Banking isn’t really about money: It is about people. People who have money, yes, but in order to build a successful career in banking, you’ve got to be able to build a trusting relationship with clients.

On that front, AI can help, creating new ways to make the relationship more flexible and productive. However, bankers do need to be careful that AI doesn’t replace the human in that relationship. Here are nine ways that AI can help bankers be more productive and create better customer relationships.

A teller that never sleeps

Your clients don’t keep bankers’ hours, so you can’t just close up and go home at 5 p.m. That’s expensive, though: Running a 24-hour call center is complex, and outsourcing can sour the customer relationship if done poorly. AI can help: Services like Interface.ai can create AI agents that are always available, always polite, and which use the same language as your own people, because they can be trained on their work. They can work with text chat, voice, or email to answer customer questions on their schedule at a lower cost to you.

Support for employees

Don’t forget that AI can work the other way around as well: Consider building a chatbot that can work with your employees, quickly answering questions they have on process, procedure, and who to ask for further information. Services like Notion and Kasisto can take information from your own internal documentation and present it in new ways that make them more efficient.

A helping hand for back office tasks

The same is true of your back office: AI is great at handling all of the mundane business of data entry, creating reports, filing regulatory paperwork, and other boring stuff. Companies like Objectway offer solutions that use AI to handle most of these tasks for small banks and wealth management services.

An always-on assistant

When you meet a client, you want to keep notes, but nobody likes a scribbler. Zocks.io provides an interesting alternative: It is an AI note taker designed for financial professionals. That means that, unlike general note takers like Otter.ai, you don’t need to train it to understand your language, and it already has the appropriate security certifications (SOC 2). Once the meeting is done, it automatically produces a summary and creates follow-up and to-do items. All of which are automatically stored in your CRM system.

Although this is likely to get lost in the post-election rush, the White House recently announced that they are looking to clamp down on chatbots and AI “doom loops,” where customers get caught in a never-ending cycle of push-this-button-then-this-button. Their reasoning is sound: Many companies see chatbots and AI as a way to deflect customers, to replace real people with bots that don’t have any real answers. Remember: Customers aren’t stupid, they know and remember when a system is designed to obfuscate rather than help.

Help with incoming paperwork

New clients tend to mean one thing: paperwork. Especially when they come from another company that just sends you a big pile of paper. Feed it to an AI designed for Intelligent Document Processing (IDP) from a service such as Eigen Technologies, which can input it, read it, categorize it, and move it into your own systems.

Offering new perspectives on risk

In an uncertain world, bankers have to manage risk. That requires information, which is often changing on a daily basis. While it is near impossible for people to track this, AI loves this sort of challenge. The Department of Homeland Security, for instance, has a program called FloodAdapt that is using AI to help create new flood risk analysis maps and predictions using the latest Machine Learning and modeling tools. The same is true of fire: The DHS is working with UC San Diego to apply AI analysis to wildfire data collected by firefighters themselves in California in a program called WiFire, which is using this data to produce new fire risk maps for the San Diego area.

A tool for seeing where customers are—and will be

Credit reports are the bête noire of banking: They provide a great indication of where your customers are, but they don’t tell you where they are  going to be in the future. And customers remember: Nobody likes a bank that is only helpful when you have money. AI can provide a way to look beyond the credit score by using multiple sources of data to provide a more nuanced view of credit risk.

Speaking your customers’ language

Running a multilingual help desk can be an expensive business, but AI is making it easier. Systems like Microsoft’s Conversational Language Understanding can speak multiple languages: Microsoft claims it can work in up to 96 languages from around the world. This isn’t just a translator that works without context: You use the transcripts of your staff with customers to train the system so it responds like they do, but in different languages.

A better mechanism for customer feedback

The marketing people like to call this sentiment analysis, but the idea is simple: What are people saying about you on the Internet? Honestly, it can be a depressing business, but services like Alchemer Pulse and Qualtrics XM make it easier, using AI to listen to the lovers, haters, and trolls and analyze what they are saying.