Consumers today have high expectations for customer service. They want service that is fast, personalized, and available wherever they are. Companies want to provide great service, but with new channels such as web chat, messaging apps, and in-app support cropping up, it’s hard for companies to know where to invest their customer service dollars.

Enter bots. In many ways, bots have been positioned as the answer. Many customer inquiries are routine and, in theory, could be easily (and inexpensively) handled by bots. Bots also have been seen as a panacea for overworked customer service teams who are struggling to scale fast enough to meet demand.

Despite this potential, some companies have reservations. Some of the earliest and most high-profile bot roll-outs have suffered mishaps that can create lasting damage for a brand. Aside from that, companies worry about whether customers will know if they’re talking to a live agent or a bot. Can they guarantee a good experience with a bot? Will bots will be able to parse sentiment like sarcasm or frustration? And how will bots work with a company’s existing call center infrastructure?

I always tell companies who are considering deploying customer service bots two things. One, we’re in the early days of bots. There are things that bots are good at and things that bots aren’t yet ready to do and, if deployed, would result in a sub-par customer experience. The best bot use cases today are simple and straightforward. Two, bots should be added to your customer service mix to help agents, not replace them. Adding bots is not a zero sum game. Deployed correctly, bots can dramatically scale your service operation and help your agents to be more productive and ultimately happier, leading to happier customers.

What bots today can do well

Simple bots are designed to handle mundane, repetitive tasks. For instance, rules-based bots can be assigned operational tasks like gathering a customer’s contact information or their reason for calling. Rules-based bots can triage initial requests before assigning cases to trained customer service agents, who can handle more advanced inquiries. Rules-based bots can process and log vast amounts of data, but human agents are still vital for resolving more complex tasks and understanding the customer’s tone and sentiment.

A similar transition happened decades ago when ATMs were introduced. Some speculated that bank tellers would quickly become a thing of the past. However, the opposite happened. ATMs have continued to exist alongside human bank tellers. The result has been that customers now have far more flexibility — such as the ability to withdraw money 24/7, something that very few banks could afford to offer if ATMs didn’t exist. By automating routine transactions with ATMs, bank tellers are able to focus on more high-value interactions, like helping customers apply for a mortgage or open up a savings account.

How bots can use AI

In the future, bots and artificial intelligence will be increasingly interwoven. Bots will get smarter with machine learning and take on more complex tasks while still working in tandem with human agents. We envision a world where AI-powered bots act like an extension of the service rep, sifting through vast amounts of detailed product knowledge articles and surfacing that information to the service agent in a seamless way so the agent can continue conversing with the customer while the bot does the heavy lifting on information retrieval.

Bots that use AI and machine learning get smarter with each interaction. The sooner you adopt, the sooner you can learn more about your customers and improve your company’s customer experience. Even if your organization isn’t ready to deploy customer-facing bots, you can deploy them internally to learn what the most common customer questions coming into your contact center are.

Selecting the right bots for your customers

One area where bots show a lot of promise is in messaging apps. As apps like Facebook Messenger and Line become increasingly popular, more companies are adding messaging apps to their service mix.

Bots lend themselves well to customer conversations in messaging apps for several reasons. The experience is highly personalized. It’s asynchronous so customers can respond when they want to. It keeps a history of the conversation, and the customer can even, say, send it a picture of the broken item to illustrate their problem. Supplementing traditional voice-based communications with bot-based interactions on messaging apps can help your company reduce customer service costs while increasing customer satisfaction.

Building bots for the present and future

Building a bot doesn’t require advanced engineering. Commands can be relatively simple, with limited responses, so that once preliminary issues have been resolved, service agents can hop in. Bots aren’t a means to an end, but another way to enhance customer interactions. A recent Forrester report, “The State of Chatbots,” sums up the current relationship between bots and agents quite eloquently:

AI-based chatbots are like seeds. You plant them, but then you need to feed them with high volumes of consumer interaction so they grow. Chat developers and designers are the gardeners: they have to tend to the chatbots and coach their growth through continuous, yet gentle, correction.

Every company wants to provide a great customer experience across the channels where their customers are. Bots can help companies free up their service agents and focus their time on the more complex or critical customer inquiries. Any company looking to improve their service offerings or extend to new channels should be looking at the entire spectrum of bots — from helping agents to providing fully automated experiences — to see what works best for them.

Meredith Flynn-Ripley is the vice president of messaging at Salesforce.