Team Structure Is the Missing Link: Why the Deloitte AI Value Gap Closes With Bigger, More Diverse, Better-Connected Teams

See also (wiki): wiki/ai-talent-workforce-planning.md, wiki/ai-change-management.md, wiki/workflow-redesign.md

Executive Summary

• Deloitte Insights (Brodzik, Kulkarni, Mahto, Kreit — Feb 27, 2026, n=1,394 U.S. working professionals, July 2025 survey) identifies three team-structure variables that correlate with AI value capture: team size, cognitive diversity, and cross-functional connectedness. Teams of 10 or more report 2x the AI-driven innovation, problem-solving, and efficiency gains of teams with four or fewer.
• Cognitive diversity is the strongest lever. 91% of high-performing AI teams deliberately hire for varied skill sets (vs. 68% of other teams); 86% prioritize diverse past experiences (vs. 51%). Employees who feel their organization overlooks diversity of thought in AI design are 60 percentage points less likely to use AI tools daily.
• Trust and learning orientation shift the adoption curve. 83% of high-trusting teams use AI vs. 63% of others. High-benefit teams are 2.5x more likely to frame work as a learning opportunity and 2.1x more likely to plan role reshaping as AI evolves.
• The investment imbalance is stark. Deloitte’s companion Tech Trends research reports 93% of tech funding directed to technology itself; only 7% to training and upskilling. The 10-20-70 pattern (BCG) and the team-structure pattern (Deloitte) point to the same problem: budgets underwrite the tool, not the team that has to make it work.
• For a 200-2,000 person American company, the Monday-morning implication is concrete: before funding the next AI pilot, audit the team it is landing in. Headcount under five, single-discipline composition, and low cross-functional connection predict underperformance no matter what tool you buy.

The Team-Structure Variables That Predict AI Value

Deloitte’s survey treats AI-derived benefits — innovation, problem-solving, efficiency — as the dependent variable and tests team composition as the predictor. The three variables that move the outcome are size, cognitive diversity, and connectedness.

Team size. AI usage rises sharply with team headcount. 74% of workers on teams with 10+ members report using AI at work, compared to 54% on teams with fewer than five. Self-reported benefits follow the same curve — larger teams report roughly 2x the innovation, problem-solving, and efficiency gains of the smallest teams. The survey does not establish causation, but the pattern is consistent with a simple mechanism: larger teams have more surface area to run experiments, more peers to learn from, and more combinations of tasks to which AI can be applied.

Cognitive diversity. The gap between high- and low-performing teams on hiring practice is among the largest in the study:

• 91% of high-performing AI teams actively hire for varied skill sets (vs. 68% of other teams)
• 86% prioritize diverse past experiences (vs. 51%)
• 54% regularly incorporate diverse viewpoints into decisions (vs. 27%)

The inverse signal matters for change management. Employees who feel their organization overlooks diversity of thought in AI design are 60 percentage points less likely to use AI tools daily. In a workforce where a typical mid-market rollout targets 60-70% daily use, a team that feels excluded from design cuts that number roughly in half.

Connectedness. High-benefit teams score higher on three cultural measures: 2.5x more likely to view work as a learning opportunity, 1.9x more likely to feel empowered to make decisions, and 2.1x more likely to plan for reshaping roles as AI evolves. Trust is the most operational metric — 83% of high-trusting teams use AI versus 63% of others, a 20-point gap that maps directly onto the “pilot adoption” number most executive dashboards already track. Cross-functional teams are 30% more likely to report significant efficiency and innovation gains than single-function teams.

Why This Complements — and Sharpens — the Existing Evidence Base

The corpus already carries three adjacent findings that this study reframes rather than replaces:

• BCG’s 10-20-70 framework treats people-and-process as 70% of AI value but groups all non-technology investment into one pillar. Deloitte’s data disaggregates the “people” 70% into three structural variables a CHRO can actually design around.
• McKinsey’s March 2025 State of AI (n=1,491, workflow redesign as #1 EBIT predictor out of 25 attributes tested) establishes that how work gets reorganized matters more than which model is deployed. Team composition is one lever inside that redesign — the lever that governs whether the redesigned workflow has the right mix of people to execute it.
• NBER Copilot RCT (n=7,137, May 2025) measures individual and firm-level effects and finds 0% change in meetings or documents completed despite 31% less email time for regular users. The Faros data showed +98% PRs with zero delivery improvement. Both results suggest the bottleneck is not the individual — it is the coordination layer. Deloitte’s team-structure finding offers a hypothesis for why: in teams that are small, narrow, and disconnected, AI produces more output upstream without changing the throughput of the pipeline it feeds.

