88% of companies use generative AI in at least one business function. Yet only 6% see a measurable impact on their financial results. This figure, from McKinsey's “The State of AI in 2025” report, captures the paradox shaping this year's market: adoption is widespread, but competitive advantage remains concentrated in very few hands.
This article examines what separates companies turning generative AI into a lasting strategic asset from those accumulating proofs of concept without ever reaching production. It provides figures, actionable frameworks, and concrete benchmarks to help you make your own investment decisions.
TL;DR — Generative AI creates a competitive advantage only when it is integrated into core business processes, backed by senior management, and deployed at scale. Companies limiting themselves to peripheral uses—writing, summarizing, monitoring—are losing ground to those redesigning their workflows from the ground up. The deciding factor is organizational execution, rather than technology.
The Great Divide: Universal Adoption, Concentrated Value
88% Adopt, 6% Transform
McKinsey's “The State of AI in 2025” report leaves little room for doubt. Almost nine out of ten companies report using AI in at least one business function. But the share reporting a real impact on EBIT—above 5%—falls to 6%. This ratio is far from incidental. It means the vast majority of AI deployments remain confined to experimentation or marginal gains.
Gartner also identified this gap, predicting that 30% of generative AI projects would be abandoned after the proof-of-concept phase before the end of 2025, mainly because of poor data quality, uncontrolled costs, and ill-defined business value. Recent data shows reality exceeding that projection: another survey found that 42% of companies abandoned most of their AI initiatives in 2025, versus 17% in 2024.
The Window of Advantage Is Closing Quickly
Gartner highlights a phenomenon rarely mentioned in enthusiastic commentary: the advantage associated with generative AI erodes faster than in previous innovation cycles. Generative AI capabilities become a baseline requirement for market offerings in less than 36 months. In other words, what differentiated a business in 2023 became a prerequisite in 2025.
For executives, the implication is straightforward: adopting generative AI does not itself guarantee a competitive advantage. That advantage comes from how quickly and deeply an organization integrates it into critical processes, before competitors reach the same level of maturity.
Inside the 6%: What Winning Companies Do
They Aim for Transformation, Not Efficiency
The McKinsey report identifies a decisive characteristic: AI high performers are three times more likely than other companies to say they intend to fundamentally transform their business, rather than simply optimize existing processes. This distinction matters enormously.
A company using generative AI to write marketing emails faster saves time. A company using it to rethink its entire product design process—from analyzing customer feedback to generating prototypes—moves into a different competitive category.
High performers are 2.8 times more likely to fundamentally redesign their workflows: 55%, versus 20% of other companies. This different approach produces tangible results: a majority report improved innovation, while nearly half see improvements in customer satisfaction and competitive differentiation.
They Deploy at Scale, Not in Silos
Nearly two-thirds of organizations have not yet begun scaling AI across the enterprise. High performers do precisely the opposite: they are at least three times more likely to scale AI agent use across most business functions.
Capgemini's report on harnessing AI's value and unlocking advantage at scale confirms this trend. Generative AI adoption rose from 6% in 2023 to 30% in 2025, and 93% of companies are exploring or implementing generative AI capabilities. But Capgemini emphasizes one point: only companies adopting a “platformization” approach—shared AI infrastructure supporting multiple use cases—achieve cost-effective deployment at scale.
They Involve Senior Management
High performers are three times more likely to have senior leaders demonstrating strong commitment to and direct ownership of AI initiatives. This is more than a cosmetic factor. Without executive sponsorship, AI projects remain technical initiatives, contested between teams, underfunded, and lacking the mandate needed to transform cross-functional processes.
McKinsey also finds high performers much more likely to have defined human-in-the-loop validation processes: 65%, versus 23% of other companies. Rigorous quality control does not slow things down; it enables scaling without losing the trust of teams and customers.
The Trap of Peripheral Uses
Most Companies Remain Focused on Support Functions
Research by Bpifrance Le Lab shows that 31% of French microbusinesses and SMEs use generative AI, twice the share a year earlier. But in 68% of cases, use is limited to content writing. Monitoring, marketing, and communications remain the main areas of application. Core functions—production, logistics, product design, advanced customer service—are rarely involved.
