He is the orchestration layer.

And I say this with deep respect for him. He is brilliant at what he does. But a company should never depend on one person to hold everything together. That is a fragile setup. Everybody in that room knew it. Nobody said it out loud until I did.

So let me say it here: what this manufacturer needs, what most manufacturers need, is a technical orchestration layer. A system that does what this one person does. Every day. Every shift. Without getting sick, without going on holiday, without retiring.

The real number behind the chaos

How many applications does the average enterprise actually run? Depends who you ask. Okta’s 2025 data, based on real login activity, says 101. Zylo, tracking SaaS subscriptions, says 275. MuleSoft lands at 897, though they sell integration software, so take that number with some healthy skepticism.

For a manufacturing plant, the number that matters is usually somewhere between twelve and thirty. ERP, MES, SCADA, PLM, QMS, CMMS, WMS, CRM, ITSM, BI tools. Plus whatever the last digital transformation project left behind that nobody dares to turn off.

Most of these systems work perfectly fine on their own.

But they were never built to work with each other. MuleSoft found that only 29% of enterprise applications are actually integrated. The rest? Manual handoffs. People copying numbers from one screen into another. Decisions based on whoever last saved something to a shared drive.

A Harvard Business Review study tracked workers across three Fortune 500 companies. They toggle between applications 1,200 times per day. That burns almost four hours a week just reorienting after each switch. Roughly 9% of their working time, gone. And that is office workers. On a factory floor, where a wrong number in the wrong system means a missed shipment or a failed audit, the cost goes far beyond time. It hits trust. It hits customers. It hits the bottom line.

Where the hours actually disappear

I keep asking the same question when I talk to operations people: what happens when your ERP says one thing and your MES says another?

The answer is always the same. Someone walks over to someone else. They open two systems side by side. They compare. They make a judgment call. They move on. Nobody logs this. Nobody tracks the twenty minutes it took. It is just how things work.

Multiply that by every shift. Every handoff. Every order confirmation that requires cross-referencing two sources of truth that were never designed to agree with each other.

That is the orchestration gap. Your systems work. They just work alone.

The AI that industry actually needs (and the one it keeps buying instead)

Here is where I have to be honest about something that most people in this industry will find uncomfortable.

The instinct right now is to throw “AI” at everything. And I understand why. The marketing is loud. The demos are impressive. The pressure from the board is real.

But let me share what the data actually says.

MIT’s NANDA initiative studied hundreds of enterprise AI deployments. Only about 5% of integrated GenAI pilots achieved measurable impact on the P&L. Five percent. RAND’s research on AI project failure puts broader rates above 80%. The reasons come up again and again: misaligned expectations, wrong tool for the job and infrastructure gaps.

In manufacturing, I believe the numbers are worse. Because manufacturing does not need surprise. Manufacturing needs repeatability.

Think about it this way. Would you put a system that gives a different answer to the same question every time you ask it in charge of your quality process? Your supply chain reconciliation? Your audit documentation?

I would hope the answer is a clear no.

The distinction that matters here is simple. Probabilistic AI generates plausible outputs that vary each time. It is built to be creative. Deterministic execution gives you the same output for the same input. Always. Full audit trail. Zero hallucination. Zero creative reinterpretation of your production schedule.

For a factory, that second one is everything.

Connect what you have. Stop replacing what already works.

The pattern I see at almost every manufacturer we work with goes like this: someone proposes a platform migration. Rip out the old system. Drop in a new one. Twelve months minimum. Eight figures. The promise is consolidation.

The reality? You end up with a shiny new system that still fails to talk to the three old ones you could never switch off because somebody built a critical process on top of them years ago.

I have seen this story play out a dozen times. It always ends the same way.

The better path is orchestration. Instead of replacing systems, you connect them. An orchestration layer sits above your existing platforms and coordinates what they already do well. Your ERP stays. Your MES stays. Your SCADA stays. The orchestration layer makes them work together. Deterministically. Auditably. Automatically.

This idea is simple in principle. What changed is that it is now feasible at the speed and reliability that manufacturing actually requires. But only if you build it on deterministic AI. The kind that does the same thing right, every single time.

Why this conversation belongs in Europe

I am writing this ahead of Hannover Messe 2026 and there is a reason this topic hits harder here than in San Francisco.

