The future isn’t a distant reality – it’s unfolding right now. Companies like OpenTable and Wiley are transforming industries with autonomous AI agents that go beyond simple assistance – they analyse data, create plans and take action.
OpenTable, which seats 1.7 billion diners annually, uses AI agents to handle reservation changes and manage loyalty points. By automating these tasks, employees are freed up to focus on building customer relationships - and saving just two minutes per call can drive significant efficiency gains.
Education publisher Wiley improved case resolution by 40% in its first few weeks of AI-agent use. By triaging registration and payment issues, the technology directs students to resources and reduces service representative workloads.
Wiley knows its workers often waste precious time on routine tasks that don’t add value. So it’s supercharging the capability of employees to create a new era of efficiency. AI agents work around the clock, creating plans, solving problems and increasing productivity.
Achieve transformation the smart way
Artificial intelligence (AI) is expected to add trillions to the global economy by 2030, underscoring the importance of nailing your AI implementation. Unlocking AI’s potential could mean tapping into the biggest economic shift of our generation.
How can you empower your workforce for the next wave of AI with minimal pain and maximum gain? Here’s the good news: you don’t need to spend months or millions training your own large language model (LLM) or building custom solutions.
LLMs, while powerful, are just one piece of the puzzle. Real enterprise AI success comes from integrating data, AI and automation. This allows AI to not only answer questions and generate text but to actually perform work for you.
Why you need a complete AI system
LLMs cannot take action. They can only produce responses to questions, which means it will never be a truly effective solution for getting AI to complete valuable work. Businesses need an AI system like Agentforce, where agents are woven into company data, business processes and everyday workflows, completing tasks alongside humans. All of this is built into the Salesforce platform, which makes it easy for companies to turn on agents quickly. Saks Fifth Avenue is a great example of this - the brand launched its first AI service agent in just one week.
Here are the 5 key elements that early Agentforce adopters have found essential to success:
Dynamic data retrieval
Real-time access to quality data across the organisation is the bedrock of successful AI. Platforms like Salesforce Data Cloud connect all data, structured and unstructured, into one unified platform. This data is then connected to the LLM. The best platforms use techniques like retrieval augmented generation (RAG) and semantic search to surface the best data for the job.
Reasoning
AI agents are capable of carrying out multistep, complex tasks, like initiating a return for a shopper and re-ordering a different item.
Central to this is a reasoning engine, which generates a plan based on what a user is trying to do. It evaluates and refines the plan, pulling data from CRM and other systems. Then, it decides what business process to use based on the request, and repeats the process until it’s right.
Security guardrails
Because AI agents can take action with little human intervention, it’s critical for them to have built-in security guardrails. The agent has to “know” what it can and cannot do. If a requested task seems outside the organisation’s guardrails, or the agent cannot access the right data to complete the job, it needs to understand that and delegate the task to a human. These guardrails also encompass permission rules around who can access and share data.
Orchestrating action
Generative AI excels at answering questions and creating content. To unlock its full potential, organisations must take action by integrating it into their business workflows and tools. Salesforce’s Flow and MuleSoft platforms enable this, with Flow automating tasks and MuleSoft APIs linking systems.
This allows AI agents to operate smoothly across your business. For example, an AI agent could manage fraud detection, flag suspicious spending, send alerts to customers, cancel cards if fraud is confirmed, initiate replacements and update the fraud case in the system.
Collaboration
AI agents and employees can work collaboratively, leaning on their unique strengths to achieve more than they could on their own. Agents can process huge amounts of data and execute complex tasks, while humans bring strategic thinking, emotional intelligence, wisdom and judgment. This partnership will redefine workflows in all industries, leading to faster, smarter decision-making.
What’s wrong with training your own LLM?
Many businesses believe training an LLM with their own data will create a model that knows everything about their business and customers. However, LLM training is expensive, time-consuming and requires specialists. For most companies, it’s not feasible.
Even if you could train your own LLM, it wouldn’t stay accurate for long. It only knows what it’s trained on, so any updates - a changed customer record - make it outdated. It’s like a GPS that works until the first detour. When the route changes, you must reboot the system.
AI pilot time is over
It’s time to scale and see real results. But many companies are stuck in the pilot phase with AI, because of data quality issues, trouble identifying use cases and the belief it can’t execute more sophisticated tasks.
Leading companies know there’s little value in building solutions from scratch. Leveraging out-of-the-box tools alongside integrated data flows, automation, and workflows is the key to delivering a better customer experience – and securing a share of the trillion-dollar AI opportunity.
To learn more please visit www.salesforce.com/uk/agentforce/
