Service teams are ready for AI. But are their customers?

The buzz around AI service agents has shifted from “what’s possible?” to “where to next?” Between 2025 and 2026, adoption rose from 39% to 66% , and service professionals now trust agents for everything from simple tasks to sensitive requests, predicting demand to coaching reps, proactive outreach to case resolution. There’s a lot to be excited about.

But that doesn’t mean your customers feel the hype. Teams see increasingly capable agents that can provide on-demand contact centre support and free up reps for more complex cases. Meanwhile, customers still have a justifiable hangover from rigid chatbots, dead-end loops, and the frustrating feeling that speaking to a human is off the table.

Trust doesn’t arrive in the post once your agentic technology improves. It has to be earned through the customer experience. That starts with trusted data, clear guardrails, accountability, and AI agents that know when to mirror emotion, when to take action, and when it’s time to hand the case over to a human.

The good news is that scepticism isn’t set in stone. In this guide, we’ll dig into the data from our latest State of Service report to show how service teams can create trusted AI experiences and scale their agentic enterprise without asking customers to rely on faith.

• Australian service professionals overestimate customer trust in AI chatbots, but good agentic experiences can quickly change customer perceptions.
• The first step toward powerful AI agents is connected, trustworthy data. This gives agentic AI the context it needs to produce reliable outcomes consistently.
• Human handoffs, clear disclosure, and upfront governance help organisations scale AI safely and sustainably.
• The best way to build trust in AI is by building great AI agents. You should also be transparent about AI use and give customers a clear path to speak to a rep.
• When you get those foundations right, there’s no end to the opportunities.

Read on to see how businesses like yours are paving the way for trusted, agentic service.

• Service teams are sold, but customers need convincing
• Better outcomes build better trust
• So, where do we start?
• The context: Grounding every response in trusted data
• The agent: Clear guardrails stop false promises
• The customer: Keep people in control
• From strong foundations to extraordinary service
• Build your agentic enterprise on trust
• FAQs

Service teams don’t need convincing that AI agents are ready for real work. Among Australian service professionals already using AI agents, 90% have them delivering faster, more personal service to customers, while 86% use them behind the scenes to improve productivity and free up time for strategic work.

When service teams see these benefits, they often assume that consumers see them, too. However, the reality is a little less rose-tinted. In Australia, 68% of service professionals believe their customers fully trust AI. But when you ask customers directly? Fewer than 44% say they trust AI to handle their customer service needs.

That gap comes down to perspective and experience. Every day, service teams get a front-row view of what AI can achieve. They see autonomous AI case resolution getting better, first contact getting faster, and agents taking care of the admin that used to swallow up the working week. From that side of the fence, putting trust in AI feels natural.

But customers judge AI by personal outcomes, not long-term trends. An agent could handle their first four requests perfectly, but if it loses context and gives an incorrect answer during the fifth interaction, that failure is the experience they remember. The business sees an 80% success rate; the customer sees the one time AI got it wrong.

The encouraging part is that good experiences can change perspectives. Despite ongoing low trust, 73% of healthcare consumers , 69% of financial services consumers, and 64% of retail consumers said AI-powered customer service exceeded their expectations. Customers aren’t necessarily closing the door on AI; they just need a better reason to invite it in.

Trust grows through consistency. When customers can see that successful outcomes are the norm, this chips away at the chatbot hangover and gives them a reason to trust the next interaction that little bit more. The potential is limitless, but service teams have to put in the work today to build the foundation that will unlock it.

Consistent service experiences start with consistent AI. You need to build a framework that ensures agents have the information to be useful, the boundaries to act safely, and the accountability to earn customer trust. We like to break it down into three core layers:

1. The context: Give agents access to connected real-time data while keeping customer information protected with permissions, encryption, and masking.
2. The agent: Set boundaries around what the agent can say, decide, and do. Define when it needs to request approval or hand over the case to a rep.
3. The customer: Make AI involvement clear, keep customers in control, empower them with choice, and give them an easy route to a rep.

Let’s look at each of these layers one by one, starting with the context that agents rely on.

AI can only provide personalised responses if it has access to live, accurate customer data. If crucial context is scattered across apps and legacy platforms, stored in inconsistent formats, or telling conflicting stories, the agent has no reliable way to identify which record it can trust. That’s a recipe for incorrect responses, dead-end chats, and frustrated customers.

Your first job is to build a data foundation. You need to bring all of your customer context into one shared view, standardise it into a consistent format, and keep it current so agents can access it when required. You’ll also need to build in security and governance from day one, so sensitive data stays protected as agents process and act on it.

While a unified foundation isn’t exactly a new concept, connectivity is still posing problems for service teams. Sixty-three per cent say data readiness is a major blocker to their agentic AI adoption, and 70% say data privacy concerns delay their rollouts. There’s a clear gap between ambition and execution that teams need to address to unlock what AI can deliver.

Part of the problem is that traditional backend data integration methods weren’t built for agentic AI. Batch ETL can’t keep up with the real-time data agents demand. Point-to-point integrations link two apps, but they’re laborious to manage at scale. They also scatter governance across a web of connections, making policies harder to enforce consistently.

So, what’s the actual solution? For many businesses, it's to invest in a unified platform that creates and maintains the whole ecosystem on your behalf.

Data 360 continually ingests and syncs your structured and unstructured data using zero-copy integrations and built-in connectors. From there, the platform will help you harmonise that data, apply governance, security, and privacy policies, then activate it in real time across teams, agents, and workflows.

