AI Budgets Shift 93% to Tech, 7% to People

Corporate allocations to artificial intelligence are sharply skewed toward hardware and software: firms now devote roughly 93% of AI budgets to technology and just 7% to people and change management, according to reporting in Fortune on March 29, 2026 (Fortune, Mar 29, 2026). The distribution marks a decisive pivot from broader digital-transformation practice and is already generating measurable friction in deployment timelines and user adoption across sectors. Executives and boards are reacting to headline-grabbing model capabilities and cloud economics, but the underinvestment in reskilling, design and governance is producing a counterintuitive productivity drag in early adopter firms. For institutional investors tracking operational leverage from AI, the reallocation presents both a near-term risk to projected cost savings and a potential medium-term opportunity for businesses that rebalance spending toward human capital and integration. This article unpacks the data, highlights sector-level implications, and sets out Fazen Capital’s perspective on how capital allocators should interpret this pattern of expenditure.

Context

The 93%/7% split reported by Fortune (Mar 29, 2026) synthesizes survey results and vendor spend tallies collected from several large enterprises and consultancy studies (Fortune; research cited from Deloitte, Wharton, Harvard). That split contrasts with historical IT and transformation projects where implementation budgets commonly earmarked a materially larger share for training, change management and user experience — typically in the high single-digits to low double-digits of total program spend. The deviation toward technology reflects two simultaneous pressures: the accelerating performance improvements of foundational models that demand expensive compute, and an investor-driven focus on headline metrics (models shipped, teraflops procured) rather than adoption curves or employee retraining metrics.

This reallocation is happening against a broader labor-market and automation backdrop. The World Economic Forum’s 2020 Future of Jobs report estimated that 85 million roles could be displaced by technological change while 97 million new roles could emerge by 2025 (World Economic Forum, 2020). Separately, the OECD’s 2019 analysis placed approximately 14% of jobs at high risk of automation (OECD, 2019). Those macro figures help explain why corporate leadership is aggressive in acquiring technology: there is fear of being competitively outpaced. However, the macro shift does not negate the operational reality that hardware and model capacity deliver no business value without human processes to integrate outputs into decisions, sales, and regulated workflows.

The Fortune report also links the spending pattern to a rise in measurable implementation setbacks in the first half of 2026: longer-than-expected time to value, elevated error rates in production workflows, and user pushback in customer-facing applications (Fortune, Mar 29, 2026). These are early indicators rather than settled trends, but they are consistent with historical cases where underinvesting in the organizational side of technology has delayed ROI by 6–24 months across sectors.

Data Deep Dive

The single most striking data point is the 93% allocation to tech versus 7% to people. This ratio emerged from aggregated survey and spending data reported by Fortune on March 29, 2026 (Fortune, Mar 29, 2026), and was corroborated by interviews with chief digital officers and procurement leads cited in the piece. Layering that with sector-specific metrics reveals heterogeneity: large financial institutions and cloud-native software companies skew even more heavily to infrastructure, while consumer-facing retail and healthcare firms—where user adoption is critical—show a marginally higher relative spend on training (roughly 10–12% in those subsectors, per firm-level disclosures referenced in Fortune).

A practical way to understand the effect is to compare this allocation with typical project-performance benchmarks. Historically, program-management guidance from major consultancies has suggested allocating 10–30% of transformation budgets to people, process redesign and upskilling to secure adoption and risk controls. The current 7% average therefore sits at or below the bottom of that historical band, implying companies are compressing training and change budgets to prioritize compute and models. That compression can materially extend the time-to-value curve: empirical studies of enterprise software rollouts have shown implementation timelines can lengthen by 25–50% when stakeholder training is insufficient.

From a cost perspective, the immediate line items driving the 93% number are clear: large-scale cloud credits for model training, bespoke model licensing, GPU clusters and data engineering. For example, public filings and vendor reports over the past two years showed enterprises committing multi-year cloud agreements worth tens of millions of dollars; Fortune’s reporting notes multiple Fortune 500 firms that committed seven-figure monthly cloud expenditures in 2025–26 as part of their AI pushes (Fortune, Mar 29, 2026). The result is a front-loaded capex and opex profile for technology with deferred human-capital investment, a combination that magnifies short-term headline output while undermining durable adoption.

Sector Implications

Financial services. Banks and asset managers have been early adopters of foundation models for trading signals, compliance automation and client-service chat. The high tech-share allocation can accelerate model deployment for analytics and algo strategies, but inadequate investment in governance and human oversight increases model risk and potential regulatory scrutiny. Given that regulators in major jurisdictions issued updated AI guidance in 2024–2025 emphasizing human-in-the-loop controls, firms that underinvest in people risk noncompliance costs and remediation expenses that may erode expected savings.

Healthcare. Clinical settings require clinician buy-in and validated workflows. There, the 7% allocation is particularly problematic: integration into EMRs, clinician training, and outcomes validation demand substantial human investment. Clinical deployment delays translate directly into slower patient benefit and revenue realization. Conversely, health systems that commit a larger share of resources to training and clinical validation are reporting faster uptake and earlier realized improvements in throughput and diagnostic accuracy, per interviews cited in sector briefings.

