Artificial intelligence has already arrived in supply chain management. Forecasting tools use machine learning. Planning systems detect anomalies. Productivity assistants summarise reports and draft analyses. If the surrounding commentary is to be believed, supply chain management is somewhere between a technological renaissance and the imminent disappearance of spreadsheets.

The reality is both more modest and more interesting. The deeper shift underway is organisational rather than technological. AI is gradually connecting signals across companies and functions, changing when decisions are made and who sees the consequences. When that capability matures, the supply chains of today will look as if they belong in the operational history books.

Phase 1: AI as an Efficiency Layer

Most organisations are currently in the first stage of AI adoption.

Artificial intelligence is increasingly embedded inside enterprise software. Forecasting engines incorporate machine learning. Document processing tools extract operational data automatically. Productivity assistants help analysts interrogate large datasets without writing complex queries.

The rapid progress of generative AI accelerated this phase dramatically. Frontier models developed by organisations such as OpenAI and others are now embedded across enterprise productivity platforms, allowing planners and analysts to interact with operational data conversationally and surface insights that previously required hours of manual work.

Large organisations are already deploying these capabilities at scale. Leading retailers have expanded machine-learning forecasting models across thousands of products and stores, while global manufacturers are integrating AI-driven demand planning directly into operational workflows.

These systems can produce clear returns. Forecast accuracy improves. Analysts spend less time preparing data. Reports are produced faster.

Judging by the marketing language surrounding them, one could be forgiven for thinking that civilisation itself has been upgraded along with the spreadsheet.

The tools are smarter, but structurally the supply chain remains largely unchanged.

That is why many AI initiatives feel incremental despite the impressive technology behind them. They can produce clear value, but they remain mostly local optimisations, improving parts of the system without fundamentally changing how the organisation operates.

For many SMEs, this first wave of AI will arrive mostly through the software they already use. Planning tools, ERP platforms, and analytics systems are quietly embedding these capabilities into everyday workflows. The opportunity is real, but the strategic question remains the same as it has always been: whether the organisation itself is structured to use better information effectively.

Phase 2: Operational Agents

The next stage moves beyond analysis toward execution.

AI systems will increasingly act as bounded operational agents, performing specific tasks within defined guardrails and human supervision.

Instead of merely highlighting insights, these agents will begin taking operational action.

An agent may adjust safety stock parameters automatically when demand volatility increases. Another may monitor supplier signals and detect emerging disruptions earlier. Logistics agents may continuously refine routing plans as conditions change.

Early versions of this model are already appearing. Enterprise software platforms are experimenting with AI agents capable of executing workflow actions and assisting with operational tasks. Technology companies such as Shopify and ServiceNow have begun deploying AI assistants able to perform structured operational activities rather than simply generating analysis.

In supply chains, the first wave will likely target routine operational maintenance – and there is plenty of it.

Many planning teams spend large portions of their time correcting spreadsheets, adjusting parameters, reconciling forecasts, and chasing exceptions across systems.

Operational agents are very good at this kind of work.

As they mature, they will remove a large amount of the friction that currently dominates operational planning.

The consequence will be gradual but meaningful. Roles centred on operational maintenance will shrink, while roles focused on coordination, trade-offs, and decision quality will become more valuable.

Some of these agents may even figure in organisational charts, receive an employee ID and an email address, but probably fall short of a gym membership.

Yet even this stage is not the real breakthrough.

It still operates largely within the existing organisational structure.

The Real Breakthrough: Enterprise Intelligence

The real transformation begins when AI stops analysing isolated datasets and starts connecting signals across the enterprise and beyond it.

Most supply chain problems are not technical forecasting problems. They are coordination problems.

Information exists across the organisation. It is spread across obvious and unlikely places alike, but it rarely meets in time to influence decisions.

Marketing launches a promotion without visibility of emerging supply risks. Procurement discovers supplier constraints weeks after commercial commitments have already been made. Logistics identifies capacity issues after customer orders are confirmed.

Each function has competent people and good information, but the signals remain fragmented.

The breakthrough comes when AI systems connect operational data, communication streams, planning assumptions, supplier signals, and commercial activity into a shared operational context.

