6 Ways to Reduce the Cost of AI in Customer Service

While AI technology has become significantly more accessible to most companies over the last year, cost is often a major barrier for companies looking to implement it.

Most AI tools charge some kind of fee per resolution – Intercom’s Fin, for example, costs $0.99. That’s less than the cost of resolving tickets using offshore agents – but not by much. When companies add in the cost of implementing and managing the AI system, many lose interest in making the investment.

Yet Ultimate’s Customer Service Trends 2024 report suggests that the average company that invests in AI sees a 250% return on investment. That’s a pretty compelling statistic — and it’s part of the reason so many companies plan on investing in AI in 2024.

Finding ways to maximize your team’s efficiency first — before you implement that new AI tool — will save you a significant amount of money. Making your existing tech stack work better also helps your AI solution work better down the road. 

At Peak Support, we’ve helped dozens of clients across many different industries improve and scale their customer support, which gives us a unique vantage point into what works. We have firsthand experience with all the major support ticketing tools and AI solutions, as well as a team dedicated to helping companies implement and optimize their AI platforms. 

These are some of the best ways to reduce the cost of AI in your customer service organization, so that you can benefit from its capabilities without breaking the bank.

1. Maximize automation options in existing software

Every major customer service platform has a ton of built-in automation tools that you could be using right now. That’s true whether you’re on Zendesk, Help Scout, Intercom, or some other tool. 

The question is: Are you using those features well? 

Let’s use Zendesk as an example. Zendesk can:

• Automatically close spam tickets, reducing the total number of tickets that your agents have to touch.
• Auto-reply with suggested help center articles based on the message your customer sends
• Proactively suggest contextual help in a web widget that you can embed in your product to reduce your overall contact rate.
• Gather information in a contact form, without adding unnecessary effort for your customers by using conditional fields. You can use this extra information to route tickets efficiently or automate some aspects of the response with automations or triggers.

Most companies we work with at Peak Support know these features exist. They may even be dabbling in some of this functionality. But, speaking generally, they’re usually leaving a ton of opportunity on the table. 

We recently optimized Zendesk’s chatbot for Embark, an ecommerce company that sells pet DNA tests, and reduced chat volume by 75%, while maintaining a 97% CSAT. That was a simple chatbot, not an AI-powered one. At another client, we uncovered 18 different opportunities to improve their Zendesk instance. 

Zooming out, there are two huge advantages in starting with your existing software before implementing AI solutions:

1. The learning curve is a lot smaller. You only have to learn about a feature, rather than implement a brand new tool from scratch. 
2. The cost is usually negligible. You may already be paying for these automation features, like Zendesk’s automations and triggers. Even if they aren’t included in your package, often you’re able to purchase them as an add-on with limited extra cost (especially compared to the cost of implementing a brand new AI solution).

2. Update your knowledge base content

A great knowledge base is the best foundation to build any AI solution on. 

The best cases to automate with an AI solution are typically the ones that can be solved with a simple, easy-to-follow, and, most importantly, an easy-to-find article. That’s not to say AI can’t handle more complex scenarios — it just takes a lot more work to get there. 

Fortunately, a great knowledge base also enables other forms of self-service, which reduces your customer contacts and decreases the cost of your AI. 

If your customers can find answers independently, they won’t contact you, and you won’t have to pay your AI for a resolution. 

Maintaining your knowledge base often needs dedicated, full-time attention. Many companies are starting to create new roles specifically for keeping the knowledge base updated, often called a knowledge manager. 

Since the majority of AI tools use knowledge base content in their answers, a knowledge manager is an investment that will also ensure your bot is operating at its best.

3. Optimize your knowledge base

Having thorough and accurate content in your knowledge base is important. 

But that content won’t be anywhere near as effective if your knowledge base isn’t optimized. 

If your customers aren’t finding the info they need or if they’re struggling to put that info to work because of how it’s presented, then you need to spend some time optimizing your knowledge base’s structure and content. 