Europe’s regulatory environment with GDPR, the AI Act and data sovereignty requirements is often framed as a burden. And sometimes, honestly, it is.

But for industrial AI, it is the competitive advantage.

Data sovereignty means your orchestration layer will never silently route production data through a US hyperscaler. Good. Audit requirements mean every automated decision must be traceable. Good. Quality standards mean you can only ship something that works reliably. Good.

These constraints force you to build things that are actually solid. European industry has been doing this for decades. Precision engineering. Process discipline. Quality systems that hold up under pressure. The orchestration layer is the same philosophy applied to software. Software that reads between the tools. Just like that one guy on the floor does, but at scale, around the clock, across every system.

The companies that figure this out now will have more than a productivity advantage. They will have a compounding one. Because once your systems are coordinated, once the gaps are closed, every improvement you make to any single system benefits the whole operation.

The question nobody is asking at Hannover Messe

Every booth will demo AI this year. Every panel will discuss AI. Everyone will hear the word until it loses all meaning.

But here is the question I would actually like someone to answer honestly:

How many manual handoffs happen in your operations every single day?

It is the boring question. The unsexy one. But when you actually sit down and count, it explains exactly where your time and money are going.

That is the conversation I want to have in April.

Because at INXM, we are already building the answer.

Now let me show you what this looks like when it is real.

Imagine you are a buyer at a large enterprise. You wake up on a Tuesday morning. Coffee in hand. You have 180 suppliers under your responsibility. Contracts, quality scores, delivery performance, risk profiles, incoming goods receipts and open issues. All of it spread across SAP, your supplier portal, your quality management system, your email inbox and probably three or four Excel files that only you understand.

You have no idea that three of your suppliers are running out of contract today.

But INXM does.

INXM runs through all 180 suppliers. Every single day. It collects the data about their situation, their problems and their challenges. It pulls incoming goods receipts. It checks the contracts. It reviews quality scores, delivery reliability, the full performance picture and every risk signal in the system.

INXM knows that yesterday, three contracts were about to expire. It has already prepared everything you need for the call this morning. The supplier performance summary. The open quality issues. The delivery track record. The full negotiation preparation tailored to each supplier sitting ready for you.

And it has aligned everything in SAP. Checked. Prepared. Ready for you to review and go.

This is what orchestration looks like when it actually works. In a fast moving market, with processes and systems of record that no single human can manage alone anymore. INXM does that work. Quietly. Reliably. Every morning before you finish your coffee.

inxm.ai

“We deployed AI, but our organization isn’t getting any faster.”

It is a striking observation, considering the vast sums poured into copilots, assistants, generative search, and productivity tools. Yet the disappointment is not a failure of technology. It is the consequence of a fundamental misunderstanding of the type of work organizations actually need to automate.

Most companies today rely on two classes of AI systems. They are often blended together in conversation, but they operate on entirely different levels:

• One class improves the experience of work.

• The other changes the logic of work itself.

Confusing the two leads directly to stalled transformation efforts.

The first class: copilots, intelligent add-ons, AI search functions acts like vitamins.
- They are useful.
- They make daily tasks more comfortable.
- They reduce the friction of finding information, generating content, or summarizing meetings.
- They deliver answers, insights, and clarity.
- They reduce cognitive load.

But vitamins, no matter how effective, do not treat acute operational pain.
They do not change bottlenecks, workflows, or throughput.
They enrich knowledge.
They do not create significant movement.

And that is the crucial point.

Because what executives across industries actually seek has very little to do with “better information.”

When you speak with COOs, plant managers, heads of engineering, or sales leaders, the pattern is unmistakable.
- They do not want another tool; they have enough so far.
- They do not want another dashboard filled with insights.

What they want – though it is rarely phrased so directly – “is a system that takes work off the table”, not one that merely assists in doing it.
They want a system that works.
Not a system that helps them work better.

This is why the label “assistant” is misleading. Assistants answer questions, summarize content, and explain data. They are valuable but they do not alter the flow of operations. The number of handovers remains unchanged. The same tickets circulate. The same approvals clog the pipeline. The same Excel files travel across inboxes.

An assistant improves awareness.
It does not improve execution.

