This article is originally published on Forbes on 17 August 2026 [Link to original article]
Imagine a customer raises a complex issue. Today, it may move from customer service to operations, finance, legal and product. Each function reviews its portion, updates its own system and protects its own priorities. By the time the organisation responds, the customer has repeated the story several times and nobody fully owns the outcome.
Now imagine an AI-enabled workflow that gathers the history, identifies policy and financial implications, proposes response options and alerts the people whose judgment is required. No department has disappeared. But the department no longer determines how the work moves.
This is the deeper organisational question raised by AI. The discussion is usually about productivity: which tasks can be automated, how much time can be saved and how many agents can be deployed. AI may make the existing organisational model obsolete. Departments may remain homes for expertise while work increasingly forms around customers, products, strategic priorities and end-to-end outcomes. The leadership challenge is therefore not simply to implement AI. It is to redesign the system of work. Four systems lenses can help.
1. Diagnose How Silos Persist
The Levels of Perspective model distinguishes visible events from patterns, systemic structures and mental models. At the event level, leaders may see delayed approvals, duplicated work or conflicting recommendations. At the pattern level, the same customer issue repeatedly crosses several functions. At the structural level, separate budgets, systems, incentives and reporting lines reinforce fragmentation. Beneath these structures sit assumptions such as “my department must control this decision” or “sharing data creates more risk than value.”
Many AI projects address only the event. They automate a handoff, accelerate an approval or add a chatbot. The symptom improves, but the larger system remains unchanged. Before deploying AI, choose one high-friction outcome and examine it at all four levels. Ask what keeps recurring, which structures produce that pattern and which assumptions make those structures appear reasonable. This prevents the organisation from digitising the silo instead of redesigning it.
2. Manage The Tension Between Today And The Future
Healthy tension helps leaders hold two realities at once: an honest account of current conditions and a compelling view of the future. The current reality may be that customer issues cross five functions, decisions take 10 days and each function is measured separately. The desired future may be a coordinated workflow in which specialist expertise is drawn in when needed and one person remains accountable for the total outcome.
Leaders often reduce the tension in the wrong direction. They lower the vision until it fits the current structure. The new AI workflow then becomes another tool inside the old process. Instead, define the future operating principle before selecting the technology. For example: “Work will be organised around the customer outcome, not the departmental sequence.”
3. Map The Work As A System
A systems view asks leaders to define the larger purpose, identify the essential parts and examine their interdependencies. For an AI-enabled customer resolution system, the parts may include customer service, finance, legal, product, data, risk and the AI platform. The critical question is how tightly their work depends on one another. Some relationships require periodic coordination. Others need continuous information exchange and rapid response. Highly interdependent work should not rely on slow handoffs and occasional meetings.
Map the workflow from beginning to end. Identify where information enters, where judgment is required, where decisions are delayed and where the same data is recreated. Then classify what AI can execute, what it can recommend and what must remain a human decision. This also clarifies the future role of departments. They become capability homes that maintain standards, develop expertise and govern professional quality. Outcome teams draw on those capabilities without being trapped by the departmental boundary.
4. Design For Acceptance And Learning
A technically elegant AI system can still fail if people do not trust it, understand it or see how their role changes. The Q x A = E principle makes this explicit: The quality of the technical solution, multiplied by acceptance by the people, determines the effectiveness of the result.
Acceptance means involving the right people early enough to improve the design, surface risks and build ownership. For each redesigned workflow, clarify who owns the outcome, who sets the guardrails, who may override the system and who is accountable when the result causes harm. Run a bounded pilot around one outcome rather than reorganising the whole company at once.
Do not add a human approval at every stage merely to create a sense of control. That recreates the old bureaucracy in digital form. Human intervention should be concentrated where judgment, ethics, exceptions or material consequences are involved. Then create a feedback loop. Track one leading indicator, one outcome indicator and one balancing indicator. The final question in every review should be: What will we change because of what we have learned?
Get From Faster Silos To A Learning System
AI will not eliminate the need for finance, legal, marketing, human resources or operations. It will challenge the assumption that every piece of work must travel through those functions in sequence. However, the future organisation may operate through a dual structure. Departments preserve expertise and professional standards. Fluid teams combine human judgment, specialist capability and AI agents around an outcome.
This requires leaders to give up an old source of control. Authority can no longer depend mainly on owning information, headcount or process steps. It must come from setting direction, designing conditions, making trade-offs and protecting the integrity of the whole system. This identity shift may prove harder than the technology deployment. Functional leaders must be willing to contribute expertise without defending territory. They may also have to accept that optimising their department can sometimes weaken the performance of the larger organization.
Organisations that merely insert AI into departments may make each silo faster. Those that redesign vision, structures, interdependencies and learning loops around outcomes can become something more valuable: an organisation that works, adapts and learns as one system.