13 ways to use conversational AI to improve customer service

When your customers need you, they expect to be able to communicate with a representative straight away. In today’s market, where 43% of customers will switch to a competitor if they receive poor service, it’s important to use customer service opportunities not just to fix a problem, but to build lifelong loyalty.

While great customer service was often seen as a point of difference, it’s now table stakes. In fact, we found that 82% of service professionals agree that customer expectations are higher than they used to be.

To tackle this heavy competition and high expectations, teams are turning to conversational AI. With 69% of companies now using AI (up from 24% in 2020), customer care is entering a new era of speed and personalisation. Conversational AI helps teams manage higher case volumes, offer 24/7 assistance, and deliver more consistent experiences.

In this article, we’ll look at how teams are using conversational AI to improve customer service, using insights and statistics from Salesforce’s 7th Edition State of Service Report , where we gleaned insights from more than 6,500 service professionals worldwide on customer support in the AI era.

Here are 13 ways you can deploy conversational AI to take your customer service to the next level.

Customer service software with AI is a great option for first-line responses to customer queries. Many teams are already using it to take the resolutions they’ve previously given to customers and turn them into knowledge base articles. In fact, we found that this is the number one use case for customer service teams using AI (followed by order inquiries).

By using generative AI to create and update FAQs, customers can find accurate answers themselves without needing to contact a representative. This saves time; it also feels easier and faster for the customer.

One example of success can be seen from Fisher & Paykel, which now offers 24/7 support using Agentforce. Based on this change, Fisher & Paykel expects to more than double self-service rates, reaching more than 65%.

There are only a handful of times that a brand actually gets to speak face-to-face with its customers. Support is the frontline of customer interaction. This is why, when a customer reaches out for support, it’s a bigger opportunity than simply solving their issue. It’s also a chance to learn more about their needs and how your business can be their standout provider.

When you get this information, you can use AI-driven selling. This works by connecting your CRM and service data. From there, AI can analyse customer preferences, purchase history, and behaviour to make personalised suggestions. These recommendations turn ordinary service conversations into meaningful sales opportunities that feel natural and helpful, not forced.

We found in our State of Service Report that offering personalised product recommendations is one of the top AI use cases in service. With tools like Service Cloud, teams can use data insights to provide specialised suggestions that build relationships and sales (watch the Service Cloud demo). In practice, a SaaS company could use AI to spot when a customer asks about features not in their current plan, and then send an email offering an upgrade for what they need.

One great example of this can be seen from the jewellery brand Pandora. They built an AI assistant using Salesforce's tools and called it Gemma. Gemma helps customers find the perfect piece by learning about the occasion, recipient, and budget, and then suggests items that match. Behind the scenes, Agentforce connects this data from Service Cloud and Agentforce Commerce to make each recommendation relevant and on-brand.

Want to learn more? Watch David Walmsley, Chief Digital and Technology Officer at Pandora, share how using the full Salesforce suite has transformed their operations. You can watch this and more on Salesforce+.

Customer service teams spend hours reviewing chat logs and call notes to understand what’s already been discussed with a customer. This is repetitive work that slows down response times and can easily lead to missed details. Generative AI allows them to automatically create quick, accurate summaries of customer interactions.

Through implementing AI to compile key points from past conversations, agents can get up to speed in seconds instead of minutes. This means less time spent digging through old messages and more time focused on resolving the issue at hand.

In our latest State of Service Report , we found that conversation summaries are now the third most common use case for AI in service. Health technology company Magentus is a great example of how this can play out. Using AI to generate post-call summaries within one customer stream, they were able to reduce customer wait times by a huge 60%.

When a customer needs urgent help, it can feel frustrating to wait while an agent searches in real time for the answer. Outdated tools and messy systems make it harder to respond quickly and keep your customers happy. AI fixes this by pulling up the right answers instantly during live conversations.

Instead of switching between tabs or searching through long documents, agents can see the right information in real time. This makes it easier to solve your customers' problems the first time they contact you and keeps the advice given by different support representatives consistent.

In our State of Service Report , we found that knowledge retrieval ranked as the fourth most common AI use case in customer service.

One business using knowledge retrieval to improve its customer experience is TripADeal. Using Agentforce and Data Cloud, they built an AI agent that instantly gives answers and personalised travel recommendations. The AI agent does this by using Retrieval-Augmented Generation (RAG) to search across unstructured data. The result has been that all their customers receive quick, relevant responses no matter what timezone they are in.

Customers want to connect with businesses in a way that suits them, whether that’s by phone, chat, email, or text. However, this often means conversations don’t carry over between channels, leaving customers needing to repeat themselves or receiving different advice from different representatives. Multimodal AI fixes this problem by allowing you to connect voice, chat, and text into one continuous conversation. In our latest State of Service Report , we found that 36% of organisations using both voice and text AI support have already connected them.

