Insource, Reskill, or Hire: McKinsey Names the Three Levers of the AI-First Technology Workforce — and Why Top Performers Pull All Three at Once

Executive Summary

• Top-performing technology organizations are not choosing between insourcing, reskilling, and targeted hiring — they are running all three simultaneously. McKinsey finds nearly half of top performers plan to increase insourcing of technology talent (vs. 37% of other companies) while simultaneously deepening reskilling programs and raising the bar on selective external hiring.
• The talent profile is shifting away from AI specialists and toward senior engineers, architects, product managers, and designers. Leading companies hire fewer technologists overall but are dramatically more selective. Adding AI headcount without changing what is selected for is the failure pattern.
• Change management — not technical complexity — is the top barrier to scaling agentic AI among top-performing technology organizations. ~24% of top performers name it vs. ~15% of peers. At scale, people architecture becomes the binding constraint, not infrastructure or model capability.
• This piece is the companion to McKinsey’s CIO budget framework (“Recalibrating Technology Budgets for the AI Era,” April 2026), which identifies insourcing the change function as one of three levers that separate deliberate modernizer CIOs from the rest. The workforce article anchors the human side of the same decision.
• Audience: CIO/CTO/CHRO at 200–2,000 person American company deciding in 2026 whether to insource engineering capability, reskill existing staff, or rely on systems integrators to build AI competency they may never own internally.

The Three-Lever Problem

Every technology workforce conversation in AI eventually arrives at the same fork: build internally, buy externally, or borrow through partners and integrators. McKinsey’s framing is more direct — the top performers are not choosing a fork. They are running insourcing, reskilling, and targeted hiring as simultaneous programs, calibrated against each other.

That simultaneity is the insight. Organizations that treat this as a sequential decision — “first reskill current staff, then decide if we need to hire, then consider insourcing” — arrive at the scale decision years late. The companies that are pulling ahead are resourcing all three tracks in parallel.

Insourcing: Nearly half of top performers plan to increase insourcing of technology talent versus 37% of other companies. This is not a general trend — it is a top-performer differentiator. The underlying logic is visible in the companion budget piece: for deliberate modernizer CIOs, 16% of the total technology budget flows to internal staff working on change (1.5 to 4.0 times the rate at lagging companies). Insourcing is not an HR preference; it is how you retain the institutional knowledge that makes AI investments compound.

Reskilling: The companion budget piece flags reskilling as a mechanism, not a program. The shift is from “send people to training” to “redesign what the roles do so that reskilling is the job.” McKinsey State of Organizations 2026 (n>10,000) finds that 75% of current roles need reshaping as AI embeds across workflows, and that the organizations sustaining top-quartile performance over time invest $5 in people for every $1 in technology. That ratio has not changed with AI — it has become more visible because technology spending is rising faster than people investment in most AI business cases.

Targeted hiring: Leading companies hire fewer technologists overall but are far more selective. The demand shift moves toward senior engineers, architects, product managers, and designers — and away from the AI-specialist title that is currently attracting the most job postings. This is a meaningful counter-signal: the title “AI Engineer” is being flooded; the harder-to-fill roles are the architects and PMs who can translate business context into system design. McKinsey Global Tech Agenda 2026 (n=632) finds 31% of surveyed companies name talent gap as the top agentic-AI blocker, ahead of integration complexity (29%) and data foundations (25%).

The Change Management Finding Is the One Most CIOs Underweight

~24% of top-performing technology organizations name change management as the core challenge to scaling agentic AI — versus ~15% of other companies. At first glance that looks like a modest gap. It is not.

This is a revealed-preference finding. Top performers are the organizations that are actually scaling agentic AI — they have already passed the technical deployment stage. The fact that change management is surfacing as the constraint in the cohort that has gotten furthest means it is not a deployment-phase problem. It is a scale-phase problem.

The pattern is consistent across the 2026 corpus:

• Deloitte State of AI in the Enterprise 2026 (n=3,235): AI skills gap is the #1 integration barrier; 84% of organizations have not redesigned jobs around AI.
• Accenture Pulse of Change 2026 (n=7,000): 86% of leaders say they are preparing their workforce — 24% have embedded continuous learning operationally. The 62-point gap is the change management failure in production numbers.
• KPMG Global AI Pulse Q1 2026 (n=2,110): Organizations investing in AI talent are 4x more likely to report meaningful AI value (77% vs. 20%). The mechanism is exactly change management: AI agents require human oversight, judgment, and course-correction — capabilities that require design, not just deployment.

The implication for a 300-person company: if change management is the constraint at the organizations that have deployed the most, it is almost certainly a constraint before any organization reaches that scale. Building it early is not overhead — it is the investment that determines whether the technology investment pays.

The Hiring Signal That Runs Counter to the Job Board

Most AI hiring programs are organized around AI-specific roles — machine learning engineers, prompt engineers, AI product managers. McKinsey’s finding points in a different direction: leading companies are increasing selectivity on senior engineers, architects, and designers while reducing total headcount volume.

What this means operationally:

A mid-market CIO adding three AI-titled roles against a budget that does not support the compensation is solving the wrong problem. The KPMG data shows 66% of AI leaders hire for AI-specific roles (vs. 53% of peers) — but the McKinsey signal is that the value comes from quality and selectivity, not from the AI title on the JD. The architect who has built production-grade distributed systems and can evaluate an agent’s failure modes is more valuable than the AI specialist who can describe transformer architectures but has never shipped under production SLAs.

