Podcast | Orchestrating get-to-know sessions with customers – right from Salesforce!

Jean
Shin

Director, Strategy & Content | Podcast Host of Mobile Interactions Now

21 min Podcast
Orchestrating get-to-know sessions with customers – before the service transactions!

In this Part 2 of our conversation with Tim Schuitemaker, CIO of Gen25, an Amsterdam-based system integration company, we delve deeper into how to enable companies to build better relationships with new customers, even before their first official sales transactions. Tim will help us get started with CRM tools and communication channels that are already available today – as well as what’s right around the corner.

Jean:

Tim, welcome back to the show. Now, in this episode, as much as you will allow me to, I would love you to take us to the backend and talk about some of the technological challenges that we still need to solve and some of the ways how you are solving some of the problems for your customers. I want to quickly revisit some of those frontline changes happening on the customer side, for example, a lot of sales representatives who used to be at the store are now with their phones at home, and making them connect and still maintain certain relationships with their clients and even carry on transactions, that would take some work, front and backend.
 

Tim:

Yeah. So a couple of things spring to mind. So maybe what you see indeed is that a large part of the workforce is sitting at home due to COVID. They're still part of your company and you can utilize them in a way by providing them access to this customer data as well, and giving them the option to interact with their customers. So that means that video calling and using WhatsApp as an upsell channel or other social channels I should say is really something that you can use these people that used to be in this retail brick and mortar stores that are now no longer are available. You can really use these people to actually interact still with your customers. And we've seen that happening with a couple of our customers as well, where we made sure that online on their website, they can make an appointment for a virtual stylist call and they made an appointment that was put to one of the employees.

And then instead of going to an actual store, you hop on a video call and they could present certain products that they think would suit you. So they ask you a survey upfront saying what kind of colors do you like? What kind of products are you looking for? They ask all kinds of questions up front. And based on that, a stylist or one of the store employees would actually create something like a look for this person. And that look would be done at that point, shown during this video to the customer, which of course allows you to brainstorm together what your customer is saying. I like this, but I want it in black, or I like this, just like you're doing in the store normally I would say.

And in the end allow you to do an order on behalf of that customer. So you send them a URL or a link or an email at the end of the conversation saying, "Okay, we've discussed this. These are the products that we looked at and you said you were interested." That means that you still are enabling your workforce to actually do their job while they're staying at home.


Jean:

I love that example because I couldn't get to my hairstylist and I was having problems. But I didn't know such options existed.


Tim:

Well, official haircut is still very hard, I would say.

So these surveys that we send out upfront are actually is sort of a bot because they can just go through a flow asking certain questions like color, like style, like those kind of things, which allow this company to create some kind of profile on your answers. And of course should be included if you interact with something that the bot doesn't understand, it should you go to this agent or this employee and allow you to interact with someone in person. On top of that, these appointments can be confirmed as well by the social channels that you're using. So they can send you an invite or a URL saying, "Okay, we have a call tomorrow. Please use this URL." All of this can be automated. So it's only a matter of using the actual conversation that you're having. You only have to prepare and that you don't have to do the whole thing up front, which usually you would do in a store. You would ask someone what colors do you like, which is really handy for marketing again, because they know what you like.


Jean:

I would say they should look at my hairstyle requests as a general sentiment analysis. And I think sometimes that will paint a bigger picture about the whole mood that some people are in. The upfront experience that you're describing, now can you tell me what you had to do in the backend to make that work?


Tim:

Well, actually, not that much. This is an interaction between a couple of our products, and we've been talking about our social product up until now. But creating appointments is a product that we offer on Salesforce as well. Of course, the customer has to do something on their website to actually include a form or something where you can make this appointment. Nowadays, most systems are low in code. They allow you to create processes and they allow you to do low to no code solutions on notifying people. We've been talking about Salesforce in the last podcast, but indeed that allows you to do these kind of things as well, notify someone that their appointment has been made. And of course, trigger these events to marketing campaigns saying, "Okay, this appointment has been created. We can now send them his confirmation," or those kind of things.


Jean:

Keep going with the hairstyle example, let's say there are some people in the mood for color red. And having that kind of general trend, that kind of data to, to anticipate, are there any easier way to combine this data to inform the choices?


Tim:

Well, of course, with a bot, you sort of predefine the choices that someone might have. But you could also ask, and that's part of our solution as well is see if there is any intent in there. So if they say hair coloring, then we can pull that from a piece of text saying, okay, hair coloring is included in this piece of text and use that as a tag on the conversation. So we pull the intent from the conversation and based on that, we can do all kinds of things. We can include more AI [inaudible 00:07:13]. because you used the word red, we're now redirecting you to that hair coloring section of the box, or need to forward it to an employee or whatever. It really allows you to do more advanced stuff with this intense analysis. Also, we do sentiment, you briefly mentioned it, but especially if people become angry, then it might be a good time to redirect the conversation to someone who has a lot of experience dealing with angry people.


