AI agents, powered by advances in language models, are moving to the mainstream as organisations search for a boost in productivity, efficiency and innovation. Technological improvements are also easing the digital infrastructure transition required to unlock agentic AI capabilities. These advances range from tools for orchestrating agent-to-agent interaction to platforms that allow companies to develop, build and deploy their own differentiated agents that tap into their existing data.
However, realizing the full potential of agentic AI requires more than infrastructural advancements. Success also depends on strategic implementation, combining the power of traditional analytics and machine learning with agentic capabilities.
And as with AI more broadly, delivering transformative impact and commercial returns means moving from isolated pilots to enterprisewide deployment. A key aspect of this shift involves enabling agents to retrieve multiple data inputs and activate tasks across organisational silos. In this environment, governance, monitoring and oversight become crucially important as autonomous systems are released in the wild.
This Economist Enterprise report, sponsored by Dataiku, explores how companies can successfully implement agentic AI solutions to build competitive advantage. It is based on a series of expert interviews and independent secondary research.
Key findings
Agentic AI is poised to revolutionise business operations, allowing organisations to supercharge their existing analytics and machine learning efforts. AI is entering its third wave, moving from ‘predictive’ to ‘generative’ to ‘agentic’. Today’s agents, powered by LLMs, can move across formats and mediums, to perceive, reason, set goals and orchestrate tasks far beyond the scope of conventional automation tools. They are also easier to implement into existing machine learning environments and digital infrastructure. Their benefits come at a time when businesses are battling rising costs, a volatile business environment and talent shortages.
Navigating the agentic AI frontier requires robust governance frameworks to mitigate risks, ensure compliance and build trust. By empowering agents to operate autonomously, including unilaterally accessing data and applications, organisations could expose themselves to compliance, trust or reputational risk. In this environment, executives are looking to safeguards like adversarial testing for spotting bias and security risks, retrieval-augmented generation (RAG) for improving accuracy, and continuous monitoring and observability for tracking agentic performance changes over time. The quality of agentic performance monitoring tools is also improving, and agents—built on responsible AI foundations—can become part of the orchestration layer of overall AI governance.
The true potential of agentic AI lies not just in cost savings, but in unlocking new revenue streams and driving competitive advantage. Agents can deliver quantifiable gains not just in cost reduction and productivity, but for driving revenue growth, powering new products, delivering deeper commercial insights from data repositories, and transforming customer engagement. An ambitious vision is not just desirable but necessary; the rise of agents increases competitive threats by allowing companies to operate at greater scale and even into adjacent sectors.
Data provide the critical fuel that powers agentic AI, making data quality, sharing protocols and interoperability essential for success. Agentic AI needs large volumes of high-quality data, which require greater data-sharing within and between organisations, and the evolution of new standards, collaboration frameworks and commercial models. Companies should evaluate and refine their data sources, removing outdated or irrelevant information, as poor data quality can lead to flawed outputs, and invest in the appropriate interoperability requirements to harness their existing technology stack.
