PwC’s AI agents are now your consultants — whether you’re ready or not
PwC is putting its AI agents directly in front of clients and cutting out the traditional back-and-forth with consultants as intermediaries. This is part of the firm’s push to reshape the consulting model around AI, not just productivity gains.
PwC’s newly launched PwC One platform lets clients log in, describe a problem, and have autonomous agents perform the work, with PwC professionals reviewing outputs in the background rather than driving them.
This is a significant shift from how PwC has deployed AI until now, and how enterprise consulting has historically worked.
Previous tools were internal, helping staff work faster while clients experienced the same engagement model they always had. PwC One changes that and embeds AI into how projects are delivered to clients as a part of the core workflow.
From back office to front door
The client initiates. The machine executes. The consultant reviews. That inversion of the traditional engagement model is what PwC is betting on, and what makes PwC One something more than another enterprise AI productivity tool.
It also combines the firm’s compliance frameworks and proprietary methodologies with autonomous capabilities, Joe Voyles, Principal, AI & Emerging Technology at PwC, tells The New Stack. Voyles is a PwC partner working in their CTO’s office leading AI engineering.
“PwC One is kind of an entryway for our clients to start to consume service offerings that are meant to be more kind of autonomous and agent oriented, so more of like a self-service type of thing — like a front door into the firm,” he says.
The machine executes, the consultant reviews
Voyles says PwC’s ultimate goal is to get to a point where the machine can interact with the firm’s clients.
“We can take in the data we need to perform a task, and the agents are able to perform that task at the level of quality that we would expect from any of our professionals perform the work that’s the objective,” he says.
The idea is to minimize as many human touches as possible. However, currently humans are deliberately kept in the loop wherever system quality isn’t yet at professional-grade or where regulatory standards require it.
“So, we are deliberately designing human touches in the process, currently in specific areas where we know either, one, systems are not performing to the level we would expect, or, two, where we have a professional expectation that humans are reviewing a guiding machine,” Voyles says.
Built for autonomy, designed with guardrails
He says the goal of the client-facing portal is an always-on relationship rather than episodic email exchanges.
The platform has a multi-LLM architecture, where it uses different models depending on the use case. It also standardizes how agents connect, so a team with an existing agent can plug into PwC One within a few days, Voyles says.
“PwC One is one step in a longer journey, but an important one,” says Matt Wood, PwC US and Global Commercial Technology and Innovation Officer, in a statement. “With PwC One, AI is at the center of an engagement, working alongside our clients and our people in ways that let us surface insight, test assumptions and navigate complexity that neither would reach alone.”
Always on, always watching
Although PwC One is still in beta, the company has several ongoing use cases with clients, including financial statement analysis, tax incentives exploration, transfer pricing compliance, Greenhouse Gas emissions anomaly detection, and early-stage financial due diligence.
The consulting model question
When asked how the platform affects PwC’s consulting business model, the firm said it is rethinking pricing — not to reduce cost, but to better reflect speed to insight and quality of outcomes. On liability, PwC was direct: accountability sits with the firm and its professionals, with a human reviewer in the loop on every PwC One engagement.
The post PwC’s AI agents are now your consultants — whether you’re ready or not appeared first on The New Stack.
