Why We Don’t Simply Have One Office

The idea of separating our offices into a ‘front’ and a ‘back’ dates back to 18th-century finance in London and later on Wall Street, where revenue-generating staff were deliberately separated from those handling processing and settlement.

As each group could focus on its own responsibilities while the wider organization coordinated between them to deliver value, this structure still makes sense for a lot of organisations. The front office is there to drive revenue and customer engagement, while the back covers operations, finance, HR, and IT. But in the light of the AI revolution, is this still a useful framework?

The AI convergence narrative

At least a few industry watchers argue that an important new player in the business landscape—artificial intelligence, and particularly the new wave of autonomous ‘agents’—may be the factor that will render the front/back office distinction meaningless. The Economist suggests SAP and Salesforce—historically representing back- and front-office worlds—are now converging through AI agents that replicate each other’s strengths; PwC highlights similar potential in areas such as banking.

I can see the attraction: what business wouldn’t want to consolidate everything onto a single platform, eliminating the front-office/back-office divide and achieving seamless workflows and fully dissolved silos?

The challenge is that, no matter how many AI agents enter the picture, the underlying division in how businesses operate is unlikely to disappear. Agents have an important role to play, but the fundamental structure of business remains unchanged. And as they famously say, if it ain’t broke, don’t fix it.

Why ‘one system’ or ‘one office’ is unrealistic

To me, the big vendors’ vision of a ‘one-size-fits-all’ IT environment looks like a chimera: appealing, but difficult to justify in practice. Even if right as a long-term route of travel, such a change would not dismantle core enterprise structures, or the vendors that work so well at the different ends of the office, any time soon.

Yes, software boundaries are blurring, and AI is delivering real gains. Opening a bank account, for instance, can now be done almost instantly through digital identity verification rather than manual paperwork. But it is unrealistic to assume AI agents will collapse complex enterprise environments into a single unified system. The reality is we rely on multiple providers for sound commercial reasons as much as technical ones.

This is why regulators emphasise supplier diversification, and many experts advocate for multi-cloud strategies. Vendor lock-in is why CIOs long ago moved away from single-supplier models (essentially a ‘monopoly’ risk). Today, value comes from combining best-in-class solutions across computing, applications, data, and infrastructure.

Even vendors themselves depend on third-party integrations. That alone makes the notion that AI will transform expanded platforms such as SAP, Salesforce, or others into a single source of truth for the entire enterprise feel, shall we say, overambitious.

The missing document intelligence layer

Customers we speak to highlight that their CRM front office and ERP back office systems are essential, but neither handles the continuous flow of documents and information that underpins everyday operations.

Take SEW-EURODRIVE, a global leader in industrial automation (all baggage handling systems at Frankfurt Airport are powered by its drive technology, for example). The company receives around 5,000 custom orders daily, embedded in emails, which are processed through document intelligence. Like many large enterprises, this takes place within a highly complex IT landscape, and to deliver for its customers it needs a scaled application environment that spans multiple functions across the front and back office.

The system recognises an order embedded in an email, registers the order in SAP, spins up the CAD design system and orchestrates the order from design through to manufacturing and delivery. It can even cross-reference data in a CRM system, like Salesforce, to identify further upsell opportunities. This kind of intelligence is an ‘agentic overlay’ to the existing enterprise landscape, rather than a single ecosystem vendor trying to assert ownership.

Front, back, and maybe middle?

We should be cautious about the extent to which big vendors want to use AI agents to recreate more monolithic systems. This approach is not reflective of how modern enterprises operate.

At the same time, businesses do need a unifying agentic layer that connects disparate platforms, adding intelligence and automation at a document and process level. In my view, this is best understood as document intelligence: a common language across systems that enables organisations to extract maximum value from their data, no matter where it is.

Crucially, this layer operates independently of office geography. Long after today’s excitement around AI agents fades, the challenge of making front and back office systems work effectively together will remain central, and perhaps even for the next century or two.

At least, until robots manage to do it better.

Dr John Bates is the CEO of Doxis, a global leader in ECM and document intelligence

‍