This article is originally published on Forbes on 24 September 2026 [Link to original article]
One of the challenges in teaching systems thinking is that managers already understand the language of silos and collaboration. What they rarely experience is how intelligent, well-intentioned decisions can combine to produce a poor organisational result.
Consider what I call an “XYZ learning simulation”: a 90-minute AI-enabled business activity designed to let participants experience systems blindness rather than simply discuss it. “XYZ” in this case means an organisation with departments that work in silos, receiving an order upstream and sending it downstream after their input. These departments do not know or care about the entire flow, as their KPI is just their own processing efficiency.
In the simulation, participants take on different business roles (e.g., Marketing, Accounting, Pricing) and use AI to solve the problems facing their own function and then compare those locally optimised solutions with what becomes possible when the functions work as one connected system. The central lesson is simple: When you optimise the parts, you will sub-optimise the whole some of the time. Let us break this activity down.
Step 1: Let Each Function Optimise To Solve Its Own Problem
Participants are assigned roles like Technology, Marketing, Finance and Frontline Operations. Everyone works on the same business challenge, but each team sees it from its own functional perspective.
Technology may focus on system reliability and scalability. Marketing may concentrate on positioning. Finance may focus on pricing, revenue and risk. Frontline teams may worry about explaining the new offer while still meeting existing sales targets. None of these perspectives is wrong; each function has good reason for behaving as it does.
Each team then uses a generative AI assistant to analyse its situation and recommend what its department should do.
The crucial point is the framing. Teams are effectively asking: How can our function perform better? The results can be strong: Technology improves delivery. Marketing sharpens the customer proposition. Finance strengthens financial control. Frontline teams protect sales performance.
Yet when all those recommendations are combined, the overall business result can still be disappointing. Marketing may promise something Technology cannot yet support. Finance may make a prudent decision that discourages frontline adoption. Technology may build features disconnected from customer behaviour.
This is the first lesson from XYZ: Good departmental decisions do not automatically create a good organisational outcome.
The secret weapon: There’s a simple technology layer behind the learning.
The participant experience should remain simple. Teams can use familiar AI tools, discuss their functional challenge and submit their recommendations through a straightforward interface.
Behind the activity sits an AI-Facilitator Operating System. Its role is to manage the simulation, receive team submissions and assemble the different functional decisions into one organisational picture. It contains a rubric to assess participants’ answers and suggestion improvements.
A dashboard, one element of the operating system, gives the facilitator a live view of the business. It can show each functional team, what it has decided, where information is flowing and where important connections are missing.
This creates two useful perspectives. Participants initially see the problem from inside their own department, while the facilitator can see the broader system and reveal it to the participants after step one.
From my experience, this activity can be run in the classroom or online synchronously. The important requirement is that participants are making decisions at the same time from different functional perspectives and can subsequently see how those decisions interact.
Step 2: Change The Unit of Optimisation
The second stage changes one important instruction. Instead of asking, “How do we improve our department?” participants are asked, “How do we improve the end-to-end business outcome?”
Teams now examine one another’s decisions and identify dependencies. What information must move between functions? Where are incentives working against one another? Which decisions made in one department create consequences elsewhere?
AI is used again, but differently. Instead of helping each function optimise independently, it can help participants compare perspectives, surface contradictions and construct a shared workflow.
The result is usually very different. Technology becomes connected to adoption and usage. Marketing becomes connected to the actual customer experience. Finance becomes involved in designing pricing and incentives that enable the transition. Frontline teams become an important source of customer intelligence.
The output is no longer five departmental plans; it begins to look like an integrated operating model, with shared outcomes, feedback loops and connected decision-making. I call this the “XYZ system-level response,” an aligned workflow that examines efficacy from end to end. There’s no point creating a single speedy department if another department serves as a bottleneck; the system is considered aligned only when the workflow is optimised from start to finish.
The second lesson is therefore that integration creates value that cannot be achieved simply by making every department more efficient.
Step 3: Turn The Difference Into Learning
The final stage is the debrief. Participants compare what happened when departments optimised independently with what happened when they worked around a shared outcome. The facilitator can use the dashboard to show the difference between strong activity inside functions and stronger connections across the system.
Technology may have assumed that better capability would create adoption. Marketing may have assumed that clearer messaging would create alignment. Finance may have assumed that certainty should precede commitment. Frontline employees may simply have behaved according to the measures against which they were being judged.
That is what makes XYZ useful as a learning experience. Participants see that competent people, powerful AI tools and individually strong solutions can still create weak outcomes when everyone is optimising a different part of the system.
The bigger leadership question is therefore not simply “How can AI make each function more productive?” It is “What are we asking AI, and our people, to optimise?”
Conclusion
If AI is used only to improve individual functions, organisations may simply become faster at local optimisation. With the XYZ exercise, participants do not merely learn what systems blindness means. They create it, experience its consequences and then discover what changes when the problem is reframed at the level of the whole system.