Doxis’ Dr. John Bates agrees AI agents are fueling dreams of a unified enterprise stack, but the front/back office divide isn’t going anywhere soon. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.
In our world of IT, every few years a new technology arrives promising to simplify office lives once and for all (hands up anyone who remembers The Paperless Office or Collaboration Platforms). This time, it’s Agentic AI: software that’s been specifically architected to act autonomously across systems and processes.
It’s an undeniably seductive idea. What busy business wouldn’t jump at the chance to resolve everything onto one platform—ending the front-office/back-office divide and achieving the nirvana of seamless workflows and an end to silos?
The problem is that a three-century-old division in how businesses operate is unlikely to disappear, no matter how many AI agents enter the picture. The fundamental structure of business remains unchanged.
That doesn’t mean agents aren’t worth investigating—they are. But the vision of “one size fits all” IT environment feels like a chimera, and not one that justifies huge investment yet.
Why do we have two sorts of offices at all?
The idea of separating the “front office” from the “back office” dates back to the early days of modern finance in the 18th century, particularly in the City of London and later on Wall Street. In those environments, employees responsible for generating revenue and interacting with clients were deliberately separated both physically and functionally from those handling administrative processing and transaction settlement.
That made a lot of sense, as it allowed each group to focus solely on its own responsibilities, while the wider business acted as the catalyst bringing the two sides together to deliver value to customers.
The front office is typically associated with winning customers and driving revenue, while the back office encompasses functions such as operations, finance, human resources, and IT. One side is seen as the stage where outcomes are created, while the other operates behind the scenes, largely out of view.
Because customer-facing roles tend to receive greater visibility and recognition, while operational work remains less visible, this division continues to shape corporate priorities, funding decisions, and workplace culture.
Although it once made practical sense when teams were physically separated across offices or floors, the question is whether this model remains relevant in a highly digital and automated environment. What began as a practical necessity may have evolved into a broader way of thinking that is now so embedded in business language it goes largely unquestioned.
As organizations continue to evolve through modern digital technologies, is this now perhaps an overly simple framework for distinguishing between customer-facing roles and the operational activities that support them? If so, greater results might be achieved by ending the distinction altogether.
Why not remove the separation and instead operate as a single, unified organization—where customer-facing and operational functions are no longer treated as fundamentally different domains, but as parts of the same continuous system?
Some vendors argue that, while previous attempts fell short, this convergence can finally be achieved through AI. In particular, the emergence of autonomous business software. The Economist, for instance, claimed long-established back office provider SAP and front office specialist Salesforce, both investing heavily to broaden their capabilities, are on a collision course. Both aim to deploy AI agents that allow each platform to replicate the strengths of the other, effectively creating a unified technology stack capable of supporting all core business functions and erasing the historical divide. PwC, sees similar benefits for a key industry vertical such as banking where AI could be applied in the same way.
As a keen believer in the transformative power of AI in the enterprise, I remain unconvinced. This is the same species of thinking that led to the recent SaaSpocalypse, where some traders believed business software as we know it will soon be replaced entirely by AI, with traditional categories fading away.
Even if true, the demise of the front- and back-office divide, and of the large vendors that currently own those respective domains, is unlikely to materialize at anything like this pace. Let’s examine why.
A Single System to Rule Everything?
Yes, the boundaries between software categories are becoming less distinct, and increased convergence is delivering benefits for both organizations and their customers. In an AI-enabled world, for example, opening a new account can now be completed almost instantly, replacing outdated processes that once relied on manual data entry from paper forms or faxed documents.
Even so, it is important to approach claims about AI with caution. While agents are powerful, it is unrealistic to assume they will seamlessly transform complex, multi-layered technology environments into a single, unified system capable of handling every aspect of enterprise data processing.
This is because a single system to rule everything is a vision, or a vendor sales message, that does not align with how businesses actually operate. In practice, organizations rarely rely on a single provider for all their technology needs. There are strong commercial and regulatory reasons for this: even in cloud environments, regulators discourage overconcentration, and many experts advocate for multi-cloud strategies.
Anyone who has experienced vendor lock-in will appreciate why. Functional considerations also matter: the era when companies expected a single supplier for all their IT needs, be that IBM, Digital Equipment Corporation, Groupe Bull or ICL, to meet every requirement, has long since passed. Today, CIOs know they can unlock greater value for their CEOs by combining best-in-class solutions from multiple providers across computing, apps, data, networking and storage.
In fact, even the manufacturers acknowledge this. All enterprise platforms depend on third-party integrations to deliver their full capabilities. Given external components are already necessary to make these platforms effective, the idea that AI will transform expanded ecosystems like SAP or Salesforce into a single, definitive source of truth for the entire enterprise seems overly ambitious.
My perspective is informed not only by hard-won experience in this area, but also by conversations with customers, who consistently highlight the same challenge. They recognize the importance of both back-office ERP systems and front-office CRM tools, yet also understand that neither is particularly effective at managing the continuous flow of documents and information that underpin everyday business operations.
Consider HR. We are frequently brought into organizations that already use comprehensive, market-leading HR platforms to help these companies handle employee records, streamline HR workflows, enable internal queries, and archive critical data. In addition, we enable capabilities such as information retrieval, automated document creation, communication workflows, and interactive employee engagement.
Even applications designed to lead their categories often require these additional layers—interfaces, support mechanisms, and connective technologies—to function effectively in real-world environments. Document intelligence, the next generation of enterprise content management, is a powerful approach that provides that connective “glue,” binding disparate systems and addressing complex processes such as invoice management. It enables organizations to understand and classify documents, automate document processes, store, search and manage massive document repositories, and generate and communicate documents across multiple enterprise ecosystems (e.g. bridging SAP and Salesforce).
We Always Need Front, Back & Middle Office
I believe the big ERP vendors are attempting to use AI agents to recreate the large, monolithic systems of the past.
While it is understandable that vendors are pursuing this direction, it does not reflect how modern organizations actually operate. For business technology to deliver meaningful value, it requires a unifying layer that connects disparate platforms—and, in my view, this is best met through document intelligence. Such a layer would function as a common language across systems, enabling monitoring, organization, and the extraction of maximum value from organizational data and applications.
Crucially, document intelligence operates independently of traditional organizational boundaries, supporting processes wherever they occur. Long after the current excitement around AI agents subsides, the importance of ensuring back- and front-office systems perform as effectively as possible while also finding ways to help both sides complement each other will remain a key CIO priority for a long time to come—perhaps even for the next century or more.