The turning point comes when organizations recognize that what they lack is not a new interface, but an entity that can operate their existing systems on their behalf.

That realization marks the entry point into a fundamentally different category:

Orchestration.

🚫Orchestration is not a chat window
🚫It is not a sidebar feature
🚫It is not an Add-on “AI inside a product”

Orchestration is

• ✅an Operational Layer

• ✅that converts information into actions,

• ✅and actions into completed work.

It is the context in which INXM operates.

INXM does not build another copilot or add-on. We provide a Cognitive Orchestration Layer that works inside a complex enterprise environment with the same standards of security, governance and auditability that define mission-critical operations. By connecting directly to your systems PLM, MES, QMS, ERP, engineering assets, regulatory rules and capacity data, all under the customer’s control, INXM enables AI to participate in the actual execution flow rather than observing it from the outside. It gives enterprises the mechanism for turning AI-generated recommendations into structured action across their core systems.

Where a copilot produces text, INXM. Orchestrator produces PLANs:

• structured, executable sequences of steps

• that run deterministically across your systems
  (ERP, PLM, QMS, CRM, ticketing)1

In this model, tasks are decomposed into atomic steps – each one linked to data, conditions, dependencies, and system interactions. These steps form a plan that is reproducible, auditable, and capable of spanning multiple systems end-to-end. The machine executes what it can autonomously and hands off to humans only where judgment, risk assessment, or accountability require it. This distinction is not theoretical. In daily operations, it is transformative.

• A customer request is not merely summarized and drafted into a reply. It is classified, checked against product data, validated against available capacity, simulated against material constraints, converted into multiple delivery scenarios, and prepared as a ready-to-approve offer.

• A production deviation is not merely explained. It is contextualized with specifications, logs, and historical cases; evaluated for risk; converted into a structured QMS entry; mapped to CAPA actions; distributed to the right systems; and followed through to effectiveness verification.

Through orchestration, operational work becomes a distributed system of PLANs:

• one plan extracts and analyzes incoming requests

• another checks capacity and material availability

• another creates deviations

• another executes and tracks CAPA steps

• another generates an offer

• another routes follow-up actions when thresholds are breached

• another triggers new plans based on completed steps

• …more 100x

No individual plan is revolutionary.
But collectively, they absorb a significant share of the operational load.
This is where organizations finally start to feel faster.

The real surprise for many executives is that they have not “implemented enterprise grade AI” – they have implemented another tools – copilots, assistants and add-ons. Helpful, impressive, often delightful assistants. But not systems that carry operational weight.

This explains why productivity remains flat despite AI expansions.
The problem is not model quality.
It is architectural.

Organizations that invest heavily in speech interfaces, copilots, and summarization tools without investing in orchestration will consistently find themselves better informed but not more capable. The bottleneck is not knowledge – it is coordination.

The strategic shift ahead is therefore not about better models, but about building an operational layer that can execute the work those models recommend. Copilots remain useful and should continue to be deployed. But they do not replace the need for a structural system that moves processes forward across the enterprise stack.

In this new frame, the difference between both classes condenses into a single sentence:

Copilot: 💡 “Here is the information you need.”

Orchestrator: ✅ “Here is the completed result. I’ve created and saved the related documents in PLM, started the Engineering Change Request, informed each project team domain about their specific impacts and updated the daily tomorrow including the project dashboard.” 

This is why orchestration is not just another AI category. It is emerging as the missing infrastructure that turns fragmented enterprise systems into a coherent operational engine – one capable of lifting work out of email threads, meetings, and manual handoffs, and into structured, machine-driven execution.

Companies that continue to invest exclusively in assistants will remain in a world of Vitamineffekte: helpful, pleasant, but ultimately superficial.
Companies that invest in orchestration will begin to shrink their operational workload in a way that no copilot, no matter how advanced, can deliver.

When executives today say, “We have AI, but we are not faster,” the solution is rarely another feature or another generative tool.
It is almost always the absence of orchestration – the layer where AI stops advising and starts working.

And that is the step that finally turns digital ambition into operational reality.

About INXM

Every large industrial enterprise runs its mission critical programs on a dense landscape of PLM systems, MES platforms, QMS environments, engineering documents, regulatory rules and capacity data. These systems keep the organization running, yet they rarely work together in a way that enables fast decisions or consistent execution. INXM is built to change that.

