Articles

AI Doesn’t Respect Your Silos. Neither Should You.

AI dissolves knowledge boundaries across departments, but that doesn’t automatically fix decision-making. This article explores why cross-functional work now needs structured collision, not more noise.

March 18, 2026, 9 min read. By Next CoLab.

Extreme Collaboration Series — Article 3

For decades, we restructured into matrixed organizations and agile squads to break down silos. Mostly, we failed. Turns out, AI doesn’t care about any of that. It just synthesizes across domains without asking permission.

The Meeting Nobody Can Decide

Your VP of Product spent three hours yesterday in a cross-functional AI strategy meeting. Engineering presented technical feasibility. Marketing showed customer research. Finance ran the numbers. Operations outlined implementation risks.

Everyone shared dashboards. Nobody made a decision.

“We need another meeting to align on this.”

This is the new crisis. Not information access. Integration.

Why Silos Are Dissolving (Finally)

Cross-functional collaboration has failed in the past due to translation friction. For a developer and a salesperson to work together, they had to spend hours explaining basic constraints. Engineering sees risk where Marketing sees opportunity. Finance speaks in margins while Product speaks in features.

The cognitive cost was so high that people retreated to their silos.

AI just eliminated that barrier.

Harvard Business School’s March 2025 field experiment with 776 Procter & Gamble professionals proved this (Dell’Acqua et al., 2025). Without AI, R&D suggested technical solutions. Commercial teams pushed market proposals. Classic silos.

With AI acting as a “Cybernetic Teammate,” both groups produced balanced solutions integrating technical and commercial perspectives equally. Teams using AI were significantly more likely to generate top 10% ideas.

AI dissolves the knowledge boundaries between departments. The universal translator has arrived.

Now we get to discover whether silos were a translation problem or a power problem.

When Everyone Can Access Everything

Microsoft’s 2025 Work Trend Index quantifies what happens next: 80% of workers lack the time or energy to do their jobs. Interruptions hit 275 times per day (Microsoft, 2025).

This is the hidden cost of “breaking down silos.”

When anyone can access any document, when anyone can invite anyone to any meeting, when AI can synthesize information from five departments in minutes, the bottleneck shifts:

Old bottleneck: Can I get Engineering’s data?

New bottleneck: Can I process conflicting input from Engineering, Marketing, Finance, Legal, and HR without losing my mind?

More collaboration inputs do not equal better collaboration outputs.

Research on collaborative overload shows the most connected 3–5% of employees become bottlenecks, burning out while everyone else waits (Cross, Rebele, & Grant, 2016). AI democratized connection. It didn’t democratize the capacity to synthesize it.

We’ve moved from information scarcity to integration overload.

The Question Nobody’s Asking: Whose Expertise Wins?

Here’s what we don’t discuss when celebrating “collaboration across silos”:

Silos weren’t just walls. They were hierarchies.

“Hard” skills like Engineering and Finance typically outrank “soft” skills like HR and Communications. Quantitative analysis gets more airtime than qualitative insight. The person who “speaks data” dominates the person who “speaks culture.”

If AI enables cross-disciplinary synthesis, we have to ask: Whose logic runs the algorithm?

When AI summarizes a product meeting: Does it lead with technical constraints (Engineering’s frame) or customer pain points (Customer Success’s frame)? The sequence determines whose expertise becomes foundational versus supplementary.

When AI optimizes a workflow: Does it optimize for efficiency (Operations) or employee experience (HR)? Which domain’s constraints become non-negotiable versus “nice to consider”?

When AI generates strategic options: Which discipline’s vocabulary gets privileged? Whose trade-offs get foregrounded?

These aren’t technical questions. They’re power questions.

Microsoft found that 67% of leaders are familiar with AI agents, compared with 40% of individual contributors (Microsoft, 2025). This isn’t a learning gap. It’s a power gap.

If you don’t democratize AI literacy across disciplines and levels, AI won’t break down silos. It will automate whichever discipline already has the most status.

The technology dissolves the walls. The hierarchies remain, running at AI speed.

From Open Doors to Structured Collision

If AI removes natural barriers between departments, you need intentional protocols. Not walls that block information. Filters that determine whose expertise gets decisional weight.

Structured Collision is the answer: explicit frameworks for when cross-functional teams convene and whose expertise leads on which decisions.

Not “let’s get everyone in a room and see what happens.” But “here’s the decision, here’s whose expertise has primary weight, here’s how we adjudicate.”

What Structured Collision Actually Looks Like

Scenario: A university deciding whether to launch a new academic program.

