AI-assisted simulation. Operator-controlled decisions.
AI helps draft; it does not authorize.
In TaiGong 1.3.4, it means licensed AI assistance inside an authorized simulation workflow—for example, helping generate or refine content through a configured external model. It does not mean an autonomous AI Agent can independently design, approve, schedule, launch, train, and report an entire program.
This distinction matters. A model can accelerate a draft, but it cannot supply written authorization, understand every employment or privacy constraint, verify the intended audience, or accept operational responsibility for a campaign.
The current support boundary is explicit.
| Capability | Status in 1.3.4 | Boundary |
|---|---|---|
| AI-assisted content generation | Licensed | Requires an external model/provider and configured key. |
| AI-assisted content optimization | Licensed | Operator reviews the output before use. |
| Autonomous campaign approval or launch | Unsupported | Authorization and launch remain operator responsibilities. |
| Conversational full-program AI Agent | Roadmap | Future direction, not a delivered 1.3.4 capability. |
Read the complete versioned support matrix →
Keep a human decision gate.
- Define written authorization and the permitted audience.
- Select the scenario and data the model may receive.
- Generate or refine a draft through the configured provider.
- Review accuracy, tone, safety, locale, privacy, and organizational context.
- Configure delivery, controls, schedule, and follow-up in TaiGong.
- Approve launch through the organization’s normal change process.
Model output should be treated as untrusted draft material. Never assume a fluent message is accurate, appropriate, or permitted.
The model is an external dependency.
Availability depends on the selected provider, credentials, network access, quota, model behavior, and the applicable license. Before enabling the feature, decide what content can leave the environment, which provider terms apply, how prompts and outputs are retained, and who can configure the key.
A self-hosted TaiGong instance may still call an external model if that is how the customer configures the licensed feature. Review the complete data path, not only where the application server runs.
Future Agent work stays labeled as future.
Conversational campaign planning and broader Agent-led orchestration are product directions. They are not presented on this site as shipped in version 1.3.4, and v1.3.5 is not treated as the current public release. Evaluation should be based on the published 1.3.4 matrix.