How Machine Learning Can Empower Customer Success Managers

There’s a new buzzword (or phrase really) humming in everyone’s ears...machine learning. We’ve all heard about it or read some articles on what machine learning is trying to accomplish and how it will start to replace the work of humans, as well as how it can help us work smarter and better.

And this is an exciting time for those of us working in the SaaS industry to see the evolution of machine learning. But there is also some uncertainty around it; what is machine learning capable of today and what do we hope to be able to accomplish with it in the future? What will it’s role be in our day to day work lives...and regular lives, for that matter?

I wanted to learn more about machine learning and AI, and some of the implications it has for SaaS companies, customer success, service, and support roles today and what could be to come in the near future. So I reached out to Cristina Fonseca, the co-founder of Talkdesk, technology entrepreneur, engineer, and machine learning and AI enthusiast. She’s currently left the day-to-day operations of Talkdesk to study machine learning and AI and how it will impact businesses.

She has a fantastic way of explaining what machine learning is and how it can be used to empower, not replace humans. And how in customer success, service and support roles the human will never be able to be replaced, but can be assisted by machines and AI to do their jobs better.

Meg: Thank you so much for joining me, Cristina. So before we get started into the topic of machine learning, SaaS and customer success I’d just like to get a little bit of your background - how you started your career and how you got where you are now.

Cristina: So my career...I don’t know if I have a career at this point [haha]..When I left university in 2010 I was not really happy with my professional options at the time and I think the Social Network movie was kind of famous and had come out at that time...and me and my co-founder, Tiago who studied with me, we thought, what if we programmed something, put it on the internet and become Mark Zuckerberg? Which was the good person to be at the time, not it’s a bit more challenging [haha].

So that was basically the starting point, for us basically to not accept any job offer we had and to start testing applications on the internet. Of course, that was not an easy path. One year we launched several online applications...nothing was working because it’s not just build something and people will come...there’s a lot more things you need to take into account.

And after 3 failed years of projects, we participated in a contest by Twilio and we won - then we were invited to present the idea in San Francisco at the first Twilio conference. And the first motivation was to just win a MacBook Air. That was the prize and we said well we are tired, we’re desperate let’s at least get the computer…

And [there is where] we developed the first Talkdesk prototype - which was a way for companies using all the systems in the cloud to have phones in the cloud as well. Because you had your billing system in the cloud, you had Zendesk, you had email in the cloud, you had your CRM (Customer Relationship Management) in the cloud but everytime someone would speak to you, if you would take a couple of minutes to figure out: is this person in my database, is this person paying, what’s the plan. So it was possible but time consuming. We could do it manually, but there was nothing aggregating the information about a customer, the history that person has with your company, and then give you that information in real-time especially on the phone. So we built the first prototype of Talkdesk, where you know everything about the person who is calling you and you can pick up in the browser.

We went to San Francisco and we won the Twilio competition, then we got into 500 Startups and here we are. Like a couple of years later we raised around $25 million and the company is now more than 300 people, between Portugal, where me and Thiago are from...but the company is based between Portugal and San Francisco.

Meg: That’s an amazing story and I like that the motivational starting point was a MacBook..

Cristina: Yeah that’s always the funny part! Because there’s [often] so much planning put into starting the company, developing this big idea...usually it’s not, well I just wanted to win the computer.

So I left the daily operations of [Talkdesk] in 2016 and since then I’ve been studying at Singularity University which is this crazy university where you learn about the world’s biggest problems, the latest technologies, and you combine both. Which led me to the artificial intelligence journey.

So first I thought everyone talks about it, but no one really understands what AI and machine learning are and what they can do. So I became very obsessed in understanding where the limit is today.

And I was working with an engineer building something that I thought was very very simple to build, and I was frustrated because he couldn’t make it work, and he was frustrated because things are not that simple. So it resulted in a lot of frustration and I thought, ok if I want to be successful in the next decade I need to know what are the limitations in the technology right now, what’s possible, what’s not possible, and get behind the hype that’s going on. That’s why I became interested in the AI space.

Meg: So for our listeners, do you have a quick way of explain: What is machine learning and AI, and what are some of the differences?

Cristina: So basically machine learning is having computers learn from data in order to perform tasks that humans would do. I can also tell you what machine learning is not...which is magic. There’s no way you’re going to give a computer or an algorithm a bunch of data and magic will come out of that. The steps to make this work are tough.

One, you need to have a well defined task to perform. I mean I can teach a computer how to recognize if a picture is a dog or a cat, but first I need to teach a computer what a dog is and what a cat is. So that includes getting enough data to tell the computer what that is.

