The most useful data point in the 2026 AI-strategy literature is not a model benchmark. It is the **IBM 2026 CEO Study**, conducted with **Oxford Economics** on a sample of **2,000 chief executives**, which found that **only approximately 25% of artificial-intelligence initiatives are delivering the expected return on investment**, that **16% have been scaled across the enterprise**, that **61% of CEOs are actively adopting AI agents**, and that **only approximately 29% of CEOs say they can measure AI ROI with confidence**. The same study reports that organizations that **integrate control and governance directly into their AI systems** experience approximately **25% fewer incidents** than those that do not.
This article is written for the CTO, CIO or board member who is reading those numbers and trying to reconcile them with the market context: **Gartner** anticipates that approximately **40% of enterprise applications will embed AI agents by the end of 2026**, and credible market-sizing places **AI-agent software spend at approximately 206.5 billion US dollars in 2026, up roughly 139% versus 2025**. The contradiction is real: enterprise AI spend is accelerating *while* the share of AI initiatives that demonstrably pay back is stuck around one quarter.
Our position at Call IT Dev — and the operational point of this analysis — is that the ROI gap is **not** a technology problem. The frontier models are good enough; the agentic frameworks are mature enough; the cost of inference has fallen materially. The gap is a *culture, governance, workflow-design and data-strategy* problem, and the buyers that close it most reliably are not building everything in-house. They are buying delivery discipline. We discuss the geographic and economic context that makes that delivery affordable — institutional Morocco in 2026 — in our companion piece on <a href="/en/blog/casablanca-tech-valley-morocco-nearshore-hub-buyer-guide-2026">Casablanca Tech Valley and what the new nearshore hub means for the 2026 outsourcing decision</a>.
A close reading of the IBM 2026 CEO Study, the Gartner agentic-AI forecasts and the public after-action reports from large 2025–2026 deployments surfaces a remarkably consistent pattern. The pilots that fail to scale rarely fail on the model; they fail upstream and downstream of it.
The diagnosis matters because it changes the build-vs-buy calculus. If the ROI gap were a model gap, the answer would be to buy a better model. Because the ROI gap is an orchestration, governance, data and workflow gap, the answer for most mid-market and upper-mid-market buyers is to buy disciplined delivery — and to keep the in-house team focused on the strategy, the data assets and the change management that no external partner can do for them.
The decision is not a single switch. It is a question per layer of the AI stack. The framework below is what we use with prospective buyers in scoping conversations; it is deliberately operational, not philosophical.
Read together, the framework produces a profile that is **neither pure-build nor pure-buy**. It is **build the strategic layers, buy the delivery discipline, partner on the heavy lifting**. The buyers who score the IBM 2026 CEO Study's 25% ROI-delivery rate are, almost without exception, operating this hybrid. The 75% who do not are typically over-building one layer (often the model or the data) and under-investing in another (often governance or change management).
The market is now saturated with AI consultancies. The mid-market buyer reading vendor decks in 2026 has reasonable difficulty separating disciplined delivery from rebranded staff augmentation. The five behaviors that distinguish a delivery partner that closes the ROI gap from one that does not:
A buyer who can ask a prospective partner these five questions and receive concrete artifacts in response — the baseline ROI sheet, the audit-log demo, the workflow design template, the data-quality gate, the senior bench — has a partner whose operating model is already calibrated to the 2026 problem. A buyer who receives a deck instead has a marketing response.
The 2026 AI-agent software spend forecast of ~206.5 billion US dollars (+139% versus 2025) is a number that mid-market buyers cannot meet at Western European or US East Coast day rates without sacrificing one of the layers of the framework above. Either the data layer gets cut, or the governance layer gets cut, or the change management gets cut. In each case, the cut is precisely the layer the IBM 2026 CEO Study identifies as the determinant of ROI.
Nearshore Morocco — CET-aligned, multilingual (English, French, Spanish, Arabic, with growing German and Italian), with a senior orchestration bench that has shipped against EU clients since 2022, on a labor cost basis approximately 60% below Southern Europe — is the geography that lets a buyer afford the *whole* framework. The partner pod handles the data consolidation, the workflow design, the governance build and the production delivery on a unit cost that absorbs the AI-agent software spend without forcing a layer cut. The in-house team retains the strategy, the policy and the change management. The ROI gap closes because no layer is sacrificed.
For buyers scoping the next AI delivery engagement, the practical entry points are our <a href="/en/services/ai-automation">AI automation</a> practice for agent and workflow build, our <a href="/en/services/software-development">software development</a> practice for the data and integration layer, our <a href="/en/services/dedicated-development-teams">dedicated development teams</a> for the long-horizon senior-weighted pod model, and our <a href="/en/why-morocco">Why Morocco</a> overview for the geographic context.