Source Credibility

Credibility: MEDIUM-HIGH with vendor caveat. Deloitte Insights is an in-house research publication of Deloitte, which operates a consulting practice that benefits from workforce-redesign engagement recommendations. The conclusions of this study point toward exactly the kind of organization-design work Deloitte sells. Apply that lens.

Methodologically, the study is a self-reported cross-sectional survey — it describes associations, not causal relationships. A team that already uses AI well is likely larger, more diverse, and more connected for reasons independent of AI. The direction of causation cannot be determined from these data alone. These case studies are vendor-published and represent selected wins with no control group and no independent verification. Cross-reference against: METR RCT (experienced developers 19% slower), CMU study (40.7% code complexity increase), Atlan 200-deployment analysis (median +159.8% ROI requires workflow redesign first).

The sample is sound: n=1,394, weighted toward leaders and managers (53%) but with balanced team-member representation (47%), across multiple industries and organization sizes, collected in July 2025 — TIER 1 freshness, current-generation models.

Key Data Points

What This Means for Your Organization

Before approving the next AI pilot, audit the team it is landing in. The Deloitte data suggests three questions a CIO, COO, or CHRO should answer jointly before funding:

1. Is the team large enough to run experiments and share learning? Deloitte’s evidence points to a ~10-person floor for compounding AI-usage effects. A four-person team is not disqualified from AI adoption, but it should not be your proof point.
2. Does the team have cognitive diversity by design, or by accident? High-performing teams treat skill and experience variety as a deliberate hiring choice. That is not a soft cultural preference — it correlates with a 2x gain on the three outcomes boards are being asked to track.
3. Do people outside the team know what the team does, and vice versa? Cross-functional connection predicts a 30% advantage. The “AI pilot in the marketing ops team that nobody else knows about” is not a pilot; it is a rehearsal.

For mid-market companies specifically, these findings support a resource reallocation conversation. If Deloitte’s companion Tech Trends research is right that 93% of tech budget lands on technology and 7% on the people running it, a 10-20-70 target flips those ratios. You will not get to 10-20-70 in one budget cycle. But a 20-point shift at the margin — moving 20 cents of every new AI dollar from software licensing to team design, training, and cross-functional rotation — is a conversation a CFO can approve on a one-page memo.

A final caution: survey data describes the teams that captured AI value last year. It does not tell you whether expanding a team from four to ten people will reproduce that result in your organization, or whether the teams that were already going to succeed happen to be larger and more diverse. Before restructuring, run the smaller intervention first — add a cross-functional reviewer to the existing team, expand skill diversity through targeted training, and measure whether the usage and value indicators move.

If this raised questions specific to your organization — how to phase the rebalance, how to measure trust and connectedness rather than just usage, how to set up a team audit before the next AI approval cycle — brandon@brandonsneider.com is open for the conversation.

Sources

1. Brodzik, S. Cantrell; Kulkarni, N.; Mahto, S.; Kreit, S. “Bridging the AI value gap: Are team dynamics the missing link?” Deloitte Insights, Feb 27, 2026. https://www.deloitte.com/us/en/insights/topics/talent/ai-roi-and-team-structure.html. Primary source. Sample n=1,394 U.S. working professionals surveyed July 2025. Vendor caveat applies — Deloitte Consulting benefits commercially from workforce-redesign recommendations. TIER 1 freshness.

2. Deloitte “Tech Trends 2026” — companion research cited for the 93%/7% tech-funding-vs-people-funding split.

3. BCG “AI at Work 2025” (n=10,635) — 10-20-70 framework (10% algorithm / 20% tech / 70% people and process) providing the corpus baseline for why team design matters more than tool choice.

4. McKinsey “The state of AI: How organizations are rewiring to capture value” (Mar 12, 2025, n=1,491) — workflow redesign is the #1 EBIT predictor out of 25 tested attributes; the team composition layer inside that redesign.

5. NBER Working Paper 33795 (n=7,137, 6-month RCT, May 2025) — individual Copilot use reduces email time 31% but produces zero change in meetings or documents completed at the team level; independent RCT counterpoint on where AI value accrues without workflow redesign.

Brandon Sneider | brandon@brandonsneider.com April 2026