AI Won’t Fix Silos — Org Design Will

Everyone is racing to adopt AI. Developers are coding faster with Copilot. Analysts are producing reports with GPT. Recruiters automate job descriptions, and marketers launch campaigns in minutes.

But most of these initiatives only create local productivity boosts. Teams move faster in their own silos, while the big initiatives — the bold moves that actually matter to customers and strategy — remain stuck.

Why? Because the wrong organizational design doesn’t just slow you down; it actively blocks progress. Layer AI on top of dysfunction, and the dysfunction only accelerates.

Resource Topology: Faster Silos, Slower Outcomes

In a Resource Topology, work is divided into specialized units (people, departments, teams). Analysts hand over specs, developers code, testers check, and an integration team stitches it all together at the end. This design optimizes for utilization, predictability, and centralized control.

Add AI into the mix, and each unit becomes faster. Analysts can deliver specs faster, leads can do more efficient planning and work prioritization. Developers produce more Android code per day, and testers can automate scripts in seconds. There is definitely progress.

But when we look closer, we see that the overall speed of the system doesn’t improve. That is because adding AI does not change the way the system works. We still have handoffs between units, and they multiply the delays in the work. Decision-making remains unclear, creating noise in the system as work is bounced between units with limited mandates. Distributed decision rights, tied to risk and audit processes, restrict automation instead of enabling it.

The performance gain is local, resulting in Faster queues, with slower outcomes. A classic law of systems thinking comes into play:

“The performance of a system depends on how the parts fit together, not on how they perform taken separately. When you optimize the parts, you sub-optimize the whole.” (Russell Ackoff)

The logical path of evolution is obvious. The narrow, repetitive roles in these silos are replaced by AI agents. That optimizes local productivity gains and cost savings. But it doesn’t make the organization as a whole more adaptable. In fact, it leaves the system as rigid as before, creating queues of work between AI agents.

In a Resource Topology, AI typically delivers local wins. Assistants are deployed inside existing silos, which makes individual units faster but has little effect on end-to-end outcomes. The handoffs, approvals, and fragmented data flows remain unchanged, so queues continue to grow. Because decision rights are distributed across many groups, automation is limited and often slowed down further by risk and audit processes. The focus remains on optimizing local output, rather than improving the system's overall performance.

Delivery Topology: Faster Streams, but Still Rigid

Many organizations have evolved into Delivery Topologies. Cross-functional teams are restructured into value streams. Instead of analysts and developers working separately, they sit together and ship features from start to finish. In the best scenario, they have all the skills in the team needed to deliver end-to-end items of work.

This is a genuine improvement. When we add AI to these teams, it will accelerate each value stream — features get delivered more quickly, backlogs shrink, and the teams feel empowered. The mortgage team, for instance, can now process applications end-to-end with remarkable speed.

But there’s a hidden ceiling. Each stream is still a silo, locked into its own domain. The mortgages team won’t deliver insurance. The insurance team won’t deliver business loans. When a new customer need arises, the organization must spin up a new silo to serve that need.

This design optimizes for speed at the team level and delivers what is already known. This system is fast at the value stream level, but not highly adaptive. Every strategic pivot requires reorganization.

And although the value streams are fast, in the long run, they will stop producing value. That is due to the law of diminishing returns: once AI has extracted most of the speed gains from each stream, the cost of changes begins to outweigh the value, even when change is cheap. There is no more value for the customer in yet another change to the mortgage processing.

Leaders will apply cost-saving measures — replacing people with AIs in routine parts of the value stream. Again, it creates local efficiency, but not systemic adaptability.

In a Delivery Topology, AI produces mostly local improvements. AI assistants are deployed inside individual silos, which speeds up work but has only a limited effect at the value stream level. Innovation ultimately caps out because it is confined within these narrow value streams. The emphasis remains on optimizing local outputs, not on unlocking broader organizational performance.

Adaptive Topology: Where 10X Becomes Possible

True transformation comes with the Adaptive Topology. This design is not very common. It requires giving teams of teams the mandate to work on the whole business problem or customer need, not just a slice of it. And equipping them with all the skills and tools required to succeed.

In an Adaptive Topology, people are multi-learners. They acquire knowledge across domains, enabling the team of teams to deliver outcomes autonomously. They refine work together, resolve dependencies directly, and coordinate face-to-face rather than through endless external handoffs.

Here, AI shows its full potential. It is not trapped inside silos. It is applied to shared outcomes. Models are trained on customer data to spot shifts in demand. Generative AI accelerates prototyping and validation across the whole problem space. Moreover, cross-cutting concerns such as legal, security, and compliance are embedded, so everything produced complies with the standard requirements by design.

In this environment, there is no need for disruptive reorganizations when strategy changes. Adaptation is built in. People and AIs collaborate holistically to learn, innovate, and deliver.

And because people are multi-specialists — “M-shaped” instead of narrow — they are not candidates for replacement. They are indispensable. AI doesn’t reduce their relevance; it amplifies it.

In an Adaptive Topology, AI is applied to shared business outcomes, which amplifies collective learning and impact. Because AI is integrated holistically, there is no overhead from fragmented, decentralized implementations. The system is holistic from the start, and AI strengthens that quality.

First Design, Then AI.

The archetypes in each quadrant of the map will have specific benefits of AI adoption:

Benefits of AI Adoption in the four Archetype GroupsIn each quadrant, people will work alongside AIs, but with different outcomes at the whole company level.INCOMPLETECOMPLETEReduction of planning time; Data-driven insights for strategic decisioningAmplified innovation; Data-informed strategic decision making; Supply-chain fusionEnhanced productivity; Reduction of cognitive load; Consistent execution qualityReduction of silos through integration; Alignment of outputs with strategic goalsSCOPE OF WORK MANDATEmorelessSCOPE OF SKILLS MANDATElessmoreOUTCOMESOUTPUTSLegend: human, AI agent, pool, team, team of teamsOrg Topologies™ by Alexey Krivitsky, Craig Larman, Roland Flemm. Version 2.0, 02.2025. CC BY-NC-SA

Hover over or tap any text on the drawing to read it.Tap any text on the drawing to read it.

Every organization has a design. Yet most leaders underestimate how profoundly the org design shapes the impact of AI. Without redesign, applying AI guarantees disappointment, because the expected gains are unrealistic.

The contrast is clear:

• Resource Topology → AI delivers local efficiency, but the system remains stuck.
• Delivery Topology → AI makes teams faster, but the organization stays rigid.
• Adaptive Topology → AI amplifies adaptability, enabling the whole business to move in sync with strategy.

AI does not fix design flaws. It amplifies them. To unlock its promise, leaders must first design organizations fit for change — and only then bring in AI.

More on AI and value streams in this video recording.

More on Strategic AI.

More on elevating to the Adaptive Topology.