The 2026 State of AI in the Enterprise report — based on interviews with over 3,000 senior leaders across 24 countries and six industries — finds a market divided into thirds. Each third is capturing productivity gains. Only one third is building genuine competitive advantage. The distance between those thirds is widening.

The Deloitte report's most useful contribution is not a single headline number but a taxonomy. The 3,235 executives surveyed divide almost evenly into three groups based on how they are using AI:

34% — Reimagining: Creating new products, reinventing core processes, or fundamentally changing their business model.

30% — Redesigning: Redesigning key processes around AI without altering the broader structure of the business.

37% — Surface: Layering AI onto existing systems with little or no change to existing processes.

All three groups are capturing productivity and efficiency gains — the technology works. But only the first group is building competitive advantage that compounds. The efficiency gains in the surface group are real but non-durable: as the tools become commodity, the productivity gains equalize across the market.

Worker access to AI tools expanded 50% in one year — from under 40% to around 60% of the workforce having sanctioned access. But fewer than 60% of workers with access use AI in their daily workflow — a figure that has not changed year over year. The adoption access problem has been largely solved; the adoption usage problem has not.

The agentic AI pipeline figure is striking: nearly three-quarters of companies plan to deploy autonomous AI agents within two years. Yet only 21% report having a mature model for agent governance. The same governance gap Monte Carlo documented at the practitioner level is confirmed at the C-suite level.

Talent readiness is the report's starkest finding. Only 20% of organizations say their talent is highly prepared for AI — declining from prior years because the capability requirement is rising faster than training programs can respond. The number-one way companies have adjusted their talent strategies is education, not workflow redesign. Employees are being taught how to use tools without anyone reworking how the work actually gets done.

The revenue data is the most forward-looking signal. 74% of organizations hope to grow revenue through AI in the future. Only 20% are already doing so.

Sources: Deloitte 2026 State of AI press release / Techstrong.ai analysis / BigDATAwire, March 2026

the share of enterprises using AI to truly reimagine their business — versus 37% still at the surface.

The 34% figure from Deloitte's 3,235-executive survey is the clearest benchmark in the enterprise AI landscape. It puts a precise number on the transformation gap that PwC's 74/20 finding described from a financial returns perspective in this series' opening edition. PwC found that 20% of organizations capture 74% of AI's economic value. Deloitte finds 34% are reimagining their business. These numbers are not in contradiction — they describe the same market from different angles. The organizations "reimagining" in Deloitte's taxonomy are the population that "captures most of the value" in PwC's. The remaining 66% are splitting the remaining 26% of economic returns. The transformation gap has a name, a number, and now a survey of 3,235 executives confirming it is structural rather than temporary.

Microsoft published its Q1 2026 Global AI Diffusion Report on May 7, measuring AI usage for 17.8% of the global working-age population — up from 16.3% in Q4 2025. The UAE leads the National AI Leaderboard at 70.1%. The US ranks 21st at 31.3%. The global North-South gap has widened: 27.5% in the North, 15.4% in the South. Key developer employment finding: total US software developer employment hit 2.2 million in 2025, rising 8.5% YoY — a record high. Q1 2026 data shows developer employment still 4% above March 2025. This challenges the dominant displacement narrative and points to a compositional story: AI is creating new developer roles faster than it is eliminating traditional ones in aggregate.

Source: Microsoft On the Issues, May 7, 2026

Nvidia has committed more than $40 billion to AI equity investments in the first five months of 2026, led by a $30 billion stake in OpenAI. This week alone: rights to invest up to $3.2 billion in Corning (building three US facilities for Nvidia's optical technologies) and up to $2.1 billion in IREN (deploying up to 5 GW of Nvidia's DSX-branded infrastructure). Across Marvell, Lumentum, Coherent, Nebius, CoreWeave, and roughly two dozen private rounds, the pattern is consistent: capital flows to companies that buy Nvidia GPUs at scale and re-rent them to hyperscalers and frontier AI builders. Nvidia CFO Colette Kress described the strategy as ensuring compute capacity is built around Nvidia hardware. Non-marketable equity securities grew from $3.39B to $22.25B in a single year.

