Despite regulatory uncertainty, big corporates are already quietly incorporating the technology into their teams.

An anonymous FTSE-100 company is already implementing agentic AI, according to a senior data leader at the firm, who emphasised the technology is complementing, rather than replacing, human expertise.   

At this year’s Big Data LDN, the exhibition spaces thronged with vendors promising to ‘liberate’ and ‘democratise’ corporate data. The technology at the centre of this hype is agentic AI, an advanced form of artificial intelligence focused on autonomous decision-making and activity across platforms. Amid the vendor optimism, however, concrete case studies and use cases from in-house teams were notably lacking from the event’s two-day agenda. 

This absence reflects widespread uncertainty and apprehension towards the technology; despite its interest, only 1% of companies claim to have completely implemented their AI strategy. In a recent survey by research company Gartner, half of over 3,000 business leaders said they were planning only a conservative investment in agentic AI. In-house data practitioners have described a context where many firms lack consistent governance structures or enough staff with expertise to manage adoption, exacerbated by lagging regulatory frameworks, data leaks and a fear of being left behind.  

The data leader, who wished to remain anonymous, said their team is already experimenting with agentic AI, and sees the tools as complimentary to human skillsets, rather than replacing expertise. “AI will allow us to free up time for more creative work. We are not using it to replace anybody, but to support our teams.” 

Such caution around implementing AI agents reflects a broader industry conundrum, where practitioners are forced to decide whether adopting the tools lends a competitive advantage, or makes firms susceptible reputational and regulatory risks. Vendors, however, are insistent that widespread corporate adoption of the technology is just a matter of time. Sunil Soares, founder of Tavro AI, which designs AI agents for chief data officers, says “blended teams” of humans and agents will soon become a workplace norm. One of Tavro’s clients, a bank, is already using agents to classify AI use cases. Soares claims this cuts analysis time from 45 minutes to 45 seconds.    

The reality many practitioners currently face is more complex than this. The FTSE-100 leader described how their company “started from the wrong end,” using data before verifying its quality, and now faces the harder task of building governance frameworks. They said the biggest obstacle is not the new technology itself, but people. “For many, when it comes to the labour stewardship and data ownership, it feels, for them, like getting a new job. This change management is slowing down implementation. The teams are ready, but we’re still facing resistance from the business because of a lack of mature data literacy.”  

A lack of clear oversight or unified approach has meant governance failures are a major concern for in-house practitioners today. The data leader cites the case of a peer who introduced an internal ChatGPT tool without adequate safeguards in place. “The tool ended up leaking confidential salary information to an employee.”  

The possible reputational damage of these projects going wrong is high and partially explains why so few corporates were prepared to speak at Big Data LDN this year. In the background, the regulatory landscape is fragmented, as the EU’s AI Act, due to take effect next year, imposes strict obligations on high-risk systems. The US is slashing red tape in favour of innovation, while the UK has yet to set out a clear framework.  

Despite some firms trialling agentic AI, practitioners believe human oversight will remain essential for years. But with vendors pushing for faster adoption and boards considering the costs, data leaders will soon face the strategic question of whether to move first and risk being burned, or to wait and risk falling behind.