Telstra
Dual genAI agent assist: one tool summarizes the customer history in a single sentence, the other answers the advisor's questions from the internal knowledge base
Telstra rolled out to its contact centers two genAI tools built on Azure OpenAI (One Sentence Summary, Ask Telstra): 90% of advisors report saving time, with 20% fewer follow-up contacts and 80% reporting a positive impact on the customer interaction.
Key points
- Two genAI agent-assist tools: a customer account summary and a business answer in RAG.
- Built on Azure OpenAI and Azure AI Search, with Microsoft and Accenture.
- 90% of advisors save time, 20% fewer follow-up contacts.
- Evidence B, confirmed status, extended to contact centers and stores in 2024.
Objective
Save advisor time on each contact, cut callbacks, and make the interaction smoother by giving the advisor the customer history and the business answer in real time.
The deployment
Telstra built two genAI tools on Azure OpenAI and extended them to its contact centers and store teams. One Sentence Summary condenses a customer's history, notes, and transactions into a short summary, so the advisor grasps the situation at a glance without rereading the whole file. Ask Telstra lets the advisor query the internal knowledge base in natural language and get a generated answer. On the trials run in late 2023 with about 200 advisors, 90 percent of One Sentence Summary users reported saving time and being more effective, with 20 percent fewer follow-up contacts, and more than 80 percent of Ask Telstra testers said the tool improves the customer interaction. Telstra extended both tools to all its contact center and store teams over 2024, with Microsoft and Accenture.
Results Proof B
Figures published by Telstra in its official release and echoed in a Microsoft customer story with the same values (90 percent time saved, 20 percent fewer callbacks, 80 percent positive impact). Brand release plus vendor are consistent; the measures are self-reported (advisor self-assessment on the trials), not an efficiency gain measured on incrementality.
How it works
Documented architectureThe stack in detail
- llm Microsoft Azure OpenAI Service OpenAI models hosted on Azure, used for account summarization (One Sentence Summary) and answer generation (Ask Telstra). Model version not disclosed.
- outil Azure AI Search Indexing and search of the internal knowledge base for Ask Telstra's RAG.
- infra Azure Machine Learning ML platform used to operate the summarization and generation components.
- integrateur Accenture Scaling and data modernization partner for the extension to all teams.
How it runs, concretely
For ops teams-
1Summarize the customer account AI / Azure OpenAI
One Sentence Summary compiles history, notes, and transactions into one sentence to save the advisor from rereading the whole file.
-
2Answer business questions AI / RAG
Ask Telstra queries the internal knowledge base in natural language and surfaces a generated answer to the advisor.
-
3Keep the advisor as decision-maker Human advisor
The advisor reads, checks, and decides the answer to the customer. The AI prepares and suggests; it does not speak to the customer in the advisor's place.
-
4Measure time saved and callbacks Data team + management
Tracking of time saved, follow-up contacts avoided, and perceived quality to decide the extension to all teams.
The freshness and accuracy of the knowledge base behind Ask Telstra. If it is stale, the tool returns a wrong answer that the advisor passes on to the customer.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- Consolidated customer history (notes, transactions) accessible at the advisor's desk
- A clean, up-to-date internal knowledge base
- GDPR governance over the data exposed to the LLM
Org prerequisites
- Advisors trained to verify the AI's outputs
- A process to maintain the knowledge base
Possible stack
- Azure OpenAI + Azure AI Search, or Vertex AI, or a market LLM in RAG
- A component to summarize the customer account automatically
- Integration with the advisor's desktop / CRM
The plan, step by step
- Step 1Clean up the internal knowledge base and consolidate the customer history (notes, transactions) accessible at the advisor's desk, with GDPR framing.Deliverable: An indexable knowledge base and validated data access.
- Step 2Build the automatic account summary and integrate it into the advisor's desktop.Deliverable: A summary tool piloted on a real flow.
- Step 3Pilot on 100-200 advisors with a control group; measure time saved, callbacks avoided, and perceived quality.Deliverable: Pilot results against control.
- Step 4Add the business answer in RAG on the knowledge base, with systematic human verification.Deliverable: A business-answer assistant in pilot.
- Step 5Extend to the contact center and store teams, train on verifying outputs, set up the knowledge-base maintenance process.Deliverable: Full deployment and a routine to keep the base updated.
First step: Pilot the account summary on a group of advisors, measure time saved and callbacks avoided against a control group, then add the business answer in RAG.
The same pattern, proven in other industries
ADT
genAI customer agent
Air India
genAI customer agent (automated customer service)
Bank of America
conversational financial assistant at scale
Sources
- S1 Telstra scales up AI adoption following promising pilots of generative AI solutions Primary archive pending
- S2 Telstra dials in elevated customer service with Azure OpenAI Service Interested party archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.