MuleSoft takes these capabilities even further by bringing your third-party and legacy systems into the same ecosystem. Instead of building a new one-off connector for every integration, teams can create reusable APIs and govern them in one place. This lets Agentforce retrieve live information and take action across your entire ecosystem.

• Data foundation for the Age of AI
• 5 Strategies to Activate Unstructured Data
• Data 360 Fast-Start Playbook

Unified context makes agents more valuable, but businesses can’t open the floodgates and allow every agent unlimited access to sensitive data and high-impact actions. The wider an agent’s scope, the more damage it can do if it exposes private information or promises something the business can’t honour.

Before deployment, you’ll need to define AI customer service guardrails that determine the topics each agent can handle, the actions it can take independently, what it needs to avoid, and when it's time for a human to step in. From there, test those rules extensively before launch and monitor how the agent behaves in production so you can refine them over time.

This process is far more straightforward when you have an agentic enterprise platform behind you. For instance, our Salesforce Platform provides a suite of connected tools to help you define, enforce, test, and monitor boundaries across the agent lifecycle:

• Agentforce Builder lets you build and define each subagent’s scope, instructions, and available actions using AI-supported flows and drag-and-drop capabilities.
• Agentforce Script allows you to turn business policies, checks, and human representative escalation points into fixed rules and steps the agents need to follow.
• Agentforce Trust Layer protects the data agents use and helps detect harmful or inappropriate responses before they reach the customer.
• Agentforce Testing Center lets you generate and model interactions at scale and test whether the agent is performing as it should be before it reaches customers.
• Agentforce Observability provides session-level traces, health monitoring, and alerts to help you monitor agent performance and solve issues faster.

The exact guardrails you set will depend on the sensitivity of your data, the consequences of an incorrect action, and whether the action is reversible.

As it stands, just 6% of organisations give agents no access to sensitive data. However, the rest of the respondents either provide limited access, full access, or vary their approach by agent. The important thing is that AI should only see what its assigned job requires.

• A Practical Guide to Reliable AI Agents That Work
• The Grown-Up’s Guide to Operating an Agentic Enterprise
• The Agentforce Guide to Context Engineering

In a perfect world, every customer would see AI agents through the eyes of Australia’s service teams. But the reality is that low confidence and scepticism don’t disappear overnight. The onus is on businesses to empower customers with choice and show them that AI agents can make service faster without taking away control.

The best way to build trust in AI will always be, well… building great agents. However, there are additional controls that can put customers at ease and ensure they stay in control.

• Transparency: Customers need to know when they’re speaking to AI and go into the interaction with clear expectations about what it can help with.
• Explainability: If the agent gives a recommendation, make sure it explains the reason it reached that outcome. If it references a source, ensure it’s easy to access.
• Choice: Let customers decline AI assistance, move to another channel, or ask for a person at any time. Make that option clearly visible.
• Permission: Ensure the agent asks for explicit permission before it takes actions like changing an account, placing an order, or issuing a refund.
• Continuity: Customers shouldn’t have to repeat their explanation because the service rep is starting from scratch. Make sure context carries over during handoffs.

Service Cloud helps businesses build these principles into the interaction itself. Teams can specify how an agent introduces itself, which actions it should explain or confirm, and the exact conditions that trigger escalation. You can also add AI transparency and opt-out features that let customers decline AI support or request a human rep whenever they like.

And the best part about Agentforce? When the customer asks to speak to a rep, your agent will automatically route the conversation with all of the notes attached and add all of the context to the shared record. This means the rep can see who they’re talking to, what the agent has already done, and exactly what needs to happen next.

Besides sparing customers the headache of repeating themselves, this gives your teams the breathing room to do what they do best: bring the judgement and empathy that only people can provide.

• Understanding Trusted Agentic AI - Trailhead
• Scaling AI You Can Trust
• How We Build Trusted AI

When you get those three foundational pillars in place, there’s no limit to how high you can build your agentic enterprise.

We could fill this section with statistics about how agents improve customer satisfaction, increase rep productivity, reduce average handle times, strengthen retention, and speed up first responses. But while all of those things are true, we thought it’d be a better idea to look at the businesses who are turning service ambition into agentic execution.

Take Xero, which now uses Service Cloud to resolve 60% of online queries instantly, while giving 4.9 million subscribers access to 24/7 support. Or Panasonic, which now resolves 73% of their conversations without escalating to a live advisor, helping it scale its service with less friction.

And customers aren’t the only beneficiaries. The same trusted, responsible agents can also take pressure off the people delivering the service. At Team Global Express, an agent named Āwhina provides instant parcel updates and resolves half of common requests. Calls to the service team have now dropped by 50%.

That last point is one you’ll see running through most agentic success stories: Agents absorb routine demand and keep the work moving in the background, while service teams gain time for the expertise and human judgement that AI could never replace. The result is service that feels faster and more responsive without ever feeling less human.

Trust is built on solid ground. Create a bedrock of data, set rock-solid guardrails, and make it a priority to keep customers in control. When you can get that right, the opportunities follow.

Salesforce is built to help organisations lay that foundation. MuleSoft and Data 360 bring your systems and data together, make them governable, then activate that context across the business. From there, Service Cloud puts those insights to work, resolving requests, supporting reps through complex cases, and carrying context forward.

If you’d like to dig into the data from this guide, access our State of Service: AI Agents Edition to get insights from 3,000+ service professionals on the current challenges and opportunities facing the industry today.

Or, if you’re ready to start putting theory into practice, explore Service Cloud today to see how our platform can help you build trust into every interaction.