Don’t call them chatbots, call them intelligent assistants

Unless you’ve been hiding under a rock, you’ve no doubt heard about chatbots. However, what’s not commonly known is that chatbots have been around for years. What differentiates today’s bots is the integration of back-end artificial intelligence, which enables them to do more than simply respond with the basic logic of yesterday.

I find that it’s important to make distinctions between a bot and today’s A.I.-powered intelligent assistants. Chatbots just happen to be conversational. Intelligent assistants go beyond bots to perform tasks that assist the user. The future isn’t bots, it’s intelligent assistants.

Intelligent assistants are being developed extensively in both the consumer space (Amazon Alexa, Apple Siri) and for the enterprise (Nuance Nina and IPsoft Amelia). What remains to be seen, however, is how successful these assistants will be in the long run.

My advice, as someone who follows the app and bot market closely, is to think about how to make a product that goes well behind chat.

What makes a good intelligent assistant?

The key measure of success for an intelligent assistant, whether in the enterprise or consumer space, is how much value the assistant adds. This is usually in one of two categories, either performing a task a person would find hard to perform themselves or saving a person time by performing tasks that would take them a long time to do.

Let me give two not-so-obvious examples of successful intelligent assistants. The best example in the first category is the search engine. It’s clearly impossible for the average person to search for content on the internet without using a search engine. No wonder search remains Google’s primary business! For the second category, a great example is Kayak, which seamlessly searches for flight prices across many sites and saves time. Note that neither of these examples is new or would even be considered “bots” by anyone. However, the value they deliver to users is clear.

The defining characteristic of success for an intelligent assistant is the value it delivers to users. In some cases the conversational interface (e.g., the “chat”) adds value, but in most cases the true value is what happens behind the scenes. Chat may (or may not) make it easier for the user to ask the bot to do something, but it’s never useful when a bot does something that’s easy for us.

How smart do intelligent assistants need to be?

If you’re developing a bot, you need to first understand the value your users expect — and then make an intelligent assistant, not just a chatbot. Focus on ensuring your intelligent assistant can actually perform the tasks expected.

Another dimension to consider when creating intelligent assistants is how well they can perform those tasks and how much intelligence is actually needed. For example, if the goal is to create a first-level support agent whose task is to simply route cases to the right skill group in the enterprise upon initial contact, a chatbot with some conversational capability and basic understanding about your field could do that task a reasonable amount of the time. However, if the goal is to create a second-level support agent, who can actually troubleshoot issues, you’ll need an intelligent assistant designed to understand your field with deep reasoning and learning capabilities. In either case, regardless of how much A.I. is used to build this “more than a chatbot” idea, the key criterion for success is how well it does the job it needs to do, how much time it saves, and the ROI it delivers to the firm.

A successful example is x.ai. A chatbot would simply ask when you’d like to schedule a meeting and then add the appointment to your calendar. There’s not much value over just adding it yourself. However, what x.ai does is to take your input (such as attendees and other preferences) and then find the optimal time and place for the meeting. The bot schedules it with all parties to save you time. And it’s much more efficient at doing that task than a human would be.

The only “chat” involved in this example is the response letting you know the meeting has been added to your calendar (or, if not, that there were irreconcilable scheduling conflicts). The conversational interface makes it easier to interact with the intelligent assistant, but the true value is the behind the scenes work the assistant actually performs.

The future of intelligent assistants

In many instances, the most important use cases for A.I. — augmenting human capabilities or performing tasks a person finds difficult or all-consuming– aren’t “sexy” (like shopping, travel, or dating), but they’re the types of activities necessary for business and government to function effectively, albeit hidden from public view.

A good example is using A.I. to analyze X-rays of cargo shipping containers to identify smuggled cars. A.I. is faster and more accurate than any human would be at the same task. And this technology enables cargo inspectors to spend more time on other important areas, like smuggling prevention and enforcement.

To use a personal example from AppZen, our A.I. for automated expense report auditing is able to verify data at rates and levels of accuracy that humans simply can’t match. Humans are still required for the inevitable “judgement calls” on flagged expenses, but now they can focus their time addressing the truly important items, instead of slogging through the vast majority of expenses that are of no concern.

Conclusion

We’re in the early stages in the rise of intelligent assistants. Tomorrow’s winners in this space will be the companies who prioritize utility over flashy gimmicks. By focusing on augmenting human capabilities, these companies will unlock the true value of intelligent assistants and provide benefits for both their customers and users.