INXM provides a cognitive orchestration layer that operates inside your infrastructure and augments your workforce. It does not replace people. It works as an extension of them. When INXM triggers a workflow or sends a message into Teams or Jira, it acts in the name of the authenticated user and only within the permissions that user already has. INXM sees only what the user sees and performs only the actions the user approved. This preserves enterprise governance while unlocking a new speed of execution.

The Enterprise MCP Bridge, one of our essential parts of INXM. Orchestrator the controlled gateway through which AI can interact with your systems. The bridge turns arbitrary MCP servers into secure, observable, enterprise grade interfaces. It adds capabilities many systems do not provide on their own such as filtering, OAuth, routing, traceability and clear auditability. INXM never exposes internal systems externally. The bridge ensures that retrievals, tool calls and actions stay inside your network and remain bound to your authentication and access policies. This is how INXM integrates with complex landscapes without requiring architectural change.

This architecture also introduces self healing for mission critical workflows. INXM plans run through a managed execution layer with retry logic, failure tracking, dependency checks and success trees. Transient problems such as expired tokens, timeouts or schema mismatches can often be detected and resolved within the platform. IT teams no longer need external support after every update to fix brittle integrations or broken connectors.

INXM fits directly into existing IT operating models. It can run on premises or in a private cloud. It uses your authentication provider. It respects your RBAC setup. It connects through the network boundaries you define. It works with your preferred LLM setup or with hosted open source models provided through the TGI Manager. This gives enterprises full data sovereignty and full control of the AI execution environment.

With this foundation in place, organizations can orchestrate work where speed, precision and compliance matter most. Examples include engineering evaluations, quality deviation handling, regulatory checks, technical compatibility validation, corrective actions and cross functional coordination.

INXM gives enterprises a way to let AI operate across their landscape without redesigning architecture or compromising governance. It connects every part of a mission critical workflow into something reliable, auditable and repeatable.

INXM provides the cognitive orchestration foundation. Your teams provide the expertise. Together, execution across mission critical programs becomes faster, more resilient and enterprise aligned.

Recent months have been full of conversations about AI, transformation, and the uncomfortable question of why companies invest millions in new technology but experience almost no real change. In many organizations, the topic now simmers quietly under the surface. “AI” exists, expectations are high, use cases appear endless, yet daily life looks nearly unchanged. The same meetings, the same bottlenecks, the same manual coordination loops. Leadership voices: “We deployed AI, but the organization is not getting any faster.”

This gap has a root cause that is easy to overlook. It has little to do with algorithms, models, or tools. It has everything to do with culture, power structures, and the actual logic of work. AI forces companies to reveal how work truly flows. Not how it is described on paper, but how it actually happens. Informal, distributed, fragmented, political, friction-filled. This is where a critical realization emerges: Adoption is not a software problem. Adoption is a cultural phenomenon.

The Copilot Trap: Why Assistance and CompanyGPT Isn’t Transformation

Most companies begin their AI journey with a copilot. A tool that writes text, summarizes emails, explains information, performs minor analyses. It feels modern, elegant, productivity-enhancing. The first weeks are often impressive. People test limits, enjoy quick wins, and celebrate having a smarter assistant for daily work. But the curve flattens quickly. Usage stagnates, teams drift back into old patterns, enthusiasm fades, adoption stalls. A false conclusion emerges: AI does not work in daily life.

The truth is simpler. A copilot changes supports tasks but not the dynamics of work. It sits at the edge of the system, not in the center of your workflows. It provides answers, but coordinates nothing. It supports you creating ideas and insights, providing you recommendations in best case, but does not shape flow. It is helpful, but it does not transform a process. Transformation is created by coordination, not assistance. This is the blind spot in most initiatives.

The Pivot to Cognitive Orchestration: Relieving the System

Once AI moves beyond preparing decisions and begins coordinating the steps that follow, the organization reaches a natural next stage. Responsibilities become clearer, process chains become more predictable, and automated handovers link departments in a way that reduces friction rather than creating it. What previously required dozens of small manual actions can now flow with structure and reliability. This is not a loss of control. It is an opportunity for teams to focus on decisions instead of pushing work through the system.