This is where the next real phase of AI begins. It is best described as Enterprise Intelligence.

Instead of analysing decisions after they are made, the system begins detecting patterns while decisions are still forming.

Paradoxically, this shift may favour smaller organisations: fewer layers, fewer systems, and shorter decision chains often make SMEs better positioned to act on connected intelligence once it becomes available.

Enterprise Intelligence in Practice – A Glimpse

Imagine a marketing meeting in Lisbon in 2030, late on a sunny Thursday afternoon.

The local team is discussing a regional promotion designed to revive a product that has been quietly underperforming across Southern Europe. It is a routine discussion. The sort that happens in hundreds of offices every week.

The conversation takes place in a meeting room and in a shared digital workspace. Notes are captured automatically by the organisation’s productivity platform.

But the system notices something no one in the room can yet see.

It reads the meeting notes in real time. It recognises the product being discussed and the proposed promotion timing. It connects this signal with other operational information across the company.

Elsewhere in the organisation, procurement teams have exchanged messages with a supplier about an upcoming shortage of a key raw material used in that product. The issue is still being assessed.

No transaction exists in the ERP system.

No sales forecast or procurement plan has been updated.

Yet the system connects the signals.

Within minutes, a notification reaches the regional supply planning lead and the procurement manager responsible for the supplier relationship.

The system highlights a potential conflict between the intended promotion and an emerging supply constraint.

The message is not an alarm. It is decision context, arriving early enough to influence the decision itself.

What could be a familiar cycle of several weeks of fragmented signals, planning, frustration, escalation and re-planning was abbreviated at the speed of light.

That is Enterprise Intelligence.

If this sounds like science fiction, consider a small historical thought experiment. Imagine telling a production manager in 1985 that one day he would open a screen and instantly see every unit of inventory across warehouses, every customer order in the system, and every shipment currently moving through the network.

Beyond the Enterprise: Orchestration and Network Intelligence

Once organisations begin operating with widespread agents and Enterprise Intelligence, another shift becomes possible.

AI systems can start orchestrating responses across functions.

A disruption in supplier capacity no longer requires a chain of meetings across procurement, planning, logistics, and sales. The system can trigger coordinated adjustments across planning parameters, inventory allocation, production schedules, and logistics flows within defined governance rules.

The organisation begins reacting as a coordinated operating system.

Further ahead lies an even more ambitious stage.

Supply chains are not individual companies. They are networks of independent organisations.

The next frontier will be Network Intelligence: systems exchanging AI-to-AI signals across companies and institutions to anticipate external factors, disruptions and coordinate responses across the wider supply network.

Retailers sharing early demand signals with suppliers. Manufacturers detecting capacity constraints across the network before they escalate. Logistics providers anticipating congestion across regions.

The early infrastructure for this already exists in supply chain visibility platforms and collaboration networks.

But its full potential has barely begun to emerge.

Preparing for What Comes Next

This is why the conversation about AI in supply chains should not revolve primarily around tools.

Forecasting improvements and automation projects are valuable. Many can produce genuine ROI. They are simply not the structural transformation that conference stages sometimes imply.

Preparing for Enterprise Intelligence starts elsewhere.

Organisations need clear operational processes, reliable data structures, and decision governance that genuinely connects planning, procurement, production, inventory and fulfilment. Without those foundations, connecting signals across the organisation becomes extremely difficult.

The technology is advancing rapidly. Most operating models are not.

For SMEs, the situation is slightly different from large global corporations. Big organisations can afford to run dozens of AI pilots across forecasting, automation and analytics. Smaller companies rarely have that luxury. Their advantage lies in simpler structures, shorter decision chains and the ability to move quickly once the operating model is clear.

Preparing for the next phase of AI therefore means clarifying how the supply chain actually works: how decisions are made, how information moves between departments, and where operational signals appear first.

NEXU works with organisations at this intersection between operational foundations and emerging technology. The aim is straightforward: make the supply chain coherent enough that modern analytics, planning platforms and AI can genuinely improve decisions rather than simply adding another layer of software.

Because the reality is simple.

You cannot build the next generation of operational intelligence on operational chaos.

When it arrives, the real AI transformation will not look like a tool at all.