Sources: CNBC, May 9, 2026 / The Next Web

CNBC's May 10 analysis of Alphabet's 160% stock rally over the past twelve months frames it as market validation of the "owns most of the stack" AI thesis. Alphabet's position spans TPU hardware, Gemini models across all tiers, Google Cloud infrastructure, Google Workspace (3 billion users), DeepMind, and a $40 billion investment in Anthropic. The market appears to be pricing the "owns most of the stack" position at a premium — a useful frame for enterprise procurement teams evaluating long-term cloud and AI platform commitments.

Source: CNBC, May 10, 2026

JPMorgan Chase formally reclassified its AI investments from experimental R&D to core infrastructure, with a 2026 technology budget of approximately $19.8 billion and 2,000 staff dedicated to AI development. The reclassification changes how the investment is governed, reported, and evaluated: core infrastructure budgets are evaluated on operational reliability, cost-per-transaction, and SLA performance — not on potential. When the largest US bank treats AI as infrastructure rather than R&D, it is signaling that AI reliability, auditability, and governance have crossed the threshold required for inclusion in regulated, mission-critical financial systems.

Source: Crescendo AI digest / JPMorgan filings, May 2026

The dominant narrative around AI and software developers is that AI coding tools are eliminating entry-level developer jobs. The Stanford AI Index data supports this for specific categories. And the 65% AI-generated code figures from Snap and Google suggest a genuine structural change.

But the Microsoft Global AI Diffusion Report's employment data complicates the picture: total US software developer employment reached 2.2 million in 2025 — an 8.5% year-over-year increase and a record high. Early Q1 2026 data shows developer employment still 4% above March 2025 levels.

The resolution of this apparent contradiction is likely compositional. AI is creating new developer roles — AI application developers, prompt engineers, AI systems integrators, model fine-tuning specialists, AI governance engineers — faster than it is eliminating traditional software development roles in the aggregate. For enterprise technology leaders planning workforce strategy, the compositional distinction matters: "train your existing developers on AI" is a different strategy than "hire AI-native developers," and the employment data suggests both will be needed simultaneously.

Source: Microsoft Global AI Diffusion Report, May 7, 2026 / Stanford AI Index 2026

This series opened on April 16 with PwC's finding that 74% of AI's economic gains flow to 20% of organizations. Four weeks later, Deloitte's 3,235-executive survey produces a complementary finding from a different angle: only 34% of organizations are genuinely reimagining their business with AI, while 37% are still layering it on the surface with no meaningful workflow change. The shape of the distribution is identical across two independent methodologies. The enterprise AI market is sorting into a small group of organizations building compounding advantage and a large group capturing efficiency gains that will commoditize as the tools become universal.

The Deloitte report's talent finding is the one that deserves the most attention. Only 20% of organizations say their talent is highly prepared for AI — and that figure has declined from prior years. Organizations have responded to the AI capability requirement with training rather than redesign. The result is AI-literate employees embedded in AI-agnostic processes — which is precisely the pattern that produces the 37% "surface-level" adoption cohort.

The organizations in the 34% are making a different set of choices. They are not primarily investing in AI training programs. They are investing in workflow redesign: identifying specific processes where AI can own end-to-end execution, defining the human oversight structure around those processes, and building the governance infrastructure that makes autonomous AI trustworthy. The ServiceNow AI Specialists resolving 99% of IT cases faster. The Travelers Insurance AI Claim Assistant handling 100,000 calls. The Cisco Codex deployment saving 1,500 engineering hours per month. These are process redesign wins that happened to involve AI.

The Nvidia $40B ecosystem investment story adds a structural dimension. Nvidia is betting that the physical infrastructure layer of AI will generate returns that compound as long as the industry's hardware dependency on its GPUs continues. If Nvidia is right, the organizations in the 34% will be building transformational capabilities on top of infrastructure that compounds their advantage. The 66% will be paying for the same infrastructure to generate efficiency gains that equalize over time.

The question each edition of this series has been circling is the same question the Deloitte data poses directly: what does it take to move from the 66% to the 34%? The data provides a clear answer. It is not model selection. It is not tool access — 60% of workers have sanctioned AI access and fewer than 60% of those use it daily. It is not training. It is process redesign with governance built in. That is the work that is hard to commission, hard to measure, and hard to maintain political will for during a period when efficiency gains are being celebrated as sufficient. It is also the only work that produces compounding competitive advantage.

AK / Spearhead / Building AI systems, not tools