In 2026, artificial intelligence agents are transitioning from experimental pilots to core enterprise tools driving real productivity gains, according to coverage from SiliconANGLE’s AI & Big Data Expo in London. Leaders from Salesforce and DeepL highlighted how agentic AI — autonomous systems built on large language models — is being deployed across business operations to automate tasks previously done by humans and meaningfully improve outcomes. These agents aren’t just conceptual: enterprises are using them to manage customer interactions, analyse documents, and generate business insights at scale.
Salesforce’s Paul O’Sullivan pointed to notable use cases like “Hallie,” a Salesforce-designed agent at Heathrow Airport that answers visa questions and provides security wait-time information, shifting about 70 % of customer engagement to agents, and enabling 24/7 service with reduced human workload. DeepL reported that 2,000 global customers are deploying AI agents for tasks like report analysis, sales list creation, and contract review — freeing employees from repetitive work and raising productivity.
A major factor in the growing adoption is ease of implementation. Both Salesforce and DeepL executives noted that agents can be built and deployed within days or weeks depending on the availability of structured data, rather than requiring heavy infrastructure overhauls. As these systems become more contextually intelligent, drawing on richer data to make sound conclusions, enterprises see them as extensions of the workforce, enhancing efficiency without replacing strategic human judgement.
Beyond basic automation, firms like Salesforce are emphasising governance and observability — ensuring that AI decisions are monitored, explainable, and integrated into core business processes rather than operating in opaque “black boxes.” Investments in tools like Salesforce’s Agentforce Observability reflect a broader shift toward trustworthy agentic deployments that scale across departments, signaling that AI agents may soon become a standard part of enterprise infrastructure rather than experimental add-ons.
