PwC teams up with OpenAI: the customer-care hotline becomes an AI agent
When one of the world's Big 4 audit-and-consulting firms sets up a joint center with OpenAI just to rebuild customer care, it is no longer an experiment. The old hotline - press 1, wait through the menu - is being replaced by agents that listen, understand and act.
On July 15, 2026, PwC US announced an agentic customer care solution suite built with OpenAI: voice agents and digital agents on OpenAI's multimodal APIs that listen, understand intent, act on their own and self-improve. The two firms set up a joint Center of Excellence pooling AI, engineering and customer-service experts for fast enterprise rollouts. Context: Gartner forecasts that by 2029 agentic AI will autonomously resolve 80% of common customer-service requests, cutting operating costs by 30%. For Vietnamese SMEs, the "agent answers first, humans take the hard cases" model works today on Zalo OA and Messenger.
What did PwC and OpenAI announce?
The heart of the deal is a voice agent + digital agent capability PwC built with OpenAI: customers call or message, the agent grasps real intent instead of keyword-matching, performs the work itself (order lookups, rescheduling, policy-based refunds) and learns from every conversation. PwC wraps this into what it calls the "agentic front office": marketing, sales, commerce and service running on one AI operating model instead of 4 separate systems - the same direction Salesforce announced with Agentforce Commerce a week earlier. To speed rollouts, the two firms created a joint Center of Excellence - a signal this is a long-term business line, not a demo project.
Why are voice agents the turning point?
Text chatbots are everywhere, but the most demanding customers still pick up the phone - and that is a business's most expensive channel. The new generation of voice agents is nothing like the "press 2 for accounting" IVR: it converses naturally, can be interrupted, understands regional accents, and most importantly does the work instead of transferring the call. Gartner estimates that once such agents spread, 80% of common requests will need no human - customer-care teams shift from repetitive answering to supervising agents and handling hard cases. This is exactly job 1 and job 5 in the 9 jobs an AI agent can do that we recently analyzed.
An agent only holds the desk well once FAQs, policies and order data are clean - PwC starts every project there too. Stock photo (CC0).
Chạm AI's take: Vietnamese SMEs do not need to wait for PwC
The PwC-OpenAI deal targets corporations, but its logic scales down to Vietnamese SMEs. One, chat channels first, voice later: Vietnamese customers already message on Zalo and Messenger - a Vietnamese-language agent on those 2 channels covers most of Gartner's "80%" at far lower cost than voice. Two, data before the agent: FAQs, return policies and order status must be clearly structured before an agent dares to act on its own - that is 70% of the effort in every project, as we describe in the AI agent implementation process. Three, measure by after-hours cases: start with one countable job (does the customer messaging at 10 pm get an answer and a closed order?), get the numbers, then expand - you can estimate the ROI quickly with our free tool before deciding.
Sources: PwC US - press release "PwC to Help Organizations Transform Agentic Customer Engagement and Service with OpenAI", July 15, 2026 · Gartner, forecast published 3/2025. Global forecast figures, offered as reference for the Vietnamese market.
Frequently asked questions
What did PwC and OpenAI just announce?
On July 15, 2026, PwC US announced an agentic customer-care solution suite built with OpenAI: voice agents and digital agents using OpenAI's multimodal APIs to listen, understand intent, act autonomously and improve over time. The two firms set up a joint Center of Excellence to bring the solution to enterprise customers faster.
What does Gartner forecast for AI in customer service?
Gartner (3/2025) forecasts that by 2029 agentic AI will autonomously resolve 80% of common customer-service requests without a human, cutting operating costs by 30%. Unlike chatbots that only answer, agents will act: cancel service plans, reschedule deliveries, negotiate shipping fees.
What can a Vietnamese SME without a big call center learn?
3 things: voice and chat channels will be staffed by agents first with humans on hard cases - a model that works today on Zalo OA and Messenger; FAQ, policy and order data must be clean and structured before an agent can hold the desk; and start with one measurable job such as after-hours replies or order lookups.