Salesforce: How Agentforce Brings AI Agents to Life

Salesforce is betting that AI agents will become as essential to enterprises as customer relationship management itself.

For Paul O’Sullivan, CTO for the UK & Ireland and SVP of Solution Engineering, that shift is already well under way.

Speaking to Technology Magazine at the Agentforce World Tour in London, Paul points to the inflection point where digital labour – delivered through AI agents – is moving from experimentation to large-scale production.

Customers from Thames Valley Police and Pandora to FedEx and PepsiCo are all looking to automate routine work while deepening human engagement through the power of Agentforce.

In short, Agentforce is Salesforce’s AI platform for designing, deploying and managing autonomous agents that plug directly into core business workflows across sales, service, marketing and beyond.

What sets it apart is its ability to draw on an organisation’s own data to deliver relevant, secure and proactive automation that goes far beyond a simple chatbot – enabling systems that can genuinely reason, act and collaborate across departments.

“If we go back a year, we launched Agentforce – the first agent platform globally – which is incredible,” Paul says.

“Fast forward to today and our customers have completed 1.2 billion triggered workflows through agentic processes and agents, which is remarkable, in just 12 months.

“Looking ahead another 12 months, we’ll see even wider adoption. Customers will increasingly leverage the power of AI across everything they do.

“Agentic AI will transform the entire value chain – from marketing and prospecting, through outreach and sales conversion, to service and customer care. 

“Agentforce truly enables end-to-end transformation.”

The impact of Agentforce

For Paul, the customer stories that bring him the most joy and fulfilment are those where AI agents sit directly between organisations and the people they serve.

Thames Valley Police, for example, launched an AI-powered assistant – aptly named Bobbi – to handle queries from the public online and alleviate pressure on its human workforce, allowing them to focus on contact that requires compassion and sensitivity.

“Bobbi has completely transformed the way they engage with citizens,” Paul says, speaking just eight days after the agent’s launch.

It is designed to handle frequently asked, non-emergency questions and chats with users in a human-like way, despite being fully automated.

Tested by more than 200 people during its trial, it draws only on information provided by Thames Valley Police and Hampshire & Isle of Wight Constabulary to give advice or direct people to the right online services.

Built on Salesforce’s Agentforce platform, Bobbi is part of Thames Valley Police’s broader move towards agentic AI – software agents that can understand requests, reason across data and then take actions on behalf of users.

“Agentforce and agentic AI – enabling digital labour – are transforming our customers’ businesses,” Paul says. “It’s absolutely incredible.”

Simplyhealth: Reimagining healthcare access with Agentforce

Health insurance provider Simplyhealth uses Salesforce’s Agentforce platform to rethink healthcare access for millions in the UK.

Chief Customer Officer Claudia Nicholls sees it as key to delivering fast, personalised support at scale – all without losing human warmth.

“Everything starts with our purpose,” she tells Technology Magazine at the Agentforce World Tour in London.

Simplyhealth exists to improve access to healthcare for everyone, serving around two-and-a-half million customers.

For Simplyhealth, growth had long meant rising costs – but now, Agentforce breaks that cycle by automating routine tasks.

For Claudia, it was all about asking the right questions: “What can we change to make that faster? Simpler? More intuitive?”

By deploying Agentforce, Simplyhealth now speeds claims, unifies member data for agents and frees them for complex cases.

The technology acts as a digital team member.

“Our customer service teams are amazing – they’re our inspiration,” she says.

“What we’re doing with Agentforce is harnessing their empathy and warmth and building that into our technology. It’s like having a team member who does the tasks you don’t want to do. Composing an email might now take one minute rather than 12. 

“The agent can work on Christmas Day, New Year’s Day, weekends or evenings, so it improves team morale by reducing workload pressure.

“The team can focus on higher-value work, while customers get answers whenever they need them – 24/7, 365 days a year.”

There have been significant cost savings for Simplyhealth, too.

By reducing the number of roles in its customer service operation by 40% – by redeploying its team, rather than making anyone redundant – Simplyhealth’s customer service team now earns 35% more than they did three years ago.

Agentforce allows both AI agents and human team members to be more personable.

For Claudia, the technology itself was the easy part – 25% of the effort.