And in the process of learning, a computer needs a data set that's prepared...has no noise...which is very very difficult to get...If you speak with companies and with people dealing with training algorithms this is probably the biggest challenge - get data that’s good for the computers to learn. And after that the computers can learn.

The learning process for humans, it also takes a while. When we’re kids you know we try to perform certain tasks...identify objects, or understand’s not at first. And we need to get several examples within a particular context to get things right. So computers are the same.

So machine learning is basically having computers learn from data, you need to have data that’s good to teach the computer and then the computers can replace humans. And this works in very well defined tasks. I think...there’s a lot of talk about general AI, which means computers are going to be more intelligent than humans...yes we’ll get there, but it’s going to take a lot of years. Right now what we can do is this narrow AI - which is artificial intelligence for very specific tasks.

Megan: That was a really good explanation, thank you. And it’s a good’s almost like a child right now. And it’s so funny because it seems so in-human, but you’ve just made it very human which is nice as well.

You have a really unique perspective, I think, from both building a company and then also coming from this programming how do you think that machine learning can help companies become more effective and maybe even more customer focused?

Cristina: Machine learning is usually not associated with being personal and I think what defines companies today and what makes them different is the ability for them to be personal to customers.

Especially in the customer success role it’s all about having someone at the other side to take care of you and your issues. I can start by just commenting a little bit on bots, which is what everyone thinks [when they think machine learning] - at least it’s within the use cases…and bots can help a little bit but are not the solution.

What bots can do and what machine learning can do in this context, I think, is basically enable the people within organizations to have better information about their customers, so that the interaction is not stupid or irrelevant, or you’re not giving something new to the person you’re taking care of. So I think what machine learning can do for companies is help them be more efficient.

If there are things companies are doing manually that’s going to stop. Honestly, the companies that are going to win in the next decade are the ones that can really automate what can be automated. Manual work is going to be automated within companies and machine learning is going to be a great part of that.

And then there’s a couple of other topics that will benefit from machine learning...[like] trying to understand if a customer is going to churn, that can be automatic. Trying to understand the best practices, why some customers are more successful than others - machine learning can help you with that. Also understand and figure out engagement issues.

But what machine learning doesn’t do, is take care of the personal relationship.

And it makes sense to compare this machine learning with the beginning of SaaS. I remember when we started developing Talkdesk - we thought SaaS was this magical thing where we built something, put it on the internet, free trial, you put the credit card in, and customer just start coming in. That doesn’t happen that’s not real...SaaS doesn’t work like that. And machine learning is not magic, in the same way.

Machine learning is not going to solve your problems, but it can be an enabler for you to better serve your customers. Because it’s going to get you the information you need, when you need it, it allows you to be proactive instead of being reactive.

Megan: Would you be comfortable digging into a little bit of how machine learning can assist with predicting churn and some engagement issues?

Cristina:...for example, during the onboarding process, software adoption is a big issue. So we have software to fix pretty much every problem within businesses, but the adoption is the tricky part.

Sometimes people sign up for your service and then they’re lazy, or they don’t have the time...or they believe things are more difficult. But maybe they are one step away from solving a real issue, or really creating value for their business. So what I think, on the engagement part, machine learning can understand [where] customers are getting stuck...especially if you have a lot of them.

It can understand the changes you need to perform to make them more successful during them during the implementation process. It can tell you if a customer performs these 3 steps then there’s a high likelihood of this customer being a paying customer. [For example], if I look at the data it tells me how a company and a set of users use a specific product, and understand who are the ones who are going to be our users most or the ones that become champions for my product...I can try to influence other people to perform those same actions in order to spread the word about what I’m doing. So that’s just part of it.

Then there’s the churn, of course. If you notice that a customer stops coming online from time to time, or there’s specific patterns that have been identified by a model that has been trained with your data then you can pay attention, and be proactive and speak to that customer and kind of reactivate the relationship if needed.

Of course I’m giving you some generics but this needs to be different depending on the business...but there’s a lot that can be done with historical data. And there’s models already trained for that. And honestly, in a couple of years, all of these will be embedded into every software.

And then there’s advanced stuff like education - what if we could generate videos, personalized for you that would guide you through the onboarding process. Another area [where] there are some companies appearing in that field [of machine learning education] self-service. Companies have bots to help them solve customer issues, but the self-service solutions are still not there. So most of the time, customers would like to help themselves but the information is not organized or it’s not structured or sometimes it’s not there. So there’s a lot of ways, I think, machine learning can help companies.

Megan: So you’re saying that there’s a bot, that people are already working on, that will take what you’re asking for or what you need help with and create a video for you based on that?

