Clicks Don’t Teach. Questions Do.

Julia Nimchinski:

And joining us next, Richard Dumas, co-founder and CMO at Slate CX. Welcome back, Richard, how have you been?

Richard Dumas:

I’m good. How are you, Julia? excited.

Julia Nimchinski:

Got it. Yeah.

Richard Dumas:

Okay, good. Another good session today.

Julia Nimchinski:

Yeah.

Richard Dumas:

I, I guess I have the privilege, if people are on the West Coast, of, presenting during your lunchtime. So I hope everyone gets a sandwich or something to drink and enjoys the session. Awesome, that’s good.

Julia Nimchinski:

Granted.

Richard Dumas:

Let me get into it. All right. So, let me just… Make sure you can see my deck.

Julia Nimchinski:

Yep.

Richard Dumas:

Okay, so, as Julia mentioned, I’m Richard Dumas, I’m one of the co-founders of a company called Slate CX, and we build a buyer engagement platform that helps you to engage with prospects and with your customers across multiple channels, whether it’s website, through email, through paid advertising. So, the topic of my presentation today is Clicks Don’t Teach, Questions Do.

And so I’m going to talk a little bit about, how to understand your buyer’s questions, and how to optimize those in a recursive way, which is the theme of the session today. So, I’m gonna, you know, piggyback a little bit on a new book that just came out, that was written by Kip Bodner and Karen Flanagan over at HubSpot, based on their loop marketing, philosophy.

And what they propose is that, and many people have done this over the last few years, that the funnel is dead. I don’t exactly agree that the funnel is dead. I mean, it still serves a certain purpose. You still have to build brand awareness. with your potential prospects.

You have to explain to them what problem you solve, and then get them into the process of actually evaluating your product, comparing you to other offerings that are on the market, and making a decision. So, I mean, the funnel still exists, it still lives. But what they propose is that in marketing, or specifically for marketing campaigns.

If you think about things in terms of a loop, and this can be a recursive loop, that things change. And their stages are, one, express. You know, how do you test and express who you are, your voice, your style, your point of view? How do you personalize each interaction? How do you use AI to make each interaction relevant?

And then under Amplify, what’s changed from the old funnel methodology is that people would see an ad or do a search on Google and come to a landing page, and that was basically where they would learn about you. But now your prospects are learning about you across all kinds of channels. They’re on Claude, they’re on ChatGPT, they’re in other places, so how do you amplify that reach? And really, how do you evolve?

How do you learn and improve? So the basic philosophy that they are proposing is that it’s no longer the company that spends the most in order to optimize that funnel, but it’s the company that learns the fastest, and based on those learnings, feeds it back into their recursive loop in order to improve performance. So I think that makes a lot of sense, and I agree with it.

So I think it raises a few questions, practical questions of, you know, how do you do this? So specifically, what should you learn from? What and how should you personalize your interactions, and how do you bring the feedback into the loop so that it improves over time?

So, as the title of this presentation describes, you know, are you looking at the clicks and the journeys from your customers, or are you actually learning from the questions that they actually ask you? You know, can you change a page, or do you actually change the answer? And what’s the right way to actually take that feedback and build it into your loop?

So, one of the things I want to point out is that, today, you know, most buyer questions never reach marketing. You have different roles in the… in a B2B buy and purchase cycle. For instance, you have your champion, you may have someone in the IT team, someone in finance, an executive. And they’re all asking different questions in different places.

So… How do you get to the bottom of what they’re most concerned with, what their pain points are, what the obstacles are for moving the deal forward? You really need to dig as deeply as possible into the kinds of questions that they’re asking across the buying process. And how does that differ?

Again, if you go back to the old kind of funnel methodology, what we would all do is, you know, look at the prospect’s journey. How many times did they visit my pricing page? How many times did they open or click through on an email? Did they download an asset for each of those things?

We would give them some more points, and basically, once they reached a, you know, a certain score, 85 in this case, we would decide they were an MQL, and we would happily ship them over to sales and say, hey, I’ve got a new qualified lead for you. But it’s really changed in the new world, so when you look at a buyer’s questions now, they may ask on our site, you know, how does this work if we’re not on HubSpot?

