Commonwealth Bank is rolling out ChatGPT Enterprise to nearly 50,000 employees, pairing broad access with hands-on training and guardrails. The goal: faster customer service, smarter case summaries, and stronger fraud response—embedding AI fluency as a core capability across frontline teams and operations.

What’s New & How It Works

From pilots to scale. CBA treats AI fluency as a core skill, not a niche project—giving staff secure access, daily prompt practice, and leadership forums to normalise usage.

Customer service uplift. Agents use AI to summarise conversations, draft compliant replies, translate on the fly, and propose next best actions—cutting average handle time and improving first-contact resolution.

Fraud response assist. Analysts use AI to condense transaction histories, surface patterns, and draft case notes—accelerating triage while human experts retain final decisions.

Enterprise guardrails. Data controls, role-based access, audit logs, model/version pinning, and red-team reviews keep usage safe and compliant.

Practical Rollout (Step-by-Step)

1. Access & enablement
   Provision enterprise accounts; configure policies (no PII in prompts beyond approved scopes), set retention controls, and integrate SSO.

2. Training at scale
   Deliver role-based pathways: frontline service, fraud ops, lending, marketing. Use daily prompt challenges, lunch-and-learns, and “office hours” for managers.

3. Workflow patterns

   • Service: conversation summaries, compliant reply drafts, translation, tone adjustment.

   • Fraud: timeline compression, anomaly callouts, evidence pack drafting.

   • Ops: meeting notes, action extraction, process documentation.

4. Human-in-the-loop
   Require approvals for high-risk actions; keep a visible audit trail and escalation routes.

5. Measurement & iteration
   Track AHT, FCR, CSAT, queue abandonment, case closure time, and false-positive/negative rates in fraud. Run controlled pilots → phased scale-up.

Example Use Cases

• Contact centre: AI drafts a compliant response and next steps; agent approves and sends in seconds.

• Fraud triage: AI summarises 90 days of activity, highlights risky sequences, and creates a structured case note.

• Branch support: Staff generate policy-aware answers for customers and translate complex terms into plain English.

Risks & Governance

• Accuracy & hallucination risk: ground answers in approved KBs; require human review for sensitive cases.

• Privacy & security: minimise personal data; apply DLP; disable training on customer data by default.

• Regulatory alignment: document controls, testing, and auditability; maintain model/reporting change logs.

FAQs

How many employees will be trained?
Nearly 50,000 employees are being enabled with ChatGPT Enterprise and role-based training.

What outcomes are expected?
Faster, more consistent customer service; better case documentation; accelerated fraud triage with humans making final decisions.

Why choose OpenAI/ChatGPT Enterprise?
Enterprise-grade controls, speed, and tooling for secure, audited usage at scale—plus strong support for prompt workflows and retrieval.

Is this safe for a bank?
Yes—when deployed with strict data controls, role-based access, human oversight, and continuous testing.