You’ll find that most of the optimizations you do to help customers will make it easier for AI solutions to use your content as well. This could mean things like:

• Rephrasing headlines into questions. Questions are often more direct and make it easier for customers to identify an article that relates to the issue they’re experiencing.
• Breaking up multi-topic articles into single-topic ones. More specific articles reduce the need to scroll or read through a lot of irrelevant information to get to the one sentence that answers a customer’s question.
• Adding keywords and search terms into your content. AI solutions are getting better at understanding and connecting similar concepts, even when they’re phrased differently. That said, you’ll still reduce the training time and improve how easy it is for customers to find relevant articles by using the terminology they use in your articles.

4. Aim to reduce your contact rate

Whether you’ve already implemented an AI solution or not, it’s never too late to look at the most common issues that are driving tickets (or AI resolutions). 

Can you head off those tickets by making product improvements, adding to your knowledge base, or using alternative self-service options?

For example, some companies require customers to contact customer service to cancel a subscription. Some people might argue that’s a bad experience, but let’s set that aside for a moment and strictly think about the numbers.

If this describes your business, it’s probably worth testing if (and how) the renewal rate changes if customers are able to cancel directly on their own. How much revenue would you lose? How much time would it free up for your support team? How many AI chatbot conversations would it eliminate? 

Or say you have a technically complex product and new customers typically reach out a number of times in their first few months. 

A knowledge base will help here, but you might also have an opportunity to invest in building community as a self-service option as well (assuming you have a strong brand identity and a healthy, dedicated customer base). 

In fact, you might see more of a positive long-term impact from building a customer community  than from implementing an AI solution, because the situations your customers are surfacing require some back-and-forth discussion and creative problem-solving.

5. Assess AI like any other agent

AI needs consistent training and maintenance. 

At the very beginning, when you start implementing an AI solution (especially a generative AI one), you should treat it just like you would a new human agent. 

Your new AI will need the same level of context, time, data, product information, and attention to provide a good experience to your customers. You have to invest serious time in training it. That might be a significant upfront cost, but it’s necessary to see a good return on your investment.

On an ongoing basis, your AI solution will also need the same level of attention (maybe even more) that you would give an agent to help them develop and improve over time. That means you need to:

• QA its interactions to ensure they’re accurate and helpful
• Assess where it needs additional training (then deliver that training)
• Proactively train it to answer questions about new features or issues that come up

You might need a specialized role dedicated to analyzing your AI’s performance and improving it. The maintenance required should go down over time (as long as your product isn’t changing too rapidly) and your AI solution should also get more effective over time.

Doing this important work helps to ensure that your AI is actually delivering against the expectations you’ve set for it. It may not directly reduce the cost of your AI, but it ensure you aren’t burning money without getting anything in return.

6. Scale your AI solution slowly

There’s been so much hype about AI that a lot of companies go all-in very quickly. 

That might mean signing a contract with a high number of resolutions or setting a high automation goal of 50% or even 80% and then staffing your team around that assumption…only to find that that isn’t working like you’d hoped.

There are other common mistakes companies make when implementing AI, but the impact of all of them are the same: they’re risky, costly, and incredibly stressful for your team.

A smarter approach is to start small and scale consistently. After you’ve optimized your existing tools, adopt an approach like this:

1. Identify a small number of cases that have a high likelihood of being solved by an AI solution — if you’re an ecommerce org, maybe those are return or exchange requests. 
2. Train your AI to manage those cases first.
3. Run a small pilot for 2-3 months to see what type of impact your new solution is having. 
4. Repeat with a few additional cases.

This process allows you to scale up in a controlled manner, while also adjusting your staffing accordingly.

You’re not paying your AI vendor for resolutions, just to have customers reach back out because the issue wasn’t fully resolved. And you’re also not taking on the incredibly costly expense of downsizing or restructuring your team, only to find out you made the wrong choice and need to rehire a few months later. 

If you have extra capacity in your team from successfully implementing AI, there are usually opportunities to shift those team members towards revenue-generating tasks like proactive chat, customer onboarding, or something else.

Level up your customer experience with AI

AI solutions can be implemented quickly and in a way that’s customer-centric and cost-efficient — but it usually takes some prior experience.

At Peak Support, we offer the expertise and support needed to swiftly implement new AI tools or optimize your existing ones. We’ve helped clients solve up to 96% of chatbot interactions without human intervention. 

We’ve built a highly experienced team that can work with a variety of tools to develop tailored solutions for your company’s needs. If you want to level up your customer experience with AI, contact our team today!