You know, can our security team see where the data goes? You know, how long would it take for a team of 3 to get this deployed. So, some of these lead back to positioning, some lead through to. RFP-style questions that need to get answered in order to move the deal forward. Some may identify content that’s missing.

Some may identify opportunities for highly personalized answers that you can offer, for specific types of buyers. So what you learn by looking at the question is who’s involved, what’s blocking them, and how to fix it. And, just to revisit a survey that we did a little while back, even buyers who visit your website or visit ChatGPT mostly come up short when they’re looking for answers.

We did a study recently, and you can find it on our website. About half of serious buyer questions get answered on a typical site. And 4% of those answers actually explain how the product works. On most sites, we just make bold claims, we’re gonna reduce costs, we’re gonna lower your cost of ownership, but we don’t actually explain how the product works.

And then many sites still aren’t using, just basic conversational AI across the site. Many don’t have chatbots at all, or for those that do, they still use them in the old-fashioned way, which is answer a question or two, and then try to get the prospect to book a meeting.

So again, the more that you can Offer those answers and engage with your prospects, the more you’re going to learn and the more you’re going to be able to optimize. So again, every question points to a stage in the loop, from the loop marketing concept. So, express, positioning statement. So, for instance, they may ask on our site, is this just a chatbot?

Well, that exposes that, you know, we haven’t explained what a buyer engagement platform is well enough if we’re still getting that type of question. How long would it take for a team of 3? That’s an opportunity for personalization. Missing content, you know, what happens to our data if we cancel?

And… By publishing that content on your website, you’re also publishing it to ChatGPT and Claude, because what they do predominantly is when someone asks about your particular product, they crawl your site and look for answers to those questions. And then evolve, you know, you can look at questions and understand product opportunities, or gaps, or what to build next.

So this is an actual example we had at one of our last events. Can this be used as a portal for our channel partners? And I don’t think we explained that clearly enough on our website. that it could be, how it could be, so it exposed a gap in some of our content, and also an opportunity for us to market to a new set of buyers, and I’m actually going to show you a little bit about that at the end of the presentation.

So again, a lot of what we focused on in the old funnel world was personalization meant kind of personalizing a page on your website. You know, you understand that it’s IBM coming to your website, and you change the headline and the logo so that it kind of pretends to talk to that buyer, but, you know, the buyer’s still basically reading the same content.

But AI, you know, can change and personalize the answer specifically to that particular company and buyer, and I’ll show you a little bit about how that works in a minute. And then before I jump in the demo, I think it’s important for this session, you know, we talk about how do we gain feedback and what should we do with it. We think about it in terms of three steps: automatic, AI-assisted, and approved.

So, automatic, the system, every time when they come in to your site, or they respond to an email, the buyer agent can personalize the answer to what that buyer tells it, what it knows about them, what industry they’re in, what they did previously on your website.

It can spot a question that it can’t answer, it can identify, it can label it as a content gap, and it can create a task for you to follow up with them, or provide more content in the future that answers that question.

So, in terms of AI-assisted, you can take all of the, you know, conversations that you’ve had across all these different channels with your buyers and group them into different categories and themes, so you can understand, you know, what are the most frequently asked questions, where are gaps, and create flags. Now, what we still do, and we still advocate, is approved. You know, anything that a buyer will read.

We don’t want the AI approving its own homework and automatically generating new content and new answers that we haven’t seen, and I don’t think most of our customers do.

We actually ran into this a lot in the contact center world, where big banks and those that had lots of regulation around them Wanted to be, very confident that anything that they published or any answers that they provided had been reviewed and approved, and often by a legal department.

So, we think that this recursive loop is one where the AI can assist you, it can help you identify gaps, but ultimately the person, is the decision maker. And that’s actually exactly what HubSpot’s own loop guide says. The same thing, the people own the strategy and the judgment, and a human approval should gate what AI recommends. So, we believe that as well.

So, I’m about to show you, a demo, and what you’ll see in the demo is one buyer, four acts. So in this case, I’m going to show you an email that went… that goes out to a buyer, and when they open their slate, or their workspace. They they can ask the agent any questions, they can see cards that are on the slate. You’ll see that she asks and brings in someone else from the buying committee, in this case a CMO.

