Four out of five companies use generative AI but see no effect on their results. This is not a technical problem. It is a method problem.

A synthesis of the latest studies from Deloitte, McKinsey, and PwC reveals a spectacular gap: 78% of American and European companies deploy generative AI in at least one function, but 80% observe no material impact on their financial performance. While organizations get bogged down in marginal gains, the rare companies that redesign their processes around autonomous agents record productivity gains 2 to 10 times higher.

The Essentials

  • 78% of companies use generative AI, but 80% see no effect on their financial results
  • Companies that redesign their processes around autonomous agents achieve gains of 2 to 10x compared to 20-40% for the assistant approach
  • Middle management could decline significantly by the end of 2026 according to early feedback
  • The gap widens between companies that automate and those that merely assist

AI Assistants Peak at 20% Gains

The dominant approach consists of grafting AI assistants onto existing workflows. A salesperson uses ChatGPT to draft emails. A lawyer has Claude summarize contracts. An accountant automates calculations with Copilot. Result: productivity gains of 20 to 40% on specific tasks, but no transformation of overall performance.

This logic of assistance reproduces the pattern of all previous technological revolutions. The first electric factories reproduced the organization of steam-powered workshops. The first websites copied paper brochures. The computerization of the 1980s initially mechanized bureaucracy without rethinking it.

Companies install AI the way they installed Excel in the 1990s: tool by tool, function by function, without questioning the overall architecture. A European bank surveyed by Deloitte deployed an increasing number of different AI tools over a short period. Measured productivity: marginal improvements across all operations.

Autonomous Agents Redesign the Organization

Conversely, a minority of companies is experimenting with “agents”: AIs capable of conducting complete end-to-end processes. Not assistants that help a human, but systems that take charge of entire workflows.

An American insurance company entrusted an agent with the complete processing of simple automobile claims. The agent analyzes damage photos, consults technical databases, calculates compensation, contacts repair shops and validates payments. Processing time considerably reduced. Costs significantly decreased. Administrative errors virtually eliminated.

A British consulting firm uses an agent for its bids. The AI reads the specifications, identifies required expertise within the company, drafts the technical proposal, calculates prices and transmits the file. Success rate: identical to human proposals. Preparation time: drastically reduced.

This approach requires a complete overhaul of processes. No more having a sales director, a technical manager, and a lawyer passing the file between them. The agent handles everything; humans only keep final validation and client relations.

Middle Management in the Turbulence Zone

Early impact measurements reveal a transformation of hierarchies. Positions of coordination, transmission, and intermediate control lose their reason for being when agents take charge of complete processes.

McKinsey observes a significant reduction in intermediate management functions in companies that massively deploy agents. PwC projects significant acceleration by the end of 2026. Not massive layoffs, but non-renewal of positions and redeployment toward strategic or relational functions.

A French retail group eliminated three hierarchical levels in its purchasing. Before: junior buyer, senior buyer, team leader, purchasing director. Now: strategic buyer and AI agent managing operations. The department’s workforce declined significantly, but strategic purchasing increased considerably in volume handled.

This hierarchical compression reproduces a phenomenon observed in other sectors affected by automation. Automotive lost its foremen when robots took over assembly lines. Finance eliminated its back-offices when algorithms automated transactions.

The Post-AI Organization Emerges in a Few Sectors

The most advanced sectors outline the contours of the post-AI organization. Fewer levels, more autonomy, refocus on strategy and human relations.

In insurance, AXA is experimenting with teams considerably smaller than traditional staffing. Typical composition: a few subject matter experts, a data scientist, a customer relations manager, and several AI agents handling operations. Productivity multiplied, customer satisfaction significantly higher.

In consulting, Accenture is redeploying junior consultants toward field assignments while agents prepare deliverables. Gone are the armies of junior analysts spending their nights on PowerPoint. Humans keep only design, critical analysis, and client relations.

This transformation is not technically complex. The tools exist, skills develop rapidly. But it requires rethinking the organization from scratch. Redefining roles, redistributing power, accepting that an AI makes decisions that previously fell to management.

The Gap Widens Between Pioneers and Followers

Companies fall into two categories. Those using AI as an additional tool stagnate around 20% punctual gains. Those positioning it as an organizational transformation see their productivity explode.

This gap is not temporary. The more pioneers accumulate learning data and refine their agents, the more they distance their competitors. A bank that has been automating loans for two years now processes a file in three hours versus three days for competitors. The competitive advantage crystallizes.

Humanity is becoming accustomed to AI without measuring what it is unlearning, but companies that take the leap toward organizational overhaul gain a decisive edge.

Europe’s lag in AI infrastructure, already visible with European gigafactories facing American investments, is now compounded by an organizational lag. European companies remain more cautious about overhauling their processes, allowing American companies to accumulate operational experience.

The experimentation window will not remain open indefinitely. Companies that have not engaged their organizational transformation in the next 18 months risk discovering that their competitors now handle the same volume with three times fewer staff. At that point, catching up will no longer be an option.

Sources