RSM's $1 Billion Bet Signals the Rise of AI Agents

RSM, the sixth largest accounting firm in the world, just announced a plan to invest $1 billion in artificial intelligence over the next three years—a massive step up from its previous spending. The firm isn't just buying more chatbot licenses for its employees; it's betting on "AI agents" to automate complex tax and accounting workflows from start to finish. This move by a mainstream professional services giant signals a pivotal shift in the enterprise AI landscape: the focus is moving from AI as a helpful "copilot" to AI as an autonomous "agent" that can execute tasks on its own.

What are AI agents, and how are they different from chatbots?

For the past couple of years, the most common form of enterprise AI has been the "copilot"—a tool that assists a human user. It can summarize a document, draft an email, or find information, but it requires a person to prompt it, review its output, and take the final action.

AI agents represent the next evolutionary step. They are AI systems designed to take actions and make decisions on behalf of a user to achieve a goal. Instead of simply responding to a query, an agent can perform a sequence of tasks across multiple applications. For example, an agent could be instructed to "plan a business trip to Taipei for Computex," and it would then proceed to check calendars for availability, browse flights, book a hotel that fits a budget, and add the itinerary to the user's calendar—all without step-by-step human intervention. It’s the difference between asking for a recipe and having a robot cook the meal for you.

Why are accounting firms betting big on AI agents now?

The primary driver is a massive leap in potential productivity. RSM's experience is telling: the firm found that providing individual employees with generative AI tools resulted in a 5% to 25% productivity boost. However, by deploying AI agents to fully automate multi-step processes like compliance reviews, that boost jumped to as high as 80% in some cases.

This level of automation promises a dramatic return on investment. As AI models become more capable of reasoning and using software tools, companies see an opportunity to move beyond simple augmentation and begin fully automating entire workflows, particularly those that are repetitive, data-intensive, and rule-based. This is why firms across the professional services sector, from RSM and BDO to the Big Four (PwC, Deloitte, EY, KPMG), are now collectively plowing billions into developing and deploying agentic AI.

Who is leading the agent technology race?

The push for agents is a key priority for the world's leading AI labs. Google has been experimenting with "Project Mariner," an agent designed to navigate the web to complete tasks, and is promoting an "Agent-to-Agent" protocol to allow different bots to communicate. OpenAI has been vocal about its plans for powerful agents, and its latest models are being built with the ability to use software tools and make multiple "reasoning" steps to solve complex problems. This agent-centric vision is also fueling hardware ambitions, with OpenAI's Sam Altman and designer Jony Ive reportedly developing a new family of AI-powered "companion" devices.

This represents the new frontier of competition. The value is no longer just in the raw intelligence of the model, but in its ability to reliably and securely interact with the digital world to get things done.

What are the challenges holding agents back?

Despite the hype, the march of the AI agents faces significant hurdles. A key concern is trust and reliability. A chatbot "hallucinating" an incorrect fact is one thing; an AI agent booking the wrong flight or misinterpreting a compliance rule is a far more serious problem. A KPMG report showed that public trust in AI systems is already falling, and high-profile failures by autonomous agents could erode it further.

There are also major technical and security challenges. Granting an AI agent the autonomy to access multiple systems, use login credentials, and handle sensitive data (like financial or personal information) opens up a host of new security vulnerabilities. The industry is still working on establishing the standards and guardrails necessary to allow agents to operate safely at scale. As the research firm IDC noted, the evolution from simple chatbots to reasoning models to agentic AI will require "several orders of magnitude more processing capacity."

Reference Shelf:

RSM Plans $1 Billion Investment in AI Agents, Other Services (WSJ)

We Can’t Afford to Rush the March of AI Agents (Bloomberg)

AI Agents Are Taking Autonomous Action (Google)