You’ll see what happens when the agent can’t answer a question, and how we handle that. And then finally, we’ll show how we learn from all those interactions, both in our own system, and even through systems like Claude. So this is an example, of an email. This is an example that I sent, actually, after one of our last sessions.

All of you who are attending will probably get one of these from me after the session, after Julia gives me your the list, but what I did is I sent out a recording of my last session, and I put it on a slate so that anybody could ask questions about it, or they could watch the session again. So, consider that someone clicked on this link, and let me just… Switch over to my other system. Give me one second.

So in this case, you’ll see Mary received the email, and she lands on what we call a slate. So it’s a personalized buyer workspace, and the slate has a buyer agent on the left, where you can ask questions in text or in voice. You have different cards that can be added, so in this case, there are two cards with recordings of my previous sessions. So cards can be all kinds of different content.

They can be videos, PDFs, interactive calculators. We support things like Story Lane, if you want to do demos. So, you know, when she gets to the site, you’ll see it knows who she is, Mary, because, we included information from the HubSpot contact record about it. It knows what assets are on here. And then, let’s say, you know, she could start asking additional questions on her slate.

So I’m gonna give you kind of a walkthrough of an example of a potential buyer of ours, an end buyer, who would actually come in and start to engage and ask questions about Slate CX. So I’ll pull up for you a customer example. So, in this case, it adds a new card. This is a video that shows how Nexla uses SlateCX, so she can watch that. And then let’s say… She asks another question.

I’m cutting and pasting so I don’t have to do so much typing. So, this question is, how can I tell how well my current website provides answers to ChatGPT and Claude? So the agent brings up additional cards. In this case, we have a kit. That someone can use to actually test their existing chatbot. So there are instructions for how you use it. This is actually a webpage that it’s bringing up.

And maybe she says, well, you know. How can I tell which questions my prospects have? So, more specific. question. So, in this case, she could use the card or the application to review the performance of their existing chatbot. And in this case, it’s saying, I’ll check the knowledge base, says about, seeing what questions Ask, okay, so you can see the questions Prospect asks in the conversation view of your reporting.

It lists every conversation, sortable by last activity or message count. And what it does is it brings up another video that actually explains and shows how that works in the product. So I’ve added the conversation reporting card to your slate during a conversation. Once you tell me… okay, so… And let’s ask it another question.

Let’s say… This is all giving me information about how well the system stores the questions that buyers ask, but what about the ones that I don’t have answers for? So, the way that you deploy your slates is that you provide content that contains the answers to these questions. It’s not magic, it has to find the answers somewhere.

The beauty of it is that you do get to control the answers, because you’re providing the content. In this case, it says, questions your agent can’t answer show up in two places. There’s a possible unanswered question area, that’s in the admin UI, so it flags all of the possible unanswered questions. And then it also suggests this.

It suggests that, Mary actually grade her own website, and see how well How well it can answer questions. So, in this case… Alex? Correct. And spoiler alert, Halvex is made up, so they’re gonna get a really, really bad grade when you grade their website. So let’s go a little bit further, and let’s… Ask another question. Can we log in with Okta? That’s an interesting question.

So here’s a case of a question where the agent doesn’t have an answer, or doesn’t think it has a good enough answer. So it says, would you like me to send the question to the team so that they can follow up on it? Oh, I was like, okay, that’s great, let me… Submit this question to the team, right? Now, what you’ll see on the back end is that that generates an alert. So here you’ll see the alert pop up in Slack.

You’ll see Mary from Halvex has an open question, and if I go through the third wall now, I’m no longer the buyer from Halvex, I’m Richard at SlateCX. I can click on the link. I can review the question that she had. Can we log in with Okta? Oh, okay, let me write an answer to this question. Okta, okay, not today. But we have it on. Our roadmap. The… Okay, so… Well, and I can post the answer, right?

So this is the idea of a human in the loop, in real time. And if we come back to her answer card, you’ll see the answer from the team came up, Okta, not today, but we have it on our roadmap. And you can click back and she can find where that was in her conversation.

So, great example of human in the loop, and all of these unanswered questions also go into your reporting admin, so that you can… you can follow up on them. So let me do one more here. And we talked about it in kind of the loop marketing concept, it’s also about personalizing the answers to a particular buyer.

