged:
- Shadow
- Advisory
- Gated
In Shadow Mode, Tajalli evaluates real operating data without controlling production execution. This allows the organization to compare outcomes before introducing enforcement.
Does Tajalli need access to all of our data?
No. The architecture follows data minimization.
Tajalli requires the evidence necessary to construct the governing state for the particular decision being mediated. It should not collect unrelated business or personal information merely because that information exists.
Can Tajalli run inside our own infrastructure?
Tajalli is being architected for enterprise deployment models including private environments and controlled cloud infrastructure.
Available deployment options for a given product are confirmed during evaluation scoping.
How is a Tajalli decision audited?
A production-grade Tajalli decision is associated with:
- input evidence
- normalized state
- policy version
- engine version
- reason code
- decision output
- relevant validity boundaries
- cryptographic decision identity, where configured
This allows a decision to be reconstructed and compared against a later deterministic replay.
Can policies be changed after a decision?
Policies can evolve through explicit versioning.
Historical decisions, however, remain tied to the policy version under which they were issued. A later policy version should not silently rewrite the history of a previous decision.
How would an enterprise begin evaluating Tajalli?
The recommended first step is normally a bounded technical evaluation.
For QISTAS, this may involve shadow certification against an existing operational workflow. For RADM, it may involve placing a selected set of consequential agent actions behind a controlled mediation boundary.
Before results are evaluated, the evaluation should define:
- scope
- evidence
- policy
- baseline
- success criteria
- claim boundaries
Enterprise enquiries
A question we have not answered?
Bring it to a technical conversation. We would rather state a limit than overstate a capability.