The other 75%, she says, “is about change management, processes, and winning hearts and minds”.

She explains: “We did an exercise called the ‘wills and skills matrix’, looking at whether people had the will to work in the new way and whether they had the skill.

“If someone had the will but not the skill, we asked: how can we help? What training, support or coaching do you need? 

“The average tenure in our contact and claims teams is 23 years, so many of our colleagues aren’t digital natives – yet it’s amazing how many now love having the tech as a team member.”

​Now armed with a digital workforce, Simplyhealth can set out on its next challenge: becoming the marketplace where demand and supply for healthcare meet.

“To do that, the right technology – including AI and Agentforce – is crucial,” Claudia says.

“It will also allow us to reach more people. At the moment, only 14% of UK workers have any form of health cover – to me, that doesn’t feel right.

“As we grow into that marketplace role, we expect to serve not just two-and-a-half million members, but five, even 10 million.

“That will make a real difference to the health of the nation.”

Paul adds: “When we look at what we’ve been able to achieve for our customers, it’s remarkable.

“Simplyhealth was under real pressure managing high volumes of FAQs and case management, which took time away from their core mission – providing truly personalised healthcare support. 

“Now, with Agentforce, they can serve twice as many customers and offer deeper, more personalised experiences.”

Enhancing experiences from shopping to sport

Agentic AI does more than some intelligence systems – it is designed to operate with a high degree of autonomy, moving beyond simple text generation to actively plan, make decisions and execute multi-step tasks.

These agents can break down a user’s objective into steps, call software tools, retrieve data and complete tasks, from answering questions to updating records.

The breadth of use cases emerging on Agentforce is central to how Paul sees Agentforce becoming, well, more of a force.

Pandora epitomises this. With Agentforce, it built a personal shopper agent named Gemma, designed to support customers from discovery through to purchase.

Gemma uses customer data stored in Salesforce’s Customer 360 platform and natural language prompts to recommend products and guide buyers, helping to reduce friction in digital commerce journeys.

And what truly sets Gemma apart is the fact that customers can make purchases directly with Pandora, without leaving Gemma’s communication function.

Then, there is LIV Golf. The sports association has deployed an agent named Caddy, designed to transform fan engagement by offering personalised information and services around events.

Paul says Caddy is a prime example of how AI agents can sit on top of existing data and systems to create entirely new experiences.

These case studies illustrate a pattern that Salesforce is keen to promote: AI agents as digital team-mates that are specialised, branded and tuned to specific business outcomes

But for Paul, it doesn’t stop there.

He expects agentic capabilities to spread across the entire value chain – and sooner than you might think – reshaping front- and back-office functions, stitching together marketing, sales and service processes that have traditionally lived in separate systems. 

Trust, safety and governance

As enterprises across industries embrace agentic AI, trust, safety and governance are, understandably, top priorities.

Paul insists that, for Salesforce, this is far from a new conversation. 

“Trust has been Salesforce’s number one value since our inception in 1999,” he says.

“It’s guided our product development and the way we’ve structured our business for more than 25 years.

“We’re leading the way in deploying AI safely and securely across the enterprise.

“The trust layer in Agentforce 360 allows us to check for hallucinations and toxicity and to apply guardrails around everything.”

As well as this, Salesforce has launched functionality in Agentforce Builder that allows customers to script components within an agent, giving even more control and more deterministic outputs. 

“That means the agents our customers deploy are safe, secure and accurately represent their business,” Paul adds.

Despite the momentum of Agentforce and agentic AI more broadly, many organisations are still at the pilot stage with AI agents.

There is some anxiety there, as Paul notes: “There’s always some nervousness. Over the past few years, with the rise of ChatGPT, there’s been concern about whether AI might replace people. But the reality is that it won’t – it will elevate what people do. It enables them to focus on higher-value work.”

He lends some advice to those starting on their agentic journeys: “Start small: iterate, test and learn. Don’t be afraid to experiment.

“Try things out in specific domains, test with small customer segments and then build from there. Over time, you can expand your agent’s capabilities to unlock more value for your business.

“It’s all about starting somewhere – leaning in, asking for support. This is going to be a phenomenal pivotal moment for enterprises to really transform, to move into that agentic era.”