So here, Mary is asking, can you show me an ROI example, given that I have 10,000 monthly web visits, an average deal size of $50,000, a win rate of 30%, and by the way, can you graph that for me? So, the agent can actually go through, because we’ve given it a skill to build ROI models, and it can start to build a customized ROI model specifically for Alvex and for Mary. So that’s pretty cool.

She gets a graph, it can show, currently, with Slate, without Slate, what kind of return on their investment they would get. So at that point, let me show you… Mary has done a little bit of evaluation here. We’ll switch to speech mode as well. You can always type and… or speak. It’s easier to type when I’m doing a demo here. Welcome to voice mode. I’m Slater.

If you want a quick next step, hey, can I invite my boss to the slate so I can share all this stuff with her? Nice idea. Let’s talk about the best way to bring your boss in. Yes, you can! Use the Share This Slate option to invite your boss and give them access. If you’d like, tell me what you want them to focus on, like ROI, reporting, or testing, and I’ll help you point out the most useful parts.

So it’s… well, what should I highlight for? Start with the ROI example and the conversation reporting demo. ROI helps them see business impact, and reporting shows how you’ll track performance and outcomes over time. So I’ll stop that for now. But again, the concept here is that the workspace, or the slate, is a collaborative environment, and it can have multiple team members on it.

All with different buyer roles, for instance. So now, Mary Marketer was the one that I started as, and that’s a manager of demand marketing, and then Joe, is her boss, and she is the CMO at the company, so they’re now both seeing the same set of cards, and the same questions and answers. So, with that, why don’t I… let me just show you one more thing.

So, let’s say… For instance, Mary then… Decided later on to come back to our website. And decided to open up the chatbot on the website. What you’ll see here is the same slate with the same conversation that she has been having all along.

So this is the idea that the workspace and the place where you’re collecting all of your questions and providing all of your answers is persistent for the buyer, and their evaluation just continues no matter which channel they come back and engage with you on again. So let me show you something else that is pretty interesting. Make this a little bigger. Okay. You’re welcome. So here is, Mary’s slate. Let me update it.

I’m sorry, this is Mary’s HubSpot contact record. Mary Marketer. So you can see… that she did the Halvex demo. You see that she has a Slate Intelligence card. You can see how many Questions she asked, or how many messages she sent, how many cards were pushed on her slate, and then also how many goals were achieved.

So in this case, we had a goal that said, we’ll educate the prospect, and we set a metric that said if the prospect asks at least 3 questions, we’ll consider them to be educated. Now, you could change that to 10 questions, you can let the LLM decide on its own when it feels like this person is actually educated and understands the product. You can also view the entire conversation and message history.

So, she said, who uses a slate? You know, how can I tell how well my current website provides answers? All that information, along with all the activity, is stored in their HubSpot contact record. And what’s even cooler is, if I’m a sales rep, I can actually go in and push new things onto her slate.

So let’s say we had a phone follow-up conversation after this, and I can see, oh, the buyer answered, Raider is already on her slate, this video’s already on her slate, but, you know, she doesn’t have goal setting in Slate CX, and she was interested in that. Let me just select that, and… Push that new card over into Mary’s slate. So if I go back over to Mary… Doof. Nothing there.

And a little network error there, but okay. So now you see the new card, and the agent knows you added goal setting in Slate CX to your slate. It’s here’s a screenshot of how you set goals within the UI.

So, another kind of… Interesting way that you can, continue engagement even if the person isn’t on your website, they’re not conversing over email, but they’re actually interacting directly with one of your sales reps. So let me show you another area here. I’m not sure how much time we have left. I’m going to close that. I’ll quickly show you within Insights here as well.

We have all of the conversations that are stored, so you can see this cloud that says, most people are asking questions about what is it and how does it work. They ask a lot of questions about pricing and cost.

There’s some other areas here, people are asking about geographic regional availability, and I can click on that, and I can see all the conversations and all of the slates that asked about that topic, like, when is it going to be available in Germany? Can I store my data in Germany? Those kind of things. And then finally, let me just pop back over to… Let me close up too many windows here.