This superficial adoption creates an illusion of modernity. The company uses ChatGPT, has trained a few employees, and can tick the “AI” box in its communications. Yet this layer of AI remains disconnected from the processes that actually generate margin.
The Hidden Costs of Strategic Inaction
The trap is twofold. First, investments scattered across uses that do not reshape the business consume budget and attention without creating a competitive barrier. Second, while the company writes its LinkedIn posts with AI, competitors are using it to automate quotation workflows, personalize their offerings at scale, or shorten product development cycles.
Real-world data confirms this risk. An analysis of 200 AI deployments in France published on data.gouv.fr found a median first-year ROI of 159%. But that figure conceals considerable variation: projects targeting core business processes generate significantly higher returns than those confined to support functions.
| Criterion | Peripheral uses | Transformative uses |
|---|---|---|
| Target functions | Marketing, monitoring, writing | Production, sales, design, customer service |
| Impact on EBIT | < 1% | > 5% (McKinsey high performers) |
| Lifespan of the advantage | < 12 months (easy to replicate) | 2–3 years (embedded in processes) |
| Typical investment | €5,000–€20,000 | €50,000–€200,000 |
| Internal sponsorship | Project manager / marketing team | Executive management / CIO |
| Production conversion rate | Low (repeated POCs) | High (transformation objective) |
The French Gap: Encouraging Signs, Structural Weaknesses
France Ranks in the Global Top Five… for Individual Adoption
With adoption reaching 44% of the population, France ranks fifth worldwide for generative AI use, behind the United Arab Emirates (64%), Singapore (60.9%), Norway (46.4%), and Ireland (44.6%). According to INSEE, 43% of economically active French people report using generative AI professionally.
These figures are flattering. But they measure individual adoption—employees using ChatGPT or Copilot on their own initiative—rather than company-led strategic integration.
Structural Investment Remains Low
The contrast emerges when we examine investment. According to Bpifrance Le Lab, only 9% of French microbusinesses and SMEs invested in AI over the past three years, and barely 2% did so regularly. INSEE confirms that only 10% of French companies with more than ten employees used at least one AI technology in 2024, compared with a European average of 13% and 28% in Denmark.
The main obstacle is not financial: more than two-thirds of reluctant executives say they cannot identify a relevant use case for their specific business. The problem is strategic vision, rather than budget.
SMEs with 100+ Employees Are Widening the Gap
A divide by company size is emerging. 53% of SMEs with at least 100 employees use generative AI, compared with 29% of microbusinesses. This difference reflects both greater investment capacity and an organizational structure better suited to cross-functional projects.
For mid-sized SMEs, the risk is being stuck in the middle: too small to have an IT department dedicated to AI, too large to get by with ad hoc use. This is precisely the segment where an external technical partner can improve competitiveness.
Four Ways to Build a Lasting Competitive Advantage with AI
Lever 1: Integrate AI into the Value Chain, Not Alongside It
According to McKinsey's analysis of 63 use cases, 75% of generative AI's economic value is concentrated in four business functions: software development, marketing and sales, customer service, and product R&D. Companies gaining a competitive advantage prioritize these functions and set measurable goals.
The head of a manufacturing SME who uses AI to produce a press summary every Monday morning saves 15 minutes. The same executive using it to analyze production quality feedback in real time and adjust manufacturing parameters is building a competitive asset.
Lever 2: Redesign Workflows Before Deploying the Tool
The common instinct is to bolt an AI tool onto an existing process. High performers reverse the sequence: they redesign the workflow to make use of AI's capabilities, then choose or develop the appropriate tool.
This workflow-first approach explains why 55% of McKinsey's high performers fundamentally redesigned their processes, versus 20% of other companies. The gain is structural rather than incremental. A tender response process redesigned around AI does more than take 30% less time: it changes the nature of the response, its personalization, and how quickly it is delivered.