Four actions that are low-cost, independent of any vendor decision, and that materially improve the ROI posture of the next AI investment.
The **IBM 2026 CEO Study** — 2,000 CEOs interviewed with **Oxford Economics** — is the clearest data point in the 2026 literature: **~25% of AI initiatives deliver expected ROI, 16% have scaled enterprise-wide, 61% of CEOs are actively adopting AI agents, only ~29% can measure AI ROI with confidence**, and organizations that embed governance into AI systems experience **~25% fewer incidents**. The market is moving regardless: **~206.5 billion USD of AI-agent software spend in 2026 (+139% YoY)**, and per **Gartner**, **~40% of enterprise applications with embedded AI agents by end-2026**.
The buyers who close the ROI gap in 2026 are not the buyers with the best model. They are the buyers who build the strategic layers in-house — strategy, policy, change management — and buy disciplined delivery on the heavy-lifting layers — data, workflow design, governance build, agent build — from a partner whose operating model already absorbs the IBM Study's findings into the delivery shape. The geography that makes the full framework affordable at mid-market scale is institutional nearshore Morocco. The decision the CTO has to defend to the board in 2026 is not *whether* to invest in AI. It is *which layers to own* and *which to partner on*. The 25%-ROI cohort is built on getting that decision right.
${CTA_BUILD_VS_BUY}
**Sources:** IBM 2026 CEO Study, conducted with Oxford Economics (2,000 chief executives surveyed; ~25% ROI delivery, 16% enterprise scale, 61% active AI-agent adoption, ~29% confident ROI measurement, ~25% fewer incidents with embedded governance); Gartner public statements on AI-agent enterprise application embedding (~40% by end-2026); industry market-sizing of AI-agent software spend (~206.5 billion USD in 2026, +139% YoY versus 2025). This article is informational, summarizes publicly reported research findings, and is not investment, legal or procurement advice.
The IBM 2026 CEO Study, conducted with Oxford Economics on a sample of 2,000 chief executives, found that approximately 25% of AI initiatives are delivering the expected ROI, 16% have been scaled enterprise-wide, 61% of CEOs are actively adopting AI agents, and only approximately 29% of CEOs say they can measure AI ROI with confidence. Organizations that embed control and governance directly into their AI systems experience approximately 25% fewer incidents.
A culture, governance, workflow-design and data-strategy problem. The frontier models are good enough; the agentic frameworks are mature enough. Pilots stall on missing upstream context capture and downstream system-of-record integration, on data-quality debt, on governance bolted on at the end rather than embedded in the workflow, on cultural adoption that runs the AI in parallel with the legacy manual path rather than replacing it, and on in-house scarcity at the orchestration layer.
Build strategy, policy, change management, the high-level workflow-design authority, and the data assets themselves. Buy the model layer (multi-vendor with an abstraction so it is replaceable), the agentic framework, the heavy lifting on data consolidation, the governance build-out (audit logs, evaluation harness, AI-BOM, KEV monitoring), and the production agent delivery. The hybrid is the operating model behind the 25% who actually show ROI.
AI-agent software spend is estimated at approximately 206.5 billion US dollars in 2026, up roughly 139% versus 2025. Gartner anticipates that approximately 40% of enterprise applications will embed AI agents by end-2026. The market scale is the reason the in-house-talent shortage at the orchestration layer is structural rather than transitional, and the reason mid-market buyers cannot fund the full delivery stack at Western European day rates.
ROI measurement wired from week one with a baseline before any code ships; governance embedded in the workflow (per-agent audit log, per-decision evaluation harness, per-feature AI-BOM, machine-readable provenance, KEV monitoring); workflow design produced in the first sprint and committed to the repository; a documented data-quality gate that the partner refuses to ship past; and a senior-weighted, stable pod composition with named tech, product and operations leads from kickoff to handover.
Because the layers the IBM Study identifies as decisive — data consolidation, workflow design, governance build, production agent delivery — are precisely the layers a mid-market buyer cannot afford at Western European day rates without cutting one of them. Nearshore Morocco (CET-aligned, multilingual, senior orchestration bench with EU delivery history since 2022, ~60% below Southern European costs) lets the buyer fund the whole framework, keep change management in-house, and retain the ROI rather than financing the gap.
CALL IT DEV — Software, AI and dedicated tech teams — Casablanca | Madrid | Dubai — contact@callitdev.com — +212-537-373777