Technology and cloud providers. The vendors selling compute and models are benefiting from the allocation pattern: higher immediate revenue and tighter vendor lock-in. Yet this creates counterparty risk for enterprise customers if the resulting systems fail to deliver expected process improvements. From an investor perspective, vendor revenue growth may look robust even as end-customer economics deteriorate in the absence of adequate implementation spending.

Risk Assessment

Operational risk. Underinvesting in people and process increases the probability of implementation failure, data-quality issues, and model drift. Those operational failures can lead to direct costs (remediation, fines) and indirect costs (lost customer trust). The Fortune reporting from March 29, 2026 highlights examples of early adopters seeing elevated error rates in production workflows and longer time-to-value (Fortune, Mar 29, 2026), signifying that the risk is material and contemporaneous.

Regulatory and compliance risk. Several regulators published AI supervisory expectations between 2024 and 2026 that emphasize governance, explainability and human oversight. Institutions allocating only 7% to people are at risk of failing to meet these supervisory expectations, which can result in enforcement actions or requirements for costly post-hoc controls. For regulated industries, this mismatch can translate into balance-sheet or reputational damage that is not captured in short-term P&L metrics.

Strategic risk. The imbalance creates a potential winner-take-most dynamic: firms that properly rebalance spending toward integration and human capital may capture outsized returns as their peers struggle to extract value. Conversely, firms that double down on tech without integration may achieve scale in raw capability but fail to monetize it, creating mispriced competitive advantages and idiosyncratic downside risk for investors.

Outlook

Over the next 12–24 months, we expect to see three measurable shifts. First, companies that experience adoption failure or regulatory pushback will increase people-and-process spend, closing part of the 93/7 gap. Second, consulting and training markets should grow: providers that specialize in reskilling and change management for AI deployment are likely to see elevated demand and higher pricing power. Third, investor scrutiny will increase: public companies will be asked to disclose not just model metrics but also deployment metrics (time to first revenue, user adoption rates, and educational hours per employee), and these disclosures will alter valuations where adoption lags.

The timing of this reallocation will be uneven across regions and sectors. Firms operating under stricter regulatory regimes or with mission-critical client relationships (healthcare, financial services) are more likely to rebalance sooner. By contrast, e-commerce and ad-tech firms that can continuously iterate in production may continue to prioritize tech spend. From a market perspective, that divergence will create cross-sectional dispersion in earnings upgrades and cost-savings trajectories through 2027.

For investors and corporate boards, the practical implication is to ask for granular metrics: explicit caps on infrastructure spend relative to integration spend, targets for employee training hours, and measurable KPIs tied to adoption. Transparency in those metrics will drive better comparisons across peers and reduce the risk of overly optimistic projections about near-term margin expansion.

Fazen Capital Perspective

Fazen Capital views the 93%/7% split not as the endpoint but as a predictable phase in a broader technology adoption lifecycle. The initial sprint to acquire compute and capability is rational from a competitive standpoint, but it creates a second-order arbitrage: the firms that pivot early to build human-centered integration and governance will convert capability into durable economic advantage. We favor monitoring leading indicators—training hours per user, percentage of models in production with documented human-in-the-loop controls, and time-to-adoption metrics—rather than relying solely on gross AI spending growth. For institutional investors, the contrarian insight is that the companies with the most disciplined governance and higher relative spend on people are more likely to deliver sustainable ROI than those that report the largest near-term AI spending increases. See our work on enterprise AI and integration frameworks for further context.

FAQ

Q: What practical metrics should investors demand to assess AI program health?

A: Beyond headline spend, demand disclosure of (1) training hours per affected employee per quarter, (2) percentage of AI projects reaching production within target timelines, (3) number of models with documented human-in-the-loop checks, and (4) remediation incidents tied to model failures. These operational metrics are leading indicators of adoption and regulatory compliance and are not uniformly reported today.

Q: How does the current allocation compare with past technology waves, like ERP or cloud migrations?

A: Historically, ERP rollouts and major cloud migrations showed similar patterns: front-loaded technology spending followed by a recognition of the need for change programs. In many ERP projects across the 2000s and 2010s, underinvestment in training lengthened realization timelines by up to a year. The difference with AI is velocity and opacity; models can change behavior post-deployment, making ongoing human oversight a recurrent cost rather than a one-time training expenditure. Institutional investors should therefore expect a longer tail of people-related spend for AI than for previous major IT projects.

Bottom Line

The 93% tech / 7% people allocation documented by Fortune (Mar 29, 2026) is a meaningful early-warning signal that many AI programs may not achieve their projected near-term benefits without rebalancing. Firms and investors should track integration and human-capital metrics as leading indicators of value capture.

Disclaimer: This article is for informational purposes only and does not constitute investment advice.