Jean:

I want to pick up from what you mentioned about the human handover moment. Technology-wise, are we going to get to a point where this isn't really needed? Or is this something that we just have to embrace and try to be more efficient with?


Tim:

So I think from a technology point of view, we're sort of already there. So what you see happening is that the conversations are in a certain state. They're either in we're now having a conversation with a bot, we're talking to someone, we're open for marketing, these kind of different states. And these states, they can be switched on and off. In our software, they can be switched on and off by an employee saying, "Okay, I've done my conversation. So now it's back to the bot to take over," which is the handover.

Of course, it's important that if someone does that, that you can see what they have interacted with the bot. You want to see the conversation that they had with the bot and where it went wrong I would say. It shouldn't be something in the system. It should come into the system, but not actually trigger a conversation. But you want to know what someone answered and what kind of, what we said, what kind of hairstyle they like. As soon as someone is not actively seeking for the personal attention, then you're okay by just keeping that in the bot state as long as possible.

So maybe a good... We have one reference case that decided to replace part of the frequently asked questions with a bot and people interact with their bot. And they found out that 15% of all questions directed to this channel were, so one five, were handled by the bot instead of an actual agent or an employee. This allows them to have more time, more efficiency. And they calculated that it was up to 70% efficiency that they reached by implementing a small bot to just check the FAQ answers and respond based on the intent of the question with the FAQ answer. So that's a huge, huge improvement for the service desk because the service desk and these people can now work on other stuff at that point.

Of course, that is a big promotion to bots in general, as well. Maybe another example or use case that we've seen with one of our customers is that they not only using AI and bots to automatically respond, but also co-piloting an employee through the conversation. So it is still a conversation that you have with an actual employee, but as soon as they open up the screen, to answer that they should give is already in the text area and they just have to press enter if they agree. But they still have the option to interact with that text. So they can still say, "Okay, this is not actually what I wanted to say to this customer at this point." But you see that over time, these things learn, as you just mentioned, and the software predicts what it has to answer to the customer.

And at the end, it's just a matter of getting a certain level, making sure that the bot answers X amount of questions, X percentage of questions right the first time. And enabling that little answer to be automated, which allow you to grow in the approach on providing more automated flows into your day-to-day work, increasing your efficiency as an employee.


Jean:

I'm going to challenge you to be more technical there. So when you are talking about having some kind of a predictive engine going in the background, pushing some of the questions and helping co-piloting, because that tells me about where the future is going. Something that's being tested. Tell me a little bit more about how the ultimate picture looks like, if that works out, if the co-piloting is an interim state, tell us a little bit about what's happening in the background to make these things more mainstream.


Tim:

It all starts with the intent. So based on the intent, we see what kind of questions the customer is asking, because nobody uses the exact same questions. They're using different phrasing of the same questions of course. And usually there's a database of having answers to certain questions there, which is triggered based on this intent. So this database is constantly adjusted by someone that is responsible for the content giving the correct answer. And based on the intent, we can add different types of intents to certain answers. And at a certain level, you see how many times people use that message over typing something themselves, which increases the confidence level of that message. And if the confidence of that message is increased, then at a certain point, you can decide as a business that if we see that 90% of the time people are using this message [inaudible 00:12:44], but they don't interact with it at all, they just press enter to send it, that means that we can then go from co-piloting to actually piloting the conversation.

And therefore you build up a confidence level over these messages grow over time. And not only don't have to go full blown out with a bot at the first try, but can grow a bot over time. So it's sort of the bot, it takes over the conversation as much as possible.


Jean:

For previous episodes with another guest, I was talking with a chatbot designer, and talked about how my dissatisfaction with a certain bot in terms of how it answers my first question. After that, the followup questions are usually met with, I don't understand, would you like to talk to somebody else, or that kind of thing. And I'm just not liking it. And why is this so hard to get the followup questions, two, three questions answered versus the first question? As a developer, his answer was, "Well, right now we're good at getting the intent, first intent somehow guided or have the predictive quality to it and address the first intent. But the intent within a conversation, once the conversation starts, that's where it gets tricky." Are we going to get better at this? I would like to see it get better.