Lever 3: Build Shared AI Infrastructure
The “one AI tool per use case” model quickly reaches its limits in cost, maintenance, and consistency. Capgemini's report emphasizes a platformization approach: a shared AI infrastructure layer serving several use cases, with shared models, unified data pipelines, and common quality standards.
For an SME, this platform can be as simple as a custom application connected to a language model API, with structured prompts and a built-in feedback system. It does not carry the cost of a large corporation's infrastructure. It is a software architecture decision.
Lever 4: Invest in People and Change as Much as Technology
According to data compiled by Insight, hidden transformation costs represent 30–40% of an AI deployment's total cost. They include change management, training, AI Act compliance, and managing internal resistance.
Companies neglecting this investment see a recurring pattern: the tool is deployed, teams ignore it or use it superficially, and the project is declared “inconclusive” after six months. The problem was not AI, but the lack of support.
Checklist: Is your AI deployment targeting a competitive advantage?
- The project targets a core business function, rather than only support functions
- An executive committee member personally sponsors the project
- The workflows involved have been redesigned, rather than simply augmented
- A change management budget is planned (30–40% of the total)
- Success measures connect to business KPIs: margin, lead time, satisfaction
- A human-in-the-loop validation mechanism is in place
- Scaling is planned from the design stage
Generative AI ROI: Beyond the Flattering Averages
The Figures Being Quoted—and What They Conceal
The aggregate figures are appealing. The 2024 Microsoft-IDC study reports an average return of 3.7x the initial investment, reaching 10.3x for leading companies. The French benchmark covering 200 deployments reports a median ROI of 159%. Gartner reports average results among early adopters of 15.8% higher revenue, 15.2% cost savings, and 22.6% higher productivity.
These averages conceal a bimodal reality. On one side, the 6% of high performers capture most of the value. On the other, 80% of companies see no significant contribution to EBIT. Generative AI ROI follows a power-law distribution rather than a normal curve, with the winners taking most of the rewards.
Time Works Against Those Who Wait
The payback period depends directly on the project type. Simple, well-targeted use cases generate initial benefits in 6–12 months. Transformative projects—the ones creating a lasting advantage—take 18–24 months to deliver their full return.
For an executive hesitating to act, this timing creates a dilemma: every quarter of inaction gives competitors already underway another quarter to extend their lead. Models improve, teams develop expertise, and proprietary datasets grow. These intangible assets accumulate and are difficult to catch up with.
| Indicator | Overall market | High performers (6%) |
|---|---|---|
| Impact on EBIT | < 1% on average | > 5% |
| Median first-year ROI | 159% (French data) | Up to 10.3x (Microsoft-IDC) |
| Production deployment | ~58% of projects | > 85% of projects |
| Payback period | 6–12 months (simple use) | 12–24 months (transformation) |
| Share of budget for change | < 10% | 30–40% |
What Will Change in 2025–2026: Three Shifts to Anticipate
Shift 1: From Chatbots to Autonomous AI Agents
Gartner predicts that 40% of enterprise applications will include specialized AI agents by the end of 2026, versus less than 5% in 2025. Moving from conversational chatbots to autonomous agents capable of completing whole tasks—booking, ordering, analyzing, following up—will reshape productivity.
Companies that have not included this agentic dimension in their roadmap by mid-2026 will face a growing functionality gap compared with those that have. Gartner nevertheless warns that more than 40% of agentic AI projects will be canceled by 2027, underscoring that organizational maturity remains the limiting factor, rather than technology.
Shift 2: AI as Infrastructure, No Longer a Project
In 2025, European companies spent €33.8 billion on AI tools, an increase of 96% in one year. This level of investment marks a turning point: AI can no longer be managed as a series of one-off projects. It is becoming a permanent infrastructure layer, like cloud computing or an ERP.
For CIOs, this shift requires rethinking application architectures. Business applications developed today without native AI capabilities will be obsolete within 36 months. Custom development must incorporate AI from the design stage, rather than add it afterward.
Shift 3: Compliance as a Competitive Advantage
The gradual implementation of the European AI Act is turning regulatory compliance into a differentiator. Companies that document their AI uses, introduce bias audits, and guarantee processing traceability will be in a strong position relative to less rigorous competitors.