The practical hiring sequence for a 200-1,000 person technology organization:

1. One technical architect who can evaluate AI systems against existing infrastructure, not just build in isolation. This is the role most mid-market companies are missing.
2. One AI-fluent product manager who can translate business requirements into system design without going through a developer as intermediary. PwC’s 2026 AI Performance Study (n=1,217) finds 2x workflow redesign likelihood among top performers — that redesign has to live somewhere in the org chart.
3. Reskilling for the remaining staff before adding more headcount. Pluralsight’s data puts internal upskilling at $5,770 per employee vs. $14,170 for an external hire. A 40-person engineering team with six months of structured reskilling is a better foundation than a 40-person team plus two new hires in AI-specific roles.

The Insourcing Calculus

The insourcing finding is the most strategically consequential piece of the McKinsey framework. It answers the question every CIO faces when a systems integrator pitches an AI transformation engagement: should the organization own this capability internally, or is it fine to buy it externally every time?

The top-performer answer is clear: insourcing is rising, not declining. Nearly half of top performers plan to increase it. The reason is embedded in the budget companion piece — organizations that outsource the change function permanently are also the organizations where AI spend lands in run rather than change, because there is no internal staff to direct it toward modernization. Outsourcing the AI capability is correlated with the budget-trap outcome.

What insourcing is not: it is not refusing to use integrators. McKinsey’s framework is additive. The top performers are increasing insourcing while also using partners selectively for data and platform work — Forrester’s AIQ data (n=1,500) finds 47% of high-adopter organizations use consulting partners for data and platforms (vs. 26% of low adopters). The pattern is: own the architecture and change management internally; borrow specialized execution.

Key Data Points

What This Means for Your Organization

The insource/reskill/hire decision is where AI strategy becomes concrete — and where most 200–2,000 person companies default to the most expensive, slowest path. The expensive path is hiring AI specialists at a salary premium the organization cannot sustain, outsourcing the design work to a systems integrator who leaves when the engagement ends, and running a reskilling program that is not connected to either. The result: AI spend rises, internal capability does not compound, and the next initiative starts from scratch.

The lower-cost, higher-return path is sequenced differently: one internal technical architect who owns the evaluation function, a structured reskilling program for 30–50% of existing staff before the next round of AI deployment, and selective partner use for data platform work — with the explicit contract that the partner trains internal staff while executing, not just delivers outputs.

The change management finding from McKinsey is the most actionable signal for organizations that have already started deploying. If ~24% of the organizations furthest along name it as their constraint, and your organization has not yet formally resourced the change management function, that gap is going to appear — it just has not appeared yet. A dedicated change management owner for AI, reporting to the CIO or CHRO, is not an overhead line item at this stage of deployment. It is the capacity that determines whether the technology investment reaches production use rates.

If the talent sourcing decision has surfaced specific questions about your organization’s current engineering team structure and the AI capability you are trying to build, that is a conversation worth having before the next budget cycle closes — brandon@brandonsneider.com.

Sources

1. McKinsey “Designing an End-to-End Technology Workforce for the AI-First Era” (~April 2026). McKinsey Technology practice. URL: https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/designing-an-end-to-end-technology-workforce-for-the-ai-first-era. Credibility: MEDIUM — McKinsey Technology practice has direct commercial interest in workforce transformation engagements; prescriptive consulting framework, not an independent RCT. Specific survey n not confirmed (article paywalled/blocked). Key findings captured from prior research session. Note: three insourcing/reskilling/change-management findings are directionally consistent with McKinsey Global Tech Agenda 2026 (n=632) and companion CIO budget article (n=17); treat as practitioner-level analysis, not statistically representative.

2. McKinsey “Recalibrating Technology Budgets for the AI Era” (with Serviceware, ~April 2026, n=17 global companies). URL: https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/recalibrating-technology-budgets-for-the-ai-era. Credibility: MEDIUM — small n=17; practitioner-depth panel; directionally consistent with larger McKinsey surveys. Full analysis: research/04-consulting-firms/mckinsey-cio-tech-budget-ai-era-2026.md.

3. McKinsey “Global Tech Agenda 2026” (Reil-Jerenz, Romanelli, Jogani, Catlin, Halawa, Himatsingka — Feb 2026, n=632 C-level, 69 nations, 24 industries). URL: https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/mckinsey-global-tech-agenda-2026. Credibility: HIGH (with McKinsey vendor caveat). Full analysis: research/04-consulting-firms/mckinsey-global-tech-agenda-2026.md.

4. McKinsey “The State of Organizations 2026: Three Tectonic Forces” (Maor, Krivkovich, Srinivasan, et al. — Feb 19, 2026, n>10,000, 15 countries, 16 industries). URL: https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-state-of-organizations. Credibility: HIGH (with McKinsey vendor caveat). Full analysis: research/04-consulting-firms/mckinsey-state-of-organizations-2026.md.

5. Deloitte “State of AI in the Enterprise 2026” (n=3,235, 24 countries, Aug–Sep 2025). Credibility: MEDIUM-HIGH. Full analysis: research/04-consulting-firms/deloitte-state-of-ai-enterprise-2026.md.

6. Accenture “Pulse of Change 2026” (n=7,000 — 3,650 C-suite + 3,350 workers, Nov–Dec 2025). Credibility: MEDIUM. Full analysis: research/04-consulting-firms/accenture-pulse-of-change-2026.md.

7. KPMG Global AI Pulse Q1 2026 (n=2,110, 20 markets, Feb–Mar 2026). Credibility: MEDIUM-HIGH. Full analysis: research/04-consulting-firms/kpmg-global-ai-pulse-tech-report-2026.md.

8. Forrester “Accelerate Your AI Voyage” / AIQ Framework (n=1,500, Apr 2026). Credibility: MEDIUM. Full analysis: research/04-consulting-firms/forrester-accelerate-ai-voyage-2026.md.

Brandon Sneider | brandon@brandonsneider.com April 2026