The Old Way: Months of emails between Academic Affairs, Finance, and Marketing. Information gets shared but never integrated. Meetings multiply. Decisions drift.

Structured Collision:

  • Before the Meeting: AI synthesizes data from all departments into one briefing. Enrollment projections, financial models, competitive landscape, and faculty capacity. Conflicts and alignments are mapped visually.
  • The Meeting (One Session, 60 Minutes): The cross-functional team focuses only on judgment calls. Not “what does the data say” (AI handled that) but “Does this align with our identity? What’s our risk tolerance? What are we trading off?”
  • The Protocol: For academic program decisions, Academic Affairs has primary decisional weight. Finance and Marketing identify constraints but don’t override. For budget decisions, Finance leads. The expertise hierarchy rotates based on decision type, not org chart.
  • After the Meeting: Document which logic prevailed and why. Make patterns visible over time so you can see whose expertise consistently gets heard versus ignored.

AI handles translation. Humans handle judgment. Protocols ensure judgment doesn’t default to whoever has the most status.

Three Principles for Making This Work

  1. Invest in Synthesizers
    The scarcest skill isn’t sharing expertise or forwarding AI summaries. It’s integrating conflicting truths from different domains.
    • Identify which conflicts are real trade-offs versus semantic confusion
    • Know when diverse input needs adjudication versus synthesis
    • Translate between domains after AI does the literal work
    This isn’t project management. It’s curatorial judgment about what to combine, what to keep in tension, what to decide.
  2. Rotate Decisional Weight Explicitly
    Create protocols for whose expertise leads on which decisions. Customer experience decisions? Customer Success has primary weight. Technical architecture? Engineering leads. Strategic positioning? Marketing and Product co-lead. The discipline with decisional authority rotates based on decision type, not hierarchy. Make it transparent. Document it. Review whether the same voices always win.
  3. Convene Only for Decisions
    Information sharing is a document. Status updates are a dashboard. Brainstorming is async AI ideation plus a sync decision session. Cross-functional teams meet synchronously only when there’s a specific judgment call requiring combined expertise. Not “let’s align” but “here’s the decision, here’s the pre-processed brief, here’s whose expertise decides.”

You’re Already Facing This

  • Siloed AI Deployment: Marketing has one tool, Finance another, Engineering a third. You automated the walls.
  • Same Voices, Faster: AI accelerated how quickly the highest-status person’s framing becomes consensus. Cross-functional meetings still end with one discipline winning, just with better slides.
  • Integration Burnout: Your most connected people drown because AI generates infinite connections without matching integration capacity.
  • Meeting Explosion: Volume increased because AI makes meetings feel “cheap” to schedule and “easy” to summarize (summaries nobody reads).

The Monday Morning Test

Pick one cross-functional decision happening this month. Ask:

  • What will AI synthesize before the meeting? (All department inputs, conflicts mapped, options generated)
  • Which disciplines need to be in the room for the judgment call? (Only those whose expertise is essential to the decision)
  • Whose expertise gets primary decisional weight, and why? (Explicit protocol: for this decision type, this discipline leads)
  • How will you make visible whose logic prevailed? (Document the decision and the reasoning so patterns become visible)

If you can’t answer these questions, your silos haven’t been broken down. They’ve been made invisible while cross-functional noise multiplies.

What You’re Actually Building

Silos were inefficient but they were safe. They protected people from complexity they didn’t need to process.

AI removed that protection. Everyone’s in the same room now, accessing the same information, invited to the same meetings.

The organizations that win won’t just “connect everyone.” They’ll build capacity to integrate those connections without losing their minds.

That means protocols for when cross-functional teams convene. Explicit frameworks for whose expertise gets decisional weight. People who can synthesize across domains. And the willingness to make visible whose voices get heard versus whose get ignored.

The technology dissolved the silos. The design work of Structured Collision is how you turn that into advantage instead of chaos.


Sources Cited

  • Cross, R., Rebele, R., & Grant, A. (2016). Collaborative overload. Harvard Business Review.
  • Dell’Acqua, F., et al. (March 2025). The cybernetic teammate: A field experiment on generative AI reshaping teamwork and expertise. Harvard Business School Working Paper, No. 25-043.
  • Microsoft. (April 2025). 2025 Work Trend Index annual report: The year the frontier firm is born.

This is Article 3 of the Extreme Collaboration series. Learn more about building extreme collaboration capacity at Next CoLab, where behavioral science, business science, and AI come together to measure what’s actually happening in cross-functional collaboration and redesign it for decisions, not drift.

More in this series