- Company name: Enterprise Bot
- Years in business: Founded in October 2016
- Locations: 30 clients and employs 77 people split between Switzerland and India
- Number of employees: 77
- Principal management: Pranay Jain CEO, Pranay Jain is the co-founder and CEO of Enterprise Bot. Originally from Pune, he studied business at Bentley University in the United States and developed a strong interest in information technology during his internship. His work with AI caught British multinational Barclays’ attention, and the bank invited him and his team, Co-founder Ravina Mutha and Co-founder and CTO Sandeep Jayasankar to a hackathon in Mumbai which tasked them with creating an effective chatbot for the bank.
- Primary business opportunity (market pitch): We are in a world that is looking for faster resolution of problems and at the same time companies’ margins are getting squeezed, which is limiting their ability to scale up customer support. Enterprise Bot’s AI is able to understand user requests and complete them in a matter of seconds helping to improve the customer experience while also reducing operational costs.
- Primary business product (product positioning): Enterprise Bot provides a hyper-automation suite that is not only able to understand users requests like other chatbot companies but also complete the user requests by integrating into core systems like Salesforce, Servicenow, Workday etc.
- Revenue % growth over last year: 85%. The company is on a trajectory to generate close to $3 million in revenue in 2020, with over €4 million predicted for 2021.
- Primary business successes to date: Live with Afterpay with over 92% accuracy in multiple languages and countries handling over 25% of traffic without any assistance or escalation.
- Website: www.enterprisebot.ai
- Email: [email protected]
- Linkedin: https://www.linkedin.com/company/enterprise-bot
- Twitter: @enterprisebot
Chatbots have been around since 1964, when Joseph Weizenbaum introduced ELIZA, a computer program that mimicked a psychotherapist. You’d ask Eliza a question and it would respond with an open-ended response that invited further interaction.
Eliza was a sensation in its day, and as the decades passed, its legend grew and some cures were reported. (You can interact with Eliza on line to this day. She’s not very impressive when compared with today’s technology, but experience for yourself what the excitement was all about.
As far as the modern user is concerned, chatbots first made their professional debut in significant numbers at the beginning of the second decade of the 2000s. They usually appeared as small circles or rectangles that showed up in the left or right hand corner of your screen. They were sometimes called virtual agents, system agents, and other variants, but over time “chatbots” became the popular term.
Chatbots were heralded as a new way to connect with prospects, customers, and people seeking customer support. Early attempts to use chat systems for these purposes had proved somewhat disappointing. Chat did increase customer interaction and close rates, but it also tethered the operator to their computer, and often wasted much management time providing support (customer are notorious for not referring to FAQs if they can get their virtual hands on an actual person. To be fair, this is often the fault of providing company, as most FAQs are incomplete, poorly indexed, and offer at best 50/50 odds you’ll find the information you’re seeking).
Powered by “cutting edge” AI, chatbots promised an end to tethering and wasted time. Customers instead would be taken into expert guiding hands and quickly receive relevant answers to enquiries and solutions to problems. It all sounded very good, and soon chatbots were spreading across the internet at a rapid pace.
Unfortunately for the technology, as more and more people interacted with chatbots, they became as popular as a diagnosis of Wuhan at a nursing home. (I myself developed a deep loathing of the critters.) The bots were stupid, and their ability to provide useful answers no better than FAQs. Worse, they often provided incorrect information, drove people crazy with repetitive and often nonsensical requests for more input, insulted everyone’s intelligence by trying to pretend to be “human,” and as far as the U.S. market was concerned, churned out far too much fractured and incomprehensible English, probably inducing more people to punch out their screens than has been reported.
As fast as they appeared, the chatbots vanished. Larger companies in some cases soldiered on with the technology, but by 2016, chatbots were widely viewed with contempt and disdain, particularly in the SMB markets, who felt they’d been promised much with little delivered, a perceptions that exists to this day.
This has started to change. But the latest generation of bots, powered by more sophisticated AI, are making something of a comeback. In fact, according to some studies, a 180 degree turnaround has been achieved. Today, according to Smallbizgenious.com, the chatbots future is bright because:
- Chatbots can cut operational costs by up to 30%.
- Handle 85% of customer interaction without human agents by 2021.
- 50% of businesses plan to spend more on chatbots than on mobile apps.
- 64% of internet users say 24-hour service is the best feature of chatbots.
- 37% of people use a customer service bot to get a quick answer in an emergency.
- There were over 300,000 chatbots on Facebook in 2018.
Before going further, we’re going to say we suspect these rosy statistics are generated by the chatbot industry, and we feel safe in cutting these numbers down by 50%. Maybe 60%. Nonetheless the circles and the squares are reappearing on web pages at a speedy clip. What’s changed? And what are the lessons to be learned from the previous generation’s fiascos.
To find out more, I talked with Pranay Jain, CEO of Enterprise Bot, to find out the current state of chatbot technologies, costs, and best practices.
Softletter: Pranay, I have to be honest. I hate chatbots and avoid using them when possible. I regard them as the support service from hell.
Pranay. And you aren’t the only one. The first generation of the technology was over hyped and didn’t deliver. But our success is a strong testament to the fact that things are changing. In 2020 we generated close to $3M in revenue and are on track to double that in 2021. That growth speaks for itself.
Chatbot providers decided to stop treating the technology like it was a text-based adventure game and focus onmaking it useful. Nobody waits to talk to a chatbot. People need to solve a problem, such as withdraw money from ATM, block a compromised credit card, find out how to locate a plug-in for WordPress ecommerce system, not receive a useless stream of off topic data.
