Have you ever experienced a failing bot? You know, everything’s going well and then you just hit a big “Grumble. Ugh! Technology!” wall?
Though Bot1234 (pictured above) could be mistaken for an assistant who knows me too well, it is clearly not doing its job if I’m in a crunch. It is useless to me, and I am left fuming that I ever let my human assistant go while furiously punching numbers to call the company that made this horrible bot-monster, let me down, and wasted my precious time.
Dimon.co is here to make sure that this common problem happens less and less. While the bot economy is booming, growing even faster than apps did in their early days, a lot of companies are overlooking some huge risks in their eagerness to keep up with the trend. As they desperately try to match competitors, many companies are throwing bots together as quickly as possible and releasing them to the public far before they’re ready or have been adequately tested. To pull businesses out of this dangerous environment before frustrated customers–like angry me, above–come calling, Dimon has developed a technology to automate bot testing and increase your bot’s performance assurance. Using its own (assuredly well-tested and attentively developed) bot, Dimon generates thousands of unique scenarios and conversations to have with your company’s bot. The system then finds any and all bugs lurking in the bot’s code so your customers don’t have to.
Dimon’s co-founder and CEO Eran Bernstein was kind enough to answer a few of our questions this week all the way from Tel Aviv. Here’s a look at our conversation, getting to the bottom of bad bots and how businesses can fix them efficiently.
An Interview with Eran Bernstein
30SecondsToFly (30STF): So, to start things off, can you provide a brief description for readers of exactly what Dimon is and of the specific problem that it looks to solve?
Bernstein: Businesses take a risk when they make a bot, because when or if the bot fails, it is frustrating to users who then blame the organization who made the bot, not the bot itself. This affects the organization’s reputation. Dimon is helping bot owners increase business risks while helping validate bots through testing and ongoing production.
30STF: Interesting. Where did the idea for something like that come from?
Bernstein: Well actually, we, our two co-founders — me and Noam Shemesh, the CTO — didn’t have any idea at first. We started thinking about ideas and had a few in other industries, some others in the bot industry, but after researching and talking to people, and discovering issues here, we really got into the bot industry and found this very big, challenging opportunity in bot systems.
30STF: What would you say, exactly, that Dimon does or offers that chatbot owners can’t do on their own? What advantage do you have on them?
Bernstein: Good question. So, first of all, we are automating testing. We know how important it is to validate bots’ behavior. Our technology can take the basic bot behavior and a basic idea of what needs to be tested and from there, generate almost infinite related scenarios. Basically, we are validating your bot [assuring that bot functionality can meet users’ expectations] in the development cycle and monitoring it while on production to make sure that your bot behaves correctly, both of which reduce business risks.
30STF: Is Dimon targeting a particular type of chatbot or clientele, such as business-to-business (B2B) bots or business-to-consumer (B2C) bots?
Bernstein: No! Every chatbot, every bot that is using natural language processing — and even some that are not — we consider our potential clients. But, of course, we prefer to work with businesses that have bots like 30SecondsToFly. They are, from our perspective, bot-driven businesses. With them, we’re working with enterprises that have bots as their service providers, making bot performance assurance their highest priority.
30STF: What exactly does the program that you implement look like to your users? What platform is it accessible from, or is it its own type of software?
Bernstein: It’s a software service in the cloud. It’s very easy to use — drag and drop — and very simple to integrate. We have a special bot that can communicate with other bots. Our bot can validate yours anytime — if it’s on the developing side and on production, like a “human bot.”
30STF: And the Dimon bot can be integrated with bots in different chat environments like iMessage, Facebook Messenger, Slack, etc.? It works with all kinds?
Bernstein: We support all platforms. Facebook and Slack are a simple integration, and the other platforms we connect to using a very simple API.
30STF: Awesome! How are the tests executed exactly? Do you create a specific script for specific businesses?
Bernstein: We understand the basic bot behavior and we know to generate a lot of potential scenarios. This is our core technology, which uses deep learning, data sets, NLG (natural language generation), etc. Our platform utilizes a bot that communicates with your bot. It can validate your bot using all of the scenarios you as the user and we as Dimon build to see that your bot behaves properly.
30STF: So, then, after the Dimon bot has interacted with the client’s bot, do you follow up and monitor the bot’s interactions with human customers?
Bernstein: Yes, of course. We are doing an ongoing optimization to your bot and your bot’s behavior in order to help you build an unfailing bot.
30STF: And afterwards, do you at Dimon educate chatbot owners on how to fix issues that arise, or just notify them of their occurrence in places that they may not have detected?
Bernstein: We help them find bugs, etc., in the training process, and eventually we tell them what the issues that their bot mishandled are. We can tell them where it is in their code, but ultimately, they need to fix it alone. We are automating all of their testing, where right now they are doing it manually, which is infeasible to do. We accelerate the time-to-market process in order for the bot to go live quickly.
30STF: Have you found that there are any kinds of bugs or issues that come up frequently?
Bernstein: No, not really. Every one has its own issues. (Laughs)
30STF: (Laughs) Alright. Have you at least found any particular types of bugs perhaps that have come up in certain types of bots?
Bernstein: Well, I can tell you that most bots are failing. I like to call [what bot owners can do on their own] “guess-and-test.” Bot owners don’t really test their bots. If anything, they manually test it. Every version and every feature that is added may cause something else to crack. It’s issues on different integrations, different features, if it didn’t understand what the user meant…bots understand totally different issues while the user has another expectation. We know to fix it. There’s the famous “Sorry, I didn’t understand you,” or we see bots asking for a lot of information in cases when such information is not needed at all, and we help them to cut the flow with their users to make it shorter.
30STF: Maybe this will undercut your whole project (laughs), but do you have any basic tips for chatbot owners that are looking to avoid making these kinds of mistakes in development stages?
Bernstein: Yeah! Testing is important in every software development cycle. People need to take it seriously from the first day. They need to validate their bot in order to minimize business risks and failing bots. We at Dimon are here to help them make things that are infeasible right now possible.
A true friend to small businesses and their bots, Dimon is currently free for “small bot builders.” Eran Bernstein and his team are devoted to filling the already enormous bot community with the highest-performing bots that businesses can be proud to have representing them. Businesses with bots or chatbots in any stage of development can sign up here and see what Dimon has to offer.
Bernstein is also available by email ([email protected]) if you have any questions about Dimon or would like to know more!