The Run/Change Budget Trap: McKinsey's 2026 Framework for CIOs Funding AI Without Defunding Operations

Executive Summary

• AI now consumes up to one-third of enterprise technology change budgets while simultaneously adding to run costs — creating a structural squeeze on CIOs that no amount of incremental budget increase resolves.
• McKinsey’s analysis of 17 global companies identifies four IT archetypes. Only one — the “deliberate modernizer” — consistently frees enough budget to fund AI at scale: run costs kept at least 20% below peer levels, 57% of application spending directed toward modernization and new capabilities.
• Top-performing companies (≥10% revenue and EBIT growth over three years) allocate 16% of total technology budgets to internal staff working on change — 1.5 to 4.0 times more than laggards. Nearly two-thirds have technology leaders “very involved” in enterprise strategy, versus 52% at other companies.
• The fix is not a larger budget. It is deciding what to stop running — making explicit choices about which legacy applications, platforms, and services to retire before approving any new AI workload.
• The analysis was conducted in partnership with Serviceware (n=17 companies, global) and reflects a practitioner-level budget-granularity that most survey-based research cannot reach.

The Problem: AI Compounds an Existing Budget Squeeze

Most technology organizations already operate with stretched budgets before AI enters the picture. The CIO’s annual challenge — funding change initiatives with whatever remains after keeping the lights on — predates generative AI by decades. AI makes it structurally worse in two directions:

AI eats change budget. The analysis finds AI consuming up to one-third of companies’ change budgets. That is budget that previously funded cloud migration, application modernization, data infrastructure, and the foundational work that makes AI work later. Organizations that skip that foundation to fund AI directly are borrowing against their own capacity to scale.

AI adds to run costs. Every AI deployment requires inference infrastructure, monitoring, fine-tuning pipelines, prompt management, integration maintenance, and security review. These are not one-time costs — they become permanent run obligations. A company that deploys ten AI tools in Year 1 and does not sunset any legacy applications has increased total technology spend without reducing run complexity.

The net effect: technology budgets are rising — half of respondents plan increases of more than 4% in 2026, and a quarter of top performers plan increases above 10% — but the increase lands disproportionately in run, not change.

Four Archetypes: Where the Budget Goes

McKinsey identifies four IT budget archetypes from the analysis of 17 companies:

The deliberate modernizer pattern is the target. These organizations:

• Keep run-based infrastructure costs at least 20% lower than comparable companies — the gap that funds change
• Assign 57% of application spending to modernization and new capability development
• Earmark at least one-third of total technology expenditure for change activities
• Allocate 16% of total technology budget to internal staff working on change — between 1.5× and 4.0× what laggards allocate to the same category

The heavy IT sustainer pattern is where most mid-market companies quietly sit. A decade of cloud migrations that never fully completed, ERP implementations that expanded rather than consolidated, and tool proliferation driven by business-unit autonomy have created run-cost obligations that most CIOs cannot retire without a deliberate decision process they have never run.

What Top Performers Do Differently

The top-performer cohort (≥10% revenue and EBIT growth over three years, averaged over the past three years) shows two consistent behaviors that distinguish them from the rest of the survey:

Technology leaders are in the room. Nearly two-thirds (64%) report that their technology leader is “very involved” in crafting enterprise strategy, compared with 52% of other companies. This is not ceremonial — it is the difference between a CIO who influences which workflows get redesigned versus one who implements decisions made by business units and then inherits the run costs.

They budget for the multiplier, not the tool. The 28% of top performers planning technology budget increases above 10% in 2026 are not funding AI licensing — they are funding the infrastructure, data foundations, and internal capacity that let AI deliver a multiplier. The 3% of other companies doing the same are funding tool subscriptions. The distinction is where the internal-staff-on-change allocation goes: top performers run that category at 1.5–4.0× the rate of others.

The Three Levers

McKinsey’s framework reduces to three executable decisions:

1. Decide what to retire from run. This is the prerequisite. Without explicit retirement of legacy applications, platforms, and services, every new AI workload compounds run costs rather than replacing them. The practical question is: what application portfolio review process do you run, at what cadence, with what criteria for sunset? Most mid-market companies have no standing process for this.