That means the data loaded into the system needs to be carefully audited, indexed, and pre-packaged. With the new AI platforms coming online, that’s becoming possible. Also, verticalization is also greatly improving the customer experience. The lesson that’s being learned is build a chatbot to a specific purpose and make sure the system tell what it can provide and what it can’t.
The original economic arguments for the technology haven’t changed; in fact, they’ve become more compelling. A cost of a service or technical call handled by hired or contract personnel averages between $1.50 to $10 call, higher in some cases. For repetitive questions, this can be reduced by 60% to 70%.
Can you be more specific about what you mean by repetitive queries?
Yes. This of course is dependent on an industry, For example, in the Netherlands, we work with an insurance company and has built a chatbot that enables a policy holder to change personal details online for their policy with the assistance of the bot. The bot doesn’t try to answer details about payouts and annuities and related issues. Invoicing is another area we’ve discovered where chatbot technology. works well. Questions about invoices tend to be highly repetitve and are easily classified. In fact, looked across a broad spectrum of industry’s 65% of all questions coming into a company are repetitive.
Are there areas of discussion and enquiry you recommend not be used with chatbots?
Yes. Stay away from classes of questions that deal with:
- Rejection of requests for financial assistance or renumeration.
- Loans and loan rejection.
- High value transactions and luxury goods. It’s OK to begin a sales cycle on a potential sale of >$30K, but that whole cycle needs to be moved out of the chatbot domain once there’s been a preliminary confirmation that this indeed a prospect.
- Mental health issues and discussions. Actually, chatbots can be effective in certain circumstances, but there are too many areas where problems can arise. A person needs to be on the phone with someone having an emotional crisis.
How do you avoid the uncanny valley, the uneasy feeling many AIs generate that results that can result in people disliking chatbots?
First, the chatbot shouldn’t be too “chatty.” Too many early bots tried to hard to be personal and “familiar.” In fact, they were too familiar, if you take my meaning. Make it clear upfront the customer or prospect is dealing with a chatbot, not a person. At the same time, avoid being disconcerting. Slow the bot down so it doesn’t answer too quickly; that’s off putting. Longer responses to questions should be staged and layered to create a conversational tone, Avoid overly long sentences. And make sure your bot is customized for local idiom and spelling; people hate dealing with bots speaking a fractured version of their native language.
What about voice bots? I’ve seen some example of them action and I’m wondering about their future.
Text bots now represent 80% of our revenue. We are seeing growing demand for voice. Text is how started, tex is 80%, By 20221, we anticipate 40% of our revenue will come from voice.
BTW, have you used Siri? Or Alexa. If so, you’re using voicebot technology.
Yes, I have. My daughter has one in her kitchen. When I’m there and helping out, I’ll use it to set timers.
It’s a start. But all joking aside, chatbots are insinuating themselves into all our lives at different levels of engagement.
What were the main challenges you faced in establishing and growing your company?
The biggest challenge we faced in 2014 through 2018 was proving the technology worked. As you pointed out, the market was flooded was skeptics. And people like you, who’d decided they couldn’t stand chatbots.
But the single greatest barrier to acceptance is integrating with legacy systems. For example, an insurance company customer’s data may secured in five different places. But a chatbot user doesn’t care about that, and won’t stand for multiple questions trying to establish to “whom” they’re talking to. You don’t know me? Despite the fact you’ve required me to fill out multiple forms digging into my personal and professional fundament?
Let’s assume a company wants to implement a chatbot system. What will it cost? How long to implement?
Let’s break implementations into basic classes. First, Chatbot One. It’s dedicated to customer support. The cost will range from from $60K to $100K. Implementation must include:
- A detailed schema of what the chatbot will address.
- Provisioning the dataset.
- Integration of other stored instance of customer data into a unified source.
Chatbot Two implements a basic FAQ. The cost will be about half of the above.
They seem costly.
No. Spreadsheet analyses demonstrate a positive ROI. But this takes us back to mistakes made with early chatbots. They are not sentient. They don’t know what they’re talking about. They’re mecha. Within the scope of their defined limits, they can be very useful, but that scope must be defined by human research. Over time, within these defined limits, you will be able to stack layers of knowledge onto previous frameworks But those limits are defined by human knowledge and intuition. And that costs money. Don’t expect to sit down with a group of employees and throw a bunch of great suggestions into the dataset. That’s been tried. You didn’t like the results.
Chatbot One requires far more preparatory work than Two. But as you’ve already noted, most FAQs are terrible. Companies don’t bother to improve them and are often clueless about how they destroy sales. But build an effective chatbot on top of them and you’ll quickly understand why they don’t work well and how to improve them.
Your profile says you split operations between Switzerland and India. That’s a bit unusual. Normally, for international firms, I see the U.K. Sweden, Israel. sometimes France and Germany. What attracted you to Switzerland?
We were going to establish an HQ in the U.K. but the Swiss Stock Exchange, impressed by our work in India, invited us to establish a startup in their country. Switzerland at the time was not making much of attempt to attract startups but that’s starting to change. It easier to incorporate, attract funding from VCs and angels, and obtain visas The two hottest sectors are biotech and crypto. But plan on spending at $20K to get going.
Now, could I ask you a question?
Do you still hate chatbots?
Uh, yes, But I have avoided them whenever I can and am not up to speed on the latest iterations. I am installing a chat system on the Softletter site to see how the technology has changed and improved; it’s certainly more affordable for smaller operations than a decade ago. But that’s not a bot; that’s me.
I recently did attend a seminar for a company selling chatbots for smaller firms, but setting up the queries, the answer trees, and populating the dataset is not a trivial exercise. If you attempt to do it yourself, you’re going to spend thousands of dollars, if not out of pocket, then in time. I still think the enterprise is where it’s at for the current technology. You’re proof of that.