Many companies discover at this point that their speed limits were never caused by missing technology. They were the result of processes designed for another era, where every step required verification and every handover triggered another alignment. Cognitive Orchestration does not break these structures. It relieves them. A copilot helps individuals work faster. An orchestrator helps the whole organization work cleaner, with fewer stops and more confidence in the flow of work.

From “Help Me” to “Move the Work Forward”: Making Adoption Stick

Copilots deliver quick wins and reduce personal workload, but they do not change how tasks travel across the enterprise. Work still moves manually from person to person. Faster, yes. But not more scalable. That is why early excitement often levels off: the underlying system has not evolved.

With cognitive orchestration, that changes. Pockets of efficiency connect. What used to be parallel efforts becomes a single coherent process. Teams no longer rely on workarounds or shadow workflows. Leaders no longer see only insights; they see momentum. The organization feels the difference between “AI helps me” and “AI moves the work forward.”

This is the natural next step after copilots. Not disruptive. Not threatening. Simply the progression from individual productivity to organizational performance, the point where AI becomes part of how the business runs rather than something that sits on the side.

The Turning Point: When AI Starts Acting

Transformation begins when AI triggers actions, synchronizes data, connects departments, and orchestrates workflows. When AI becomes part of the operational rhythm. When it does not describe work but performs it.

Consider a few examples:

• A single late-delivery signal launches an orchestrated multi-agent chain.

• One deviation triggers twelve interconnected plans.

• A customer request activates more than fifteen plans in parallel and brings them together into one result.

more… (mail me to discuss more)

This is the shift. Work flows as structured execution, not recommendations. AI is not asked; AI acts. The system performs steps autonomously and calls humans only for exceptions. The organization begins to relieve itself instead of simply informing itself.

The Innovation Ceiling: The Danger of Stopping Early

The biggest mistake today is declaring victory after deploying a copilot. Statements like “we now have AI” or “we introduced a copilot” create a false sense of progress. It is like buying a car and only using the seats. A company builds its own innovation ceiling. A polished facade with an unchanged core.

The cost is high. Missed velocity gains. Missed quality improvements. Missed cross-functional synergy. And most importantly, an organization that believes it has transformed while nothing fundamental has changed.

Don’t Just Optimize the Old World. Build a New One

Where this leads is a new operational reality.

Companies using AI for assistance make people faster. Companies using AI for orchestration make work faster. This is the difference between optimization and transformation.

The operating model changes fundamentally. Fewer handovers. Fewer dependencies. Fewer wait times. Fewer searches. Fewer manual routing steps. In parallel, more clarity. More consistency. More visibility into risks, bottlenecks, and decisions.

A production logic emerges that no longer depends on manual steering. Teams shift from gatekeeping to supervising. Decisions become traceable. Processes become auditable. Quality rises through structure, not oversight.

This is where the next evolution of work begins. The game changer. The next thing companies are searching for.

The Final Picture: Optimization vs. Reinvention. Transformation is a path.

The future belongs to organizations that treat AI not as an assistant but as an operational coordination layer, not as a gadget but as process intelligence.

The true shift begins with a simple realization: typing faster does not make the enterprise faster. Workflows must change. This separates organizations that merely adapt from those that reinvent themselves. Those who settle for assistance will build a slightly improved version of their old world. Those who embrace orchestration will build the next generation of industrial operations and shape the decade ahead.

About INXM

Every large industrial enterprise runs its mission-critical programs on a complex landscape of PLM systems, MES platforms, QMS environments, engineering documents, regulatory rules and capacity data. These systems keep the company running, but they rarely work together in a way that enables fast decisions or consistent execution. INXM is built to change that.

INXM provides an cognitive orchestration layer that connects directly with systems and people, to your authentication system, infrastructure, network and model setup. This allows AI to operate inside your environment with the same security and compliance expectations that already govern your mission-critical operations.

At the center of the platform is the Enterprise MCP Bridge, which makes your systems accessible to AI in a structured and controlled way. It supports filtering, OAuth, and routing so the cognitive orchestration layer can retrieve requirements, specifications, guidelines, regulatory inputs, capacity data, historical deviations, or ticket information without exposing systems externally. All integrations remain under your ownership.