Capgemini reports that 71% of companies say they cannot fully trust autonomous AI agents for enterprise use. This distrust creates an opportunity for companies investing in AI governance: trust becomes a competitive asset.
Practical Guide: Where to Start for Your Role
You Lead an SME with 50–250 Employees
Your priority is to identify the business process whose AI transformation will have the greatest impact on your margin, rather than adopt AI for its own sake. Begin by auditing high-value processes—sales cycle, production, customer service—and target the one with the largest volume of repetitive tasks. Budget €30,000–€80,000 for an initial transformative project, including change management.
You Are a CIO or CTO
Your challenge is architecture. Avoid piling up disconnected AI tools. Design a platformization strategy—even a simple one—that lets you share models, connectors, and quality standards across use cases. Decide promptly between build, buy, and partner: Gartner puts the time to value for differentiating AI at one to two years. Every month spent hesitating is a month lost.
You Are a Founder or Product Manager
Generative AI can shorten your time to market. Companies that build AI into their product from the design stage, rather than as a secondary feature, capture market attention faster. An MVP with integrated AI capabilities—personalization, automation, predictive analytics—immediately stands out from a conventional product. The marginal cost is low if the architecture accounts for it from the beginning.
You Lead Digital Transformation
Your main challenge is internal adoption. Two-thirds of executives reluctant to adopt AI say they see no relevant use case for their business. Your role is to show them one, rather than talk about it. A successful POC on a visible process, saving measurable time for real teams, is worth a thousand PowerPoint presentations about “the AI revolution.”
FAQ
Does Generative AI Really Create a Lasting Competitive Advantage?
Yes, but only when integrated into core business processes and deployed at scale. According to McKinsey, only 6% of companies achieve an EBIT impact above 5%. The advantage comes from transforming workflows rather than adopting a tool. Gartner warns that generative AI capabilities become a market prerequisite in less than 36 months.
What Budget Should an SME Plan for Its First Transformative AI Project?
An AI deployment targeting a business process costs €30,000–€80,000 in an SME, including change management. Hidden costs—training, compliance, change management—represent 30–40% of the total budget. The observed median first-year ROI in France is 159% for well-targeted projects.
Why Do So Many Generative AI Projects Fail?
Gartner identifies three main causes: poor data quality, uncontrolled costs, and ill-defined business value. In 2025, 42% of companies abandoned most of their AI initiatives. The most frequent failure factor is a lack of executive sponsorship, leaving projects stuck at the experimental stage.
Should You Develop Your Own AI Tools or Use Market Solutions?
It depends on your objective. Commercial SaaS solutions such as Copilot and Gemini suit peripheral uses like writing and summarizing. For a differentiated competitive advantage, custom development remains necessary: it integrates AI into your specific processes using proprietary data and creates a barrier to entry competitors cannot replicate by taking out the same subscription.
Are French Microbusinesses and SMEs Behind on AI?
Individual adoption is strong: France ranks fifth worldwide, with 44% of the population using generative AI. But structural investment remains low: only 9% of microbusinesses and SMEs invested in AI over three years. The real gap is strategic rather than technological. Two-thirds of reluctant executives see no relevant use case for their business.
When Is the Right Time to Launch an AI Project in My Company?
The right time was six months ago. Gartner says differentiating AI takes one to two years to deliver value. Each quarter of waiting widens the gap with competitors already underway, accumulating proprietary data, optimized workflows, and trained teams—all cumulative assets that are difficult to catch up with.
AI Coder Squad: Turning Generative AI into a Concrete Competitive Advantage
Moving from experimentation to a transformative AI deployment requires software architecture designed from the outset for model integration, data management, and scaling. This calls for senior developers who understand both business constraints and AI's actual capabilities.
AI Coder Squad designs custom applications and AI agents for companies that want to move quickly without sacrificing quality—with senior developers and an AI-powered approach.
→ Start your project and discover how AI Coder Squad can accelerate your next delivery.