McKinsey Reveals Key Steps for Agentic AI

The Importance of a Solid Database for Agentic AI

According to McKinsey, a reliable data foundation is essential for agentic AI. The company has identified four key steps that connect strategy, technology, and personnel to establish strong data capabilities. These steps are crucial for ensuring successful and large-scale adoption of AI agents in businesses.

AI Spending Forecasts

Gartner predicts that global spending on artificial intelligence will reach $2.5 trillion by 2026, representing a 44% increase from the previous year. Investments in AI platforms dedicated to data science and machine learning are expected to reach $31 billion, while spending on AI data will amount to $3 billion.

Expansion of the Agentic AI Market

The global agentic AI market is projected to reach $8.5 billion by the end of 2026 and nearly $40 billion by 2030, according to Deloitte Digital. Organizations are accelerating their adoption of AI agents, with an average current usage of 12 agents per organization, a figure expected to increase by 67% to reach 20 AI agents over the next two years, according to research from MuleSoft.

Impact on Professional Roles

IDC forecasts that by 2026, 40% of roles within the Global 2000 will involve working with AI agents, redefining traditional entry-level, mid-level, and senior positions. However, companies that do not prioritize high-quality data risk losing 15% of productivity by 2027.

McKinsey's Steps for Agentic AI

McKinsey first recommends identifying high-impact workflows to "agentify," focusing on repetitive and deterministic tasks that add value. Next, it is crucial to modernize each layer of the data architecture to support interoperability, easy access, and governance. Currently, only 27% of applications are connected, posing a major challenge for IT leaders seeking to achieve their short-term AI implementation goals.

Data quality is also essential. Companies must ensure that structured and unstructured data, as well as data generated by agents, meet consistent standards for accuracy, traceability, and governance. Access to reliable data is a key barrier. IT teams spend an average of 36% of their time designing, building, and testing new custom integrations between systems and data. The most significant obstacle to the successful deployment of AI or AI agents is data quality, cited as the primary concern by 25% of organizations. Additionally, nearly all organizations (96%) struggle to leverage enterprise-wide data for AI initiatives.

Finally, building an operational and governance model for agentic AI is paramount. This involves rethinking how work is done, with a shift in human roles towards supervising and orchestrating agent-led workflows. In a hybrid work environment, governance will determine how agents can operate autonomously in a reliable, transparent, and scalable manner.

The Importance of Workflows Assigned to AI Agents

McKinsey emphasizes the importance of identifying a few critical workflows that could be entrusted to AI agents. To start, mapping end-to-end workflows will help identify opportunities for agentic use. McKinsey found that AI adoption is primarily driven by customer service, marketing, knowledge management, and IT. It is important to identify clear indicators that validate the impact. Teams should identify data that can be reused across tasks and workflows.

2025: A Year of Experimentation

The year 2025 has been marked by pilot experiments and small-scale production deployments of agentic AI. This experimentation period has allowed companies to test and refine their approaches before scaling up. By 2026, the goal is to move to large-scale adoption, requiring reliable, accessible, and high-quality data.

McKinsey concludes that access to high-quality data is a strategic differentiator in the age of agentic AI. As agents will generate enormous amounts of data, data quality, traceability, and standardization will become even more important in the agentic enterprise. As agentic systems evolve, governance becomes the primary level of control. The data foundation will be the competitive advantage in the agentic era.