Because INXM runs inside your infrastructure, it fits directly into existing IT operating models. It can run on-prem or cloud, connect to systems and people and respect the authentication and access policies already in place. It can use your preferred LLM setup or hosted open-source models where appropriate.

This architecture makes it possible to orchestrate work where speed, precision and compliance matter most. Companies can use INXM to support engineering evaluations, quality deviation handling, regulatory checks, technical compatibility validation, corrective actions, and cross-functional coordination.

INXM gives enterprises a way to let AI operate across their landscape without changing the underlying architecture. It ensures every part of a mission-critical workflow is connected, secured, and repeatable across the organization.

INXM provides the cognitive orchestration foundation. Your teams provide the expertise.

Together, execution across mission-critical programs becomes faster, more reliable, and fully aligned with enterprise governance.

• The old way lives inside systems that only talk to people. Every delay depends on someone catching what a system cannot see. Every correction arrives late because it relies on attention.

• The new way connects systems to each other but overwhelms people with more information than any team can absorb. Speed rises. Control falls. Teams swim in signals, not clarity.

• The emerging way introduces something different. An Orchestration Layer that understands intent, guides actions across the entire landscape, and protects the organization from the tiny cracks that quietly grow into strategic problems. It is not a replacement for systems or for people. It is a missing layer that finally understands both.

Remaining in the first two ways means slow reactions and fragmented responsibility. Decisions are shaped by what happens to be noticed. Moving into the third means the system carries the complexity. People regain clarity. The enterprise stops losing ground in moments that should never have been surprising.

The idea of a layer that serves both systems and people

Inside almost every large organization, I see a new structure is taking shape. It does not erase what exists. It elevates it. A layer that reads how work should flow, understands the rules beneath it, and acts through every system involved.

It bridges the point where tools stop being helpful and people start becoming tired. It listens. It interprets. It moves work forward without waiting for someone to piece the picture together by hand.

In this era, progress no longer depends on a person spotting a single change at the perfect moment. The layer stands between the scattered machinery of the enterprise and the people who guide it. It translates intention into action and complexity into clarity.

How the layer behaves

This layer behaves in a way no traditional workflow engine ever could.

• It speaks machine language and human language.

• It interprets, reasons, sequences, and acts.

• It connects signals to consequences across the entire landscape.

What once required messages, meetings, and manual recalculation now unfolds as one coordinated response. A shift in capacity triggers adjustments in procurement. A risk in supply ripples through scheduling. People step in only when their judgment is needed.

Picture a change in demand. In the old pattern someone edits spreadsheets, informs production, contacts suppliers, and chases confirmations. With the new layer the actions move orchestrated from system to system and people. The organization behaves like a single body instead of a set of separate departments.

How is it different from a chatbot?
A chatbot answers questions. An Orchestration Layer does something categorically different: it interprets signals across systems, resolves contradictions, sequences actions, and moves work between humans and machines. It doesn’t wait to be asked it works continuously, like an operational nervous system. No rule-based workflow engine can do this, because the complexity changes frequently.

What an enterprise Orchestration Layer actually means
An Orchestration Layer sits above all your systems and people. It listens to signals from systems and people. It interprets what’s happening, and orchestrates specialized agents to take safe, auditable actions keeping the employees in the loop. Instead of every team navigating dozens of tools, the Orchestration Layer becomes the single coordination point, understanding the full picture and ensuring every workflow runs the way the business intends.

Example to make the idea concrete
Imagine a late supplier delivery. A chatbot would merely tell you the expected delay. An Orchestration Layer automatically checks inventory buffers, recalculates the production plan, evaluates customer impact, triggers supplier escalation if needed, and asks a human only for the final approval. One instruction in natural language triggers a full chain of safe, traceable actions.

Why enterprises that adopt this layer pull ahead

Organizations that adopt this layer gain a structural advantage. They reduce the hidden workload that employees once carried. They narrow the space between understanding and action. They avoid tiny operational issues that often escalate into strategic cost.

The complexity moves into the system that can handle it. The clarity returns to the people who deserve it.

The INXM Orchestrator platform is built to deliver this layer. It provides a reliable foundation where systems and people operate together with speed, consistency, and control.