Cristina: Exactly! So I think in the next couple of years things that we don’t imagine today are going to happen. That’s kind of disrupting the customer service rep a little bit, but then of course we’re going to argue, in a lot of situations, I want to speak to the human. So we cannot entirely replace people, but we can empower them to have the right information to better assist their customers.

Megan: Sometimes you’ll still need that support person, like you said. But to be able to automate as much as you can, that’s going to help the customer success managers focus their time better, support members manage their time better…

Cristina:...So there’s customer support, which are the reactive people, and customer success are the proactive I think what machine learning can do is help you as a customer success rep be proactive when you need to be proactive and not when the appointment that’s on your calendar beeps that you have the monthly call with the customer, just because that’s the way it’s done. I think it can help people really understand who needs help, who doesn’t really need help...which sounds kind of like magic...but it’s not magic!

Megan: But it’s learning! Like a machine starts to learn ok this is a customer you actually need to have a call with them, these are the things you should talk about based off of their usage…

Cristina: Yeah! What about something that sits on your phone calls that understands the conversation and gives you the summary of the call and actions steps. Like, Ok you promised to send an email about this, you promised to follow up in 2 weeks, this is what you need to do. We’re not far from having software to do this.

It’s basically allowing everyone within organizations to be more capable of doing their job well.

Megan: That’ll be cool! Are there any circumstances where you think machine learning isn’t beneficial, or maybe where it could give the incorrect impression?

Cristina: I think right now the only thing I can think of, is when companies try to replace humans with AI (Artificial Intelligence) without fully understanding what AI can do and not do. I mean, if you put a [only] bot speaking with your customers it’s going to be a disaster.

The only thing I can think of is replacing humans entirely is not yet possible - bots are the best example and this is because general AI does not yet exist. So [with bots]...if you’re trying to order pizza, if you talk about the weather it’s not going to tell you anything special. And humans are very unpredictable.

Megan: That’s true - it’s hard to have something predicting when humans are the most unpredictable...usually! We do have some very predictable behaviors though.

Cristina: And honestly, I had a couple interactions with people in different companies and if they figure out you’re trying to replace the work by the bot they’ll create a lot of resistance. So they have these incentives to try to convince you that the technology you’re trying to sell the company doesn’t really work. Because they don’t want their job to be disrupted.

Companies need to be very careful in this transition period and work with companies to show that [machine learning is an] enabler of better work, more efficient work. Making people be free at 4pm to go home, and spend time with their families, and there’s a great opportunity for that [with AI and machine learning]!

Megan: You’re right. And I like the way you put it where it’s not replacing you its enabling you. And communicating that message rather than one of anxiety.

So, do you have a definition of Customer Success?

Cristina: I don’t know if I have a definition of Customer Success, but for me, and as I mentioned, the customer success [people] are the people that can have amazing conversations about pretty much everything and that are proactive.

So are these champions? Because honestly, sometimes customers just need to be heard and feel there’s someone at the other side. And customer success managers are the champions that can do that very very well. So that’s my definition that I just came up with.

Megan: So I hope that there’s more and more women in the tech scene going forward, and I think we’ve had a great year so far for women. What are your hopes for women in technology going forward? What do you hope to see?

Cristina: I think we all hope to see more women in tech, and I think that right now my biggest hope is, if we could solve the final problem that exists...I think we had a role model problem before. I think right now, with everything that’s been going on, that is being solved. Everyone is aware of the issue, that’s a very good first step for that issue to be solved. But there’s still not enough people getting into tech. So it’s a funnel issue.

We need to have more girls to choose those careers in high schools and that’s the problem. And because it’s a long term problem - sometimes everyone is looking for instant fixes for our issues - but I think that it needs to be a more long term strategy to solve the root cause of the issue. At least that’s what I hope.

And then it’s also a generation problem, it’s also like when parents say to their children well girls toys and boys toys are different, this needs to change. It’s not right to bias our child like that. Which is also part of the problem, but our generation people are learning and hopefully in the next couple of years things are going to change.

Megan: I agree - I think the next generation is very open and willing to change that. And it’s inspiring because you’re seeing more and more women in those higher positions and to have more and more women to look up to for inspiration is going to be nice.

Cristina: Yeah, but the good thing about this whole internet and information era is that- so before, when choosing the career you wanted to have, you’d look around you, your family, your close circle so you’re very very limited by the choices....but now we have access to all the social networks, and people can have role models that are very distant and I think that’s going to help a lot.

Subscribe to our podcast for more
On iTunes or Soundcloud
Get started today

Drop your email and let us show you how it works