In 2015, Digit founder Ethan Bloch and his team were excited about enabling anyone to build a financial cushion, and they were confident that they knew the perfect way to build a product to let them do that. Digit was, and is, an automated savings service that you hook up to your checking account. It detects when you can part with extra money and squirrels away $0.50 or $20 at a time in a dedicated savings account. It has been glowingly reviewed and the company has raised more than $36 million in capital.

Bloch wasn't going to build another run-of-the-mill app full of buttons and sub-menus. Instead, Digit would be a friendly helper that you could talk to, the way you would text or IM one of your friends.

At the time, so-called chatbots like this were the hot new thing. Silicon Valley pulsed with excitement over the idea of talking to semi-smart computers, which would help you do anything you could possibly want to do. Slack was going through exponential growth and launched a fund for investing in bots. Betaworks did likewise. Hashtag inventor Chris Messina even claimed that "conversational commerce" was the future.

Digit's team was completely on board. Chatbots were "exploding the ecosystem and huge companies were investing money," Bloch said in a phone call. The hype was contagious.

In a way, Messina et al. weren't wrong: People use the Amazon Echo or Google Home to set kitchen timers and listen to music. But the chatbots that you contact with your thumbs have definitely not lived up to the startup world's expectations circa 2015 and 2016.

Even companies that still rely on bots aren't as bullish on them as the tech commentariat used to be. Helpshift offers customer support chatbots, but the company has found in surveys that people will only tolerate bots when they speed up the communication process, which is a tall order.

"The experience has been poor because most of the bots that have launched [and then] failed rely entirely on natural language processing and natural language understanding," Helpshift chief strategy officer Abinash Tripathy explained. Helpshift's own bots are based on decision trees.

"Human beings don't trust that experience too much because the margin of error is really high," Tripathy said. It turns out that computers aren't yet good enough at understanding humans for chatbots to meet the potential that investors and founders thought they had.

Consequently, former chatbot champion Bloch has had a change of heart. "Chatbots are dead," he tells Inc. "I'm not even sure if we can say 'chatbots are dead,' because I don't even know if they were ever alive." After all, he said, no one can point to a chatbot that "all your friends were using." Such a thing simply never existed. (Remember Peach? Probably not. Case in point.)

Until now, despite the app and Web interface, the primary way of interacting with Digit has been through chat. You can chat in the app, or you can chat via text. Users tell the bot what they want -- to set a goal, for example -- and then go through a tedious question-and-answer process to make it happen. Now that Bloch has decided that chatbots are dead, the product has to be transformed.

Bloch's chatbots-are-dead epiphany came last year. In a move that he now calls a "terrible decision," Bloch decided that Digit should push even further into chatbot territory. Digit removed the balance number that was displayed at the top of the app, forcing users to make a query to find out how much money they had. To Bloch's dismayed surprise, users absolutely hated this. "It broke the camel's back for us internally on the product team. We really started to deeply question, why were we so convicted that chat was the interface?"

He likens the chatbot interface to the DOS command line that was the point of entry for PC users in the 1980s. No one who's not a programmer interacts with a computer that way anymore, for good reason: "We're using these beautifully intuitive devices that are mostly glass and we're tapping the shit out of them." The graphical user interface won, especially for consumer products. "It just wasn't clear to us back in the early days of Digit," Bloch says.

In the coming months, Digit will look different, Bloch says. The app won't get rid of SMS and other messaging options altogether, "but it's not messaging first." You'll be able to adjust your savings goals by tapping on buttons instead of laboriously giving instructions to an admittedly charming but mostly clueless bot. "We've already cut the usage of messaging by half" among users who've tried the new interface, Bloch says. "And we're seeing way more depth of engagement" across other features.

That Digit has been able to survive and thrive -- it has more than tripled its user base over the course of 2016 -- with an interface that the CEO now thinks is completely inadequate speaks to a healthy demand for automated savings. "When you have strong product-market fit, you can make a ton of mistakes," Bloch says. "A lot."

In a way, discovering that everything needed to change validated Digit as a product: Bloch recalled telling his team, "We've had such an ineffective interface, but you know what, it didn't matter, because the savings was so good. People dealt with it." That's the ultimate compliment from users: They're willing to put up with your flaws because your strengths are so very strong.

Correction: The article originally stated that Digit more than tripled its user base since 2016, but that growth actually took place during 2016. We regret the error.