Interest in AI is high in UK businesses, and with good reason. According to Salesforce’s UK AI Readiness Index, the country is ahead of many other countries when it comes to the ability to adopt, deploy and integrate the technology.
Agentic AI is particularly exciting – and globally, business leaders predict a 327% growth in agent adoption within their organisations by 2027. Functions like customer service, sales, marketing and commerce are transforming rapidly.
However, achieving ROI from investments – and scaling from pilots to broader programmes – is blocking progress. In fact, 57% of leaders say their biggest blocker to investment is being unable to demonstrate results.
As businesses find opportunities to benefit from agentic AI, it’s crucial to create tangible value as quickly as possible – which is something that we’ve pursued for ourselves at Salesforce.
We’ve always been customer zero for our own technology, and right now we have the world’s largest agentic AI deployment. Here’s what we’ve learnt so far, including what’s going well and what we wish we’d known sooner.
How we’re using agentic AI
Agentic AI is the latest evolution in artificial intelligence, with outstanding capabilities. Unlike traditional automation that assists, AI agents act – they analyse, decide, and execute autonomously, handling both low-value, routine work and high-value, multistep tasks.
Using self-contained agents, these models are powered by machine learning, and algorithm and predictive analytics to make real-time decisions – from autonomously completing tasks to managing workflows.
All of that means agentic AI can deliver greater adaptability, personalisation, informed decision-making and productivity, when compared to its predecessors.
We wanted to explore the potential of AI agents to better serve our customers. So in 2024, we began using our own technology, Agentforce, to power our customer support site.
As a 24/7 concierge, Agentforce provides fast, accurate and more actionable responses to customer inquiries. Unlike basic chatbots, AI agents can understand and interpret conversational text, deliver context-aware information based on live business data and take action autonomously.
Salesforce is now handling around 32,000 customer conversations per week with a resolution rate of 83%. Importantly, escalations to humans have been cut in half, with only 1% of customers needing to speak to a human.
Routine tasks like password resets and generic product questions are now autonomously handled by Agentforce. With the added bandwidth agents provide, team members have more time to focus on high-touch customer engagements and expand their professional skillsets, for a more fulfilling experience at work.
Lessons from our experiences
- Agentic AI deployments are most successful when all of an organisation’s data and metadata are unified, secured and accessible
Like most technology deployments, the quality of your data matters. Agents need content and context to do their job effectively — so rollouts are most successful when all of the organisation’s data and metadata are unified, secured and accessible to the large language models (LLMs) that agents use to reason and plan. Think about relevant content early and often, and get it all in one place.
- Prioritise user experience when deploying agents, to provide the most helpful service
User experience is crucial for encouraging people to engage with AI agents. In our case, initially, the interface for Agentforce looked like just another chatbot, and users didn’t realise they were using something more unique. To address this, engineers revamped the site with a search-engine-like user interface and removed the “contact support” button. Now, most customers engage with agents first when they have service and support questions.
- Educate users on what agents can do and how they differ from traditional AI assistants
Customers may have had poor early experiences of chatbots – and now feel wary of agentic AI services as a result. At Salesforce, we found it was important to educate users on what agents can do and how they differ from traditional AI bots and assistants. Understanding that agents are trained on structured and unstructured data anchored in nuanced guardrails, which allows them to employ more human-like intelligence and deliver better answers faster, helps people to embrace the benefits of agents more quickly.
- Trust comes first – from your data to end users
Trust is critical for successful agentic AI deployments – whether it’s your customers, employees or wider ecosystem. Planning how to use AI responsibly and ethically must come first, rather than last, in any planning. Our Office of Ethical & Humane Use focuses on addressing ethical challenges head-on, including a comprehensive testing strategy for all our AI technology. Remember to consider your use of data, as well as when and how agentic AI will be used.
- Proactively invest in skills to support agentic AI
Agentic AI is fuelling the shift towards digital labour – a new business model where humans and agents work together to drive success. And there will be a skills impact: more than four in five HR chiefs are, or are planning to, reskill their workers to be more competitive in a market shaped by AI agents. With greater need for skills from AI literacy to soft skills, it’s crucial to get ahead of this shifting landscape and invest in your people.
Agentic AI needs a clear ROI
Agentic AI is already changing how our company works, and promises a whole new model for growth for UK businesses.
If you’re considering embarking on this next technological wave but don’t know where to start, the Agentic AI ROI Calculator can help you understand where agentic AI will deliver the greatest ROI for your business – helping you move from pilot to profit.
Agentic ROI Calculator
Get a data-driven view of how Agentforce can benefit your business.