Even in, Claude, you can ask questions. So if you wanted to get information about this, I asked Claude, use this late connector and create a pie chart that shows which cards are pushed most frequently to my buyers. So, you know, how was Nexi built, Nexla case study, short intro demo, you can see all the different cards, that were there. And, you know, tell me… about… Halvex. Dot coms. Questions.

And you all have used Claude, so you know that it sometimes takes a few minutes for it to come back, so I’ll just see for a second if it, returns. I’ll let that run, and then I’ll switch back to the presentation for the end. So within Claude, you can ask all kinds of questions, you know, what did buyers ask this month? What questions keep coming up? And which stage of the loop does each point to?

So the connector can do all of your analysis, for instance, directly into SlateCX without you ever logging in, even though we have all those same dashboards as well. So, back to that question, the AI didn’t fix itself, but it told us exactly what to fix, and it gave someone at Slate the ability to create content and create new answers that will actually answer those questions in the future.

And I had a little bit more to cover here, but it looks like we’re running out of time. So, what I will just do is leave you on… This? Three things to think about doing this quarter. Find your answer gaps, so test your own site, find out what AI says about you on Claude and ChatGPT against the questions your buyers actually ask. Add those questions, to your website, to your pages, and to your AI agent.

And turn every question into some kind of action. Improve your positioning, understand product gaps, those kinds of things. And then, just as a teaser, I focused mostly on, you know, updating the quality of your answers here, but that’s really just the start. You can do a lot more optimization. What if your goal is How do I create more engagement?

You could ask what answers and cards lead buyers to ask more questions and return to my site most often? more shared context? How can I optimize, the questions and the content that I provide, or the flow of the conversation, so that people provide me more information about their pain points, their challenges, their timeline? Or, how do we optimize for better outcomes?

So if our outcome is to get people to book more meetings or submit more RFPs, what kind of questions and what kind of journey actually leads to those kind of outcomes? So the same loop, capture, learn, improve, recursive, and with a human in the loop to actually approve the results. And that’s it for me. I think, I probably ran a little long there.

Julia Nimchinski:

Thank you so much, Richard. Great presentation. Curious about, there was a use case that is quite rare. The AI admitted that it doesn’t know the answer to a question. So, how do you solve for that? Because, you know, more frequently than not, it would just hallucinate something. Got it.

Richard Dumas:

So the way we’ve built it, there are two ways that you can do that. We can actually create specific topics that you don’t want the agent to answer questions about. Maybe you don’t want it answering questions about pricing. And whenever a pricing question comes up, instead of just saying, I can’t answer that, or book a meeting, it can do what I just showed you.

It’ll say, I can get an answer for you, a personalized answer for you. We can also offer the opportunity to book a meeting as well to continue that conversation. So you can, create specific topics.

The way we do it beyond topics is when the agent decides that it is not highly confident in the answer. what it’ll do is it’ll say, well, I’m not sure I have this page that relates to the question you’re asking, but it sounds like I may not have all of the information you need. Would you like to Bring a human into the loop.

So that’s a way, I mean… We can also, you know, bring a human onto the slate or the conversation as well, but more and more, I think people are less interested in having live chat, or, you know, making sure they staff enough SDRs to jump on a slate in real time.

But this is kind of a, you know, in-between solution where you can put a human in the loop, they can respond, it may not be… In real time, but it may be in an hour or so.

Julia Nimchinski:

Amazing. Thank you so much. And last question is more kind of a visionary question. Do you see that, like, all the intel that you’re gathering from the marketing perspective? And all of the buyer conversation, could they eventually just recalibrate the value of the product or platform, features, workflows, or, like, how far does the recursion go?

Richard Dumas:

Yeah, no, I mean, that’s not that visionary. We’re working on that now. So, you know, what you’ve seen now is we’re giving the buyer the information to be able to do optimizations themselves. But we are working on self-optimization for the system. So, I don’t mean generating content or answers for them, but the last example I gave, you know, what leads to better goal optimization?

And can we suggest strategies that will help you, book more meetings or generate more RFPs, for instance? That’s something we can definitely learn from all of the history of the interaction to see what drives the best results and how to fine-tune the system.

Julia Nimchinski:

Awesome. Thank you so much.

Richard Dumas:

Good, thanks for having me, appreciate it.