One example of this in action is at McDonald’s . They now use emotion and tone tracking technology across more than 38,000 locations and online to monitor customer sentiment in real time. For instance, if a new menu item receives negative feedback or a store isn’t meeting expectations, McDonald’s can act on that data right away.

Humans are masters of noticing small tonal changes and can easily spot when a conversation sounds ‘off’. A friendly tone in one interaction and a robotic reply in the next can make the experience feel disingenuous. This either means your customers' trust is eroded, or your team has to spend hours each week touching up AI text.

Conversational AI helps solve this by mirroring your brand’s voice and tone, no matter where or how a customer reaches out. When you train AI models on your brand’s language library and tone guidelines, you can make sure that your messages sound consistent. In our State of Service Report , we found that 88% of companies said their AI tools are good or excellent at maintaining a consistent brand voice.

Having a consistent brand voice can be helpful across automated chat responses, follow-up emails, or social media messages. For example, if your brand leans into a relaxed vibe, it makes sense for your bot to start a conversation with “Hey there 👋”. However, this would feel inappropriate coming from a prestigious university.

Customer service should be easy for everyone to access, regardless of their primary language or ability. However, many systems still have barriers for those who don’t fit neatly into a single way of communicating. Conversational AI can help break down some of those barriers.

In our State of Service Report , 88% of service professionals said conversational AI has improved their accessibility for more diverse customer groups. These tools (like Agentforce) can do this by translating conversations in real time, offering built-in support for multiple languages.

Research has also found that it makes a big difference for neurodiverse minds . Chatbots allow people to ask as many questions as they need, take extra time to read and process information, and respond at their own pace. For many, it removes the pressure of a phone call or in-person interaction, creating a calmer and more comfortable support experience.

The result of higher levels of inclusion leads to a better experience for everyone involved (customers and service reps).

Even with the rise of automation, some customer problems are too complex or sensitive for AI to handle alone. That’s why the most prepared service models combine both the human expertise of their people with the speed of AI.

In practice, AI can triage simple requests, like password resets or order tracking, so human agents can focus on high-impact issues that require empathy, judgment, and creative problem-solving. This collaboration helps teams resolve customer issues faster while keeping morale high, since agents spend more time on meaningful work rather than repetitive tasks.

In fact, we found in our State of Service Report that 82% of service professionals said complex cases are best resolved when humans and AI work together.

A strong example of this in practice comes from One NZ, which uses Agentforce 24/7 to help its prepaid customers transition to new plans. The guardrails set in place by One NZ prevent the AI from allowing incomplete transactions, while complex questions are automatically escalated to human representatives.

A clear example of predictive AI in action can be seen from Amazon . Their new AI model predicts what customers will need to help prevent delays and supply chain issues before they happen. Through analysing trends like demand, seasons, and local weather, it can make sure it has products in the right place at the right time.

Every customer service team has to do repetitive tasks that take up time; these are often things like updating records, storing documents, or sending follow-up emails.

Using AI automation, these frequent jobs can now be done automatically. This gives service reps more time to focus on the complex cases and customer relationship-building activities that need a human touch.

According to our latest State of Service Report , service leaders using AI agents expect both their service costs and case resolution times to drop by an average of 20%. That’s a huge win for teams under pressure to do more with less.

One great example of this can be seen at BizCover, which supports more than 200,000 small businesses. Using Service Cloud, they’ve automated much of their customer communication and case management. This has contributed to an average NPS of over 74 over the past six months, with some months over 80.

AI is only as strong as the data behind it. When customer information is spread across different platforms, it’s harder for AI to make accurate predictions. When businesses are able to bring all their service data together in one place, they can get the full potential of their AI tools.

In the State of Service Report , companies that have unified their service channel data on a single platform are 1.4 times more likely to describe their AI implementation as very successful.

A great example of this success comes from Kellanova (formerly Kellogg's). They brought in data from their more than 20 systems together in Data Cloud. After doing this, they were able to reduce the time it takes to address refund claims by 99.8%.

Conversational AI can now act as a real-time coach, offering instant feedback and guidance during live interactions. For example, it can suggest how to phrase a response more clearly, when to show empathy, or when to escalate a tricky case. In our State of Service Report , we found that 82% of service professionals said AI has helped them develop new skills, and 83% said it’s improved their career prospects.

This kind of coaching helps service teams gain confidence and sharpen their skills while on the job. It also gives managers better insight into individual performance trends, making it easier to deliver personalised training and recognise top performers.

One example of this can be seen from Precina, a healthcare company that used Agentforce and Health Cloud to train new clinicians on patient intake calls. Typically, this kind of training was resource-heavy and hard to scale. When they started using AI coaching, clinicians could practice real patient scenarios and receive instant feedback. This always-on training model helped Precina grow their team while always improving the quality and consistency of their support.