st
Customer Interaction Decision WorkbenchAI customer interaction decision-making platform · V22
Connecting to the live product service
AI customer interaction decision-making platform

Decide first whether you should say it or not, then decide when to say it and which path to take.

Organize user behavior into a timeline, generate current status, behavior change probability and auditable reach/no-reach decisions, and output recommended interaction paths and planned tasks.

  1. 01
    Data readyUpload and verify user behavior data, and generate a behavior timeline sorted by time.
  2. 02
    Prediction completedGenerate current status and behavior change probabilities for target users.
  3. 03
    Decision pendingThe system outputs recommended paths or remains silent, and writes audit records.
  4. 04
    Export confirmedManual confirmation of site reports and planned tasks can be used for subsequent operational execution.
Number of timeline users generatedNo timeline data yetCount the number of deduplicated user_ids that exist in behavior_timeline_record.
Forecast coverageNo prediction result yetThe number of users whose behavior_prediction_record exists / the number of users whose timeline exists.
remain silent ratioNo decision record yetNumber of interaction_decision_records with decision='hold' / Number of total interaction_decision_records.
Task awaiting manual confirmationNo tasks to be confirmed yetThe number of interaction_decision_records with decision='contact' that have not yet been confirmed.
BUSINESS INPUT

New project

Each project represents a customer interaction decision-making scenario, such as new user conversion, dormant user recall, repurchase reminder, or membership relationship maintenance.

Loading product-specific business fields.

Not saved
TRACEABLE OUTPUT

Interactive decision-making

Results, sources, failures, approvals, and rollbacks are saved by project.

0 live runs
∅

No business results yet

Please upload the behavioral data file containing event_id, user_id, event_type, and event_time first.

PRIVACY

Data use boundary

User behavior data and user IDs are sensitive operational data. Before uploading, you should confirm that you have obtained legal authorization, and avoid importing highly sensitive data that is irrelevant to interactive decision-making, such as ID number, bank card number, precise address, and medical information. The system processes this data only to generate timelines, states, probabilities, and in-site decision records. Current 10 role review supplements: Default field whitelist: event_id, user_id, event_type, event_time and explicitly configured low-sensitivity event attributes; other fields are not stored in the database, sent to the model, or exported by default. ; Before uploading and next to the upload control, it must be noted that user_id and event_file are sensitive operational data, and it is prohibited to upload irrelevant highly sensitive data such as ID number, bank card number, precise address, medical information, government document number, unauthorized contact information, etc. ; The model call only sends a minimal summary and necessary configuration, and does not send the complete original behavior file or irrelevant high-sensitivity fields. ;The user_id is hidden or hashed by default when exported, and the original event details and complete message content are not exported; the export of personal-level details requires secondary confirmation and auditing. ;Project deletion must cover original files, derived records, input snapshots, personally identifiable audit summaries, and exported files; only irreversible aggregate statistics and operational audits without personal identifiers will be retained.

RETENTION

Data retention

Project data is retained for 90 days by default; project administrators can delete behavior files, timelines, forecast results, and audit records in advance. After deletion, individual-level details in historical reports cannot be recovered, and only summary statistics without user IDs are retained.

SAFETY

Human responsibility and rollback

This product generates customer interaction decision-making suggestions, site planning tasks and audit reports, and does not guarantee transaction, retention or relationship improvement results. AI judgment may fail due to missing behavioral data, timeline errors, sample bias, or improper threshold configuration; when it fails, the reason for the failure must be displayed, and false conclusions must not be generated.

Rolling back only means revoking the report version, scheduled task status or manual confirmation status in this site, and restoring to the previous version of the decision record; it does not mean withdrawing emails, text messages, advertisements, CRM records or any unconnected external production system changes.