2. Direct every change dollar toward shared capability, not point solutions. Shared platforms, standardized services, and data and analytics foundations reduce future run costs. Point solutions — a new SaaS tool for one team, a custom integration for one workflow — add to run costs the moment they go live. The deliberate modernizer allocates 57% of application spending toward foundations that compound; the heavy IT sustainer spends the same dollars on tools that accrete.

3. Insource the change function. The 16% internal-staff-on-change benchmark is not about headcount reduction — it is about capability location. Organizations that outsource their change work to systems integrators and vendors own neither the architecture decisions nor the operational knowledge. When the AI initiative requires rapid iteration, those organizations are waiting for vendors; the deliberate modernizers are moving.

Key Data Points

Publication date: approximately late March / early April 2026 (exact date not published; LinkedIn post confirmed ~2 weeks prior to 2026-04-17). Tier 1 — current.

Methodological note: n=17 is small for a quantitative framework. These findings are directionally credible and practitioner-validated but should not be treated as statistically representative of the broader enterprise technology population. The four archetypes and benchmark percentages are descriptive of the studied cohort. Cross-reference against McKinsey Global Tech Agenda 2026 (n=632 C-level) and Deloitte AI Infrastructure Survey 2026 (n=515) for larger-sample validation of directional claims.

What This Means for Your Organization

The run/change budget trap is not an AI problem. It is a portfolio-management problem that AI exposes and amplifies. A CIO who cannot answer “which five legacy applications are we retiring this year to fund AI at scale” does not have a funding problem — they have a governance problem. The technology budget is already large enough in most organizations; the question is what it is funding that it should not be.

The deliberate modernizer benchmark — 16% of total technology budget allocated to internal staff working on change, 57% of application spending on modernization and new capabilities, run costs kept 20% below peer levels — is the most granular operational target for CIO budget governance published in 2026. The comparable targets from the McKinsey Global Tech Agenda 2026 (n=632) and Deloitte AI Infrastructure Survey 2026 (n=515) describe the what; this framework describes the how with specific allocation percentages a CFO can test.

For a mid-market company with a $10M total technology budget, the deliberate modernizer targets imply: $1.6M in internal staff on change (not vendors, not licenses — internal capacity), $3.3M or more in change vs. run allocation, and a standing process for retiring legacy applications that runs faster than the pace at which new AI workloads create run obligations. Most companies that walk through that math discover they are heavy IT sustainers funding strained transformation — and that the gap to deliberate modernizer is less about budget and more about the portfolio decisions they have been avoiding.

If this raised questions about your current run/change allocation or your AI business case model, the conversation is worth having — brandon@brandonsneider.com.

Sources

1. McKinsey & Company with Serviceware, “Recalibrating CIO Technology Budgets for the AI Era” (~April 2026, n=17 global companies, technology-leader survey). URL: https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/recalibrating-technology-budgets-for-the-ai-era. Credibility: MEDIUM — McKinsey has direct commercial interest in technology transformation engagements; n=17 is a practitioner-depth panel, not a statistically representative survey; Serviceware co-sponsor has commercial interest in technology financial management software. Framework and benchmark percentages are directionally credible and internally consistent. Apply McKinsey vendor caveat.

2. Cross-reference — McKinsey Global Tech Agenda 2026 (Reil-Jerenz et al., Feb 2026, n=632 C-level executives, 69 nations, 24 industries): top-performer CIO strategic involvement (64% vs. 52%) confirmed in both studies. URL: https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/mckinsey-global-tech-agenda-2026. Source: research/04-consulting-firms/mckinsey-global-tech-agenda-2026.md

3. Cross-reference — Deloitte AI Infrastructure Survey 2026 (n=515, $500M+ revenue, Nov–Dec 2025 fieldwork): AI-infrastructure budgets projected to triple by 2028; 61% expect 10B+ token/month consumption. Validates directional run-cost-escalation finding from McKinsey framework. Source: research/04-consulting-firms/deloitte-ai-infrastructure-survey-2026.md

4. Cross-reference — IBM IBV “Tech Debt Reckoning” 2026 (n=1,300, Q3 2025): 29% higher projected ROI when tech debt is priced into AI business cases; 18–29% of AI implementation budgets consumed by debt remediation. Validates the legacy-run-cost drag on AI investment thesis. Source: research/04-consulting-firms/ibm-ibv-tech-debt-reckoning-2026.md

Brandon Sneider | brandon@brandonsneider.com April 2026