95% of organizations that’ve invested in AI are seeing zero return on that investment. That’s according to a recent MIT study, the State of AI In Business 2025.
The study’s been much discussed since its release. It’s been interpreted by many as evidence that AI might not live up to its potential as a transformative technology inside businesses.
But AI implementations—be they centered around Gen AI chatbots or AI agents—have not been failing because of the technology itself, but rather, as MIT finds, because of misapplication.
The divide between pilots that extract “millions in value” (about 5%) and those that are getting “zero return,” the authors write, “does not seem to be driven by model quality or regulation, but by approach.”
In other words, AI pilots are failing because businesses are using AI tools for the wrong things.
More than half of generative AI budgets are devoted to sales and marketing tools. By far the biggest proven ROI, however, has been found in using AI to facilitate back-office work—the work of internal back-office operations teams, such as procurement, legal, HR and IT.
The study goes on:
“Most [AI implementations] fail due to brittle workflows, lack of contextual learning, and misalignment with day-to-day operations. [But] a small group of vendors and buyers are achieving faster progress by addressing these limitations directly. Buyers who succeed demand process-specific customization and evaluate tools based on business outcomes rather than software benchmarks. They expect systems that integrate with existing processes and improve over time. Vendors meeting these expectations are securing multi-million-dollar deployments within months.
The highest performing organizations report measurable savings from back-office operations. These early results suggest that learning-capable systems, when targeted at specific processes, can deliver real value, even without major organizational restructuring."
Bottom line: the very best and most profitable area of the business to augment with AI is your back-office. That’s where the value is.
So what makes your back-office operations the best place to implement AI? Let’s dive in.
The back office—finance, procurement, legal, HR, IT—manages the invisible machinery that keeps a business running.
The processes that define it are also rule-driven and repetitive, cross-functional and often complex, because the systems these processes need to interact with don’t talk to each other, and the people whose approvals and inputs they need in order to run smoothly work in different departments and application environments that also don’t talk to each other. As a result, to facilitate them end-to-end often requires of back-office teams lots of inefficient manual work.
This makes such processes perfect low hanging fruit for things like Agentic AI and Agentic Orchestration.
Orchestration technology enables back-office teams to coordinate automated business processes across departments, data systems, and application environments.
It provides back-office teams a means of more seamlessly “orchestrating” all their organization’s component parts—that is, their unique mix of people, data, and tools.
Agentic Orchestration can be thought of as orchestration that empowers back-office teams to include in this mix of manipulatable components specialized, autonomous AI agents.
AI agents and Agentic Orchestration represent the kind of AI application MIT found in its study to produce real transformational value, in particular for back-office teams who’re overwhelmed by growing request volumes and complex compliance requirements. The 5% of AI pilots that are delivering ROI are focused here.
Agentic AI is a class of artificial intelligence that moves beyond passive assistance—e.g., summarizing or retrieving information—to take autonomous, goal-oriented actions on behalf of humans and to autonomously execute complex work over time.
The very best AI agents and Agentic Orchestration platforms operate as a layer that connects every moving part of your organization, and makes it possible to facilitate collaboration at the process level between humans and agents across the tech stack.
For your back-office team, AI agents powered by orchestration can operate in effect like a well-networked and highly intelligent colleague, only, back-office teams remain in control at all times over what exactly they can do, what systems and data they can access, and how they collaborate.
For back-office teams, this sort of AI implementation can unlock a step-change in capability:
It’s important to note, though, that not all AI agents are created equal. To use agents to conduct work such as that described above, you’ll need an agentic platform capable of orchestrating safely across your entire organization—not just in one UI or vertical.
Back-office teams work across ERPs, contract management systems, siloed ticketing tools, etc. Employees whose approvals and inputs back-office processes rely on work in entirely separate constellations of environments and systems. Back-office teams must be able to orchestrate agents across all of them. Agents must be capable of meeting employees where they are, for example, in order to truly empower them.
Tonkean’s AI agents, for example, help back-office teams autonomously orchestrate back-office processes. They can be used to empower the employees that back-office teams serve to do and focus on their best work—without compromising on compliance, control, and visibility.
The hype around AI and in particular AI agents remains justified—the potential of this technology is transformational.
But the transformation isn’t occurring where many initially expected it might.
It’s happening—right now—in the back-office.
Want to learn more about how Tonkean helps enterprise back-office teams like Workday get started with Agentic AI? Let’s talk.