Tim:

Maybe it's good to see how we do it because intent, indeed, usually it's determined in the first couple of messages that someone interacts with you in a conversation, because if you're talking to a company, usually you go straight to the point. You're not going to ask them about the weather and everything. You're going into your point. I haven't received my package, [inaudible 00:00:14:31]. So it's important that the intent not just is determined for a single message, but for a range of messages. And that's what we're trying to do. So we always say if you look at the first 10 messages, it should determine the complete intent of this conversation. Well, if you look at only one of these 10 messages, that might have a different intent itself. So you want to sort of combine different flows in your bot to combine it into a single message that you then pull the intent from.

You see that these bots are just interacting with services that are there. And one of these services can't give me the intent of this text. And what people tend to do is during this conversation in a bot. They say, "Okay, I have received the new message. And now I'm going to fetch the intent." But actually that's not really, what I just said, it's not important to have the intent of that whole message. It's important to have the intent of the whole conversation. And that's where you have to group these messages once again, at the end of the conversation and say, "Okay, the intent of this or the sentiment of this whole conversation was X, Y or zed."


Jean:

I do want to know what you're currently working on that’s really exciting you and in terms of what you can do with the interactions between companies and buyers out there.


Tim:

Yeah. So indeed for me, it's really exciting to see the social channels as being part of a complete flow. So the flow I've previously described where people create appointments online and then go into video calling and having conversations via social channels in between to keep up to date with each other as maybe the stylist and the consumer. That's something that gets me really excited because I think that that essentially is the goal where a company should strive for being more customer focused and therefore being more on the channel where the customers are, provide them with tools that they want to use instead of pushing your technology into how they should use it. So indeed be on the social channels that your customer are. If you have a specific market, make sure that you've targeted that market.

And maybe a good use case there is one of our customers is FrieslandCampina. They're the largest dairy cooperative in the world. So they have various brands and we provide them with tooling ways they can use to both service and use marketing. So they have agents in multitudes of brands in Indonesia or in Malaysia and Philippines and Hong Kong and Vietnam. And for all of these different countries, they are using different social channels because in each one of these countries, the main search channels is a different one. And we provide them with one tool in their Salesforce environment which they can use to actually communicate through these different channels to their customers based on their different brands in different countries that they're in.

What I also find really cool there is that there's multiple languages being spoken and we determine what languages. So we do language recognition there. And based on that, we indicate that this conversation should be handled by this person because this person speaks that language. Yeah, that's a really cool use case that really brought the communication as part of their brand identities. And with that, they also allow you to interact more closely with their brands.

If you look into enterprise customers, of course, there are a lot of people interacting with their Facebook page and with their social channels already. But if you're a smaller company or maybe not even that small, but if it's not a channel that you're actively promoting, the volumes usually are pretty okay for you to deal with, especially if you have a contact center in place or a sales center in place. We shouldn't forget that this is also a sales tool. We've mentioned a lot of customer service here, but it's indeed also in the end a sales tool to interact with potential leads and customers.

So it's important to go where our customers want to go. They do certain questions. They ask us certain questions and ask us if we can be available in certain channels, or if we can extend our service to certain... For example, this language recognition is something that really was influenced by our customers to put it in there. And you see that our roadmap is really customer driven and new channels are constantly onboarded based on where our customers are I would say. So we recently had a customer saying, "Okay, I want to target Russia now with are tooling and fiber is really popular in Russia. Can you please enable fiber?" So we're working on that currently.

A other thing we're working on is RCS because we really see that as the future as well, the new version of [inaudible 00:19:02]. And that is one of these channels where we go much faster than I would say building something yourself as a company of going when a big software provider that has this out of the box, because we can always help our customers in finding the correct way to do things in adding these new channels and adding new functionality to the platform, to the application that we are building, to make sure that this application is future proof for everyone, for us and for them.


Jean:

That was awesome because it really puts the responsibility to solution providers like us to really enable whatever is needed.

Now, just so that we get to know you better, can you tell us, Tim, what are the three things that you do most on your phone? You can name apps if you want.


Tim:

Good question. But I use a lot of Reddit. So I read a lot of Reddit and follow a lot of different subreddits on various topics, including Salesforce of course. I'm also still a geek at heart so I play a lot of Mario Kart on my mobile phone. I'm almost embarrassed to say it, but it's true. I play a lot of Mario Kart. And of course, for the rest of the apps that I use more commonly as think is what everybody mentioned is LinkedIn and Gmail and interacting with customers and employees and the company in itself.


Jean:

Lovely. Absolutely. Now I feel like I have to leave you to go back to your Reddit and reading. We are recording this on a Friday, so, have a wonderful Friday and weekend. And I thank you for talking with us.


Tim:

Thank you for having me. It was a joy. Thank you.