What James D. Thompson’s 'Organisations in Action' can teach us about AI agents in document workflows

Many things about AI feel new. But the organisational problems AI is trying to solve are often much older.

In Organizations in Action, first published in 1967, James D. Thompson argued that organizations are not simply designed to produce outputs. They are designed to keep producing outputs under conditions of uncertainty. Markets shift, information is incomplete, inputs vary and exceptions constantly appear. The challenge is not just getting work done but keeping core operations effective despite disruption.

That is exactly why AIDA’s AI agents matter.

They are not just another layer of automation. They are a practical response to a fundamental organisational challenge: how to protect core work, coordinate complex processes and respond intelligently when conditions change.

From automation to adaptive support

Traditional document automation is often built around fixed rules. When documents follow expected patterns, workflows run smoothly. But when formats vary, information is missing or exceptions arise, the process quickly depends on human intervention.

That is the reality of modern business operations. Finance teams handle invoices in multiple layouts. Legal teams manage incomplete files. Operations teams deal with delays, anomalies, missing approvals and changing priorities. Even highly digitised organisations still face uncertainty every day.

This is where Thompson’s ideas are highly relevant. He argued that organisations need structures that protect their technical core - the activity that creates real value - from unnecessary disruption.

In document-driven businesses, that core may be accounts payable, customer onboarding, claims handling, reconciliation, legal document review or any other process that must work reliably at scale.

AIDA’s AI agents help support that core. They can analyse context, make decisions within defined boundaries, trigger actions, request missing information and escalate where needed. Instead of expecting people to absorb every exception manually, the organisation gains a more intelligent operational layer.

Protecting the core from uncertainty

One of Thompson’s most useful concepts is buffering. Organisations create buffers around critical activities so that volatility does not constantly disrupt essential work.

In document processes, volatility comes in many forms: inconsistent formats, mismatched data, missing fields, reconciliation issues, urgent follow-ups or unstructured communication across teams and partners.

Without a way to absorb that variability, skilled teams spend too much time firefighting rather than focusing on high-value work.

AIDA’s AI agents can act as intelligent buffers. They can review incoming documents, detect anomalies, route items appropriately, support validation and initiate the next step in the workflow. Instead of letting every irregularity hit the operational core directly, agents help contain variation and keep work moving.

This is not only about efficiency. It is about resilience and scalability.

Managing interdependence across workflows

Another major contribution of Organizations in Action is Thompson’s view of interdependence. Some work can be standardised. Some needs sequencing. Some requires continuous mutual adjustment between teams and functions.

That matters enormously in document-driven operations.

A process may look linear on paper: receive, extract, validate, approve, archive. In practice, many workflows are far more interconnected. A delivery note affects an invoice. A contract affects compliance. A payment status affects customer communication. A discrepancy may require coordination between finance, operations, procurement and account management.

This is where AI agents become especially valuable.

AIDA’s AI agents can operate across steps, systems and decision points. Rather than treating each document as an isolated file, they can support a broader flow of coordinated action: linking information, supporting decisions, triggering follow-ups and helping different parts of the organisation stay aligned.

In Thompson’s terms, they help where standardisation alone is no longer enough.

From passive processing to action-taking

Many automation tools are good at extracting data but not at acting on it. They process information, yet still rely on a person to interpret what matters and decide what should happen next.

AIDA’s AI agents move beyond that model.

They can help understand the meaning of a document in context, determine the next step, support communications and initiate downstream actions. That shift matters because organisations do not only need faster processing. They need systems that can contribute to operational decision-making while remaining aligned with business rules, governance and human oversight.

This is one of the clearest ways agentic AI strengthens modern operations: it helps organisations become more adaptive without losing control.

Boundary-spanning in a document-driven world

Thompson also emphasized the importance of functions that connect the organisation to the outside world. These “boundary-spanning” activities deal with customers, suppliers, partners and other external pressures so that internal operations are not constantly destabilised.

Today, many of those interactions are document-led. Orders, invoices, claims, contracts, delivery notes, approvals and related communications all move across organisational boundaries.

AIDA’s AI agents can help manage this layer too. They can reduce friction, improve responsiveness, support follow-up actions and maintain flow between internal teams and external stakeholders.

That is not just automation. It is organisational coordination.

Why this matters now

What makes AIDA’s AI agents powerful is not simply that they are intelligent. It is that they fit the real needs of modern organisations.

Businesses need more than better document management and faster document processing. They need systems that help them operate with more consistency, flexibility and control in environments that are increasingly dynamic. They need technology that supports both structure and adaptation.

That is exactly where AIDA’s AI agents fit, building upon our solid intelligent document processing (IDP) and optimal character recognition (OCR) foundations.

Seen through the lens of Organizations in Action, agentic AI is not a break from sound organisational thinking. It is a natural extension of it.

The tools have changed. The challenge has not.

Organisations still need to protect their core, coordinate interdependent work and adapt to uncertainty without losing control. AIDA’s AI agents are designed to help make that possible.