This session explores how to create predictable, high-impact AI feature demos using Saleo. Christina Simms and Jason Pierce from the Saleo team walk through real examples of simulated AI responses, assisted AI, and authentic AI pass-through, along with how to leverage powerful tools in Saleo like the Demo Data Agent and tokens for instant data personalization.
Read the full discussion:
I want to thank you all for joining us today. Really excited to talk about how you can leverage Saleo for better AI demos.
We have two amazing Saleo leaders who are joining us. You probably already know them. But we have our VP of customer experience, Christina Simms, and our principal sales engineer, Jason Pierce.
So just really quickly before we dive in, two reminders. One, we’re going to have time for Q&A. Please throw those in the Q&A window for us. It helps keep things organized, so the chat we can all visit and then Q&A can be questions.
Also, today’s webinar will be recorded. We’ll be sending that out afterwards, so you’ll have access to that quickly.
And with that, I’m going to go ahead and hand things over to Christina to get us started.
Thank you, Laura. Thrilled to be here with you guys.
Love getting a chance to talk specifically to customers, so thank you for taking the time to join today. We are really excited to talk to you and share a bit about AI and how we can both help support your AI features and also share some ways that our own Saleo AI can hopefully help you scale demos, create better data, and just further all the impact you’re able to have in Saleo.
So just to go over the agenda quickly, kind of set the stage a little bit with a quick Saleo overview and high level understanding of our approach to AI. Then we’ll get into talking more about how we can support the demoing of your AI features, and Jason will lead a live demo to show you an example.
Then we’ll spend a little time talking about some Saleo specific AI features, our demo data agent, and how that can also be leveraged with our tokens. Then we’ll give you a live demo of that so you can see it in action.
And then we’ll walk through some best practices and then do a quick sneak peek of a couple of upcoming road map items that relate to what we’re talking about today and then close things out.
Alright. Well, I will go ahead and dive in. So most of you are familiar with Saleo and all of this, so I won’t spend too much time here.
But we are really proud of our focus on helping you deliver great live demos with our data injection of being a leader in G2, and you may see yourself or some familiar logos in this sampling over here of our fantastic customers that we’re just really lucky to partner with.
So when you look at the broader Saleo platform, we have our bread and butter – Saleo Live.
It’s the foundation of how we support most of our customers, but we have our capture product as well. So this enables you to share really powerful guided tours or sandbox tours with your prospects and customers and really further the impact of the great data stories that you’re able to tell.
And today, with our focus on AI, and how that’s being infused across the Saleo product to drive greater efficiencies and value for you all.
So we’ll get into how we can help you demo your AI features. We know that AI and the features that your product teams and engineering teams are developing and releasing are incredibly powerful.
They’re probably going to drive a lot of value for your customers, but it can also present some challenges to demo those live. They can be a little unreliable or difficult to demo live.
It may be that you have great examples when you’re doing live production data, but from a compliance standpoint, maybe that’s not always possible. You might have challenges with predictability.
The responses that you get could change some, and it’s tough to go into a live demo not being totally sure what you’re going to get. It can also be hard if it’s just reliant on native demo data in one instance to do any kind of personalization or adaptation so that you can tell multiple stories.
And then you also have the added layer of complexity if your copilot or AI assistant is retrieving data from multiple sources and you’re trying to do this live but in an environment that maybe isn’t truly actively used by a customer where all this stuff would be connected and easy to show. So there’s several ways that we can work with your AI features, and we’ll talk through those quickly.
So one is a full simulated AI approach. We are able to allow you to input keywords to trigger, in your prompts that will trigger specific outputs.
So you get total predictability in what your AI feature is going to surface. This can be great because it can reduce costs from running live AI queries in your product, and it really eliminates any of that risk from unreliability or uncertainty in their performance of your AI feature, and great to remove that risk when you’re demoing live.
Another approach is assisted AI. So this is a bit of a hybrid.
We can let any of your really stable features run live and natively as they would. But then if there’s certain prompts or certain parts of your AI assistant that are a bit more unpredictable or where you need to be able to show different scenarios that you just can’t support natively, we can have Saleo take that over.
And then the final is authentic. Maybe your feature does great natively and is super predictable when you demo it live.
You don’t need Saleo support. We can let it pass through as we often say, and have that authentic AI experience be demoed, but then have Saleo supporting other parts of your demo, as you’re talking to prospects.
So with that, I will turn it over to Jason so that he can share a live example of this and let you see it in action.
So, I’m going to use Salesforce and show Agentforce specifically. So this has been a major use case for them as a customer.
And, you know, one of the challenges that was first and foremost in mind is it’s not predictable. You can ask the same exact question five minutes apart and get two very different answers.
And while that’s appropriate in the real world, not ideal for a live demo. Helps to know exactly what you’re going to get.
So what I’m going to do is I am going to run a quick example of a query without the demo running, and then we’ll bring in the Saleo extension and show you how we can transform that and how we’re taking that control. So, we will bring in agentforce right over here, and we will prompt it with a question, “how is loyalty trending across our data streams?” That is a major component of this dashboard.
So we’ll see what this yields as far as response goes. Obviously, that thinking process, that interface where it shows it’s being processed, that’s key.
Want to make sure we may maintain that, which you’ll see. But the query is too vague to get specific results.
Not shocking in this case, but, again, it’s not an ideal response in real time during a live demo. So I’m going to close the Agentforce extension, and we’ll open up Saleo.
And I need to go to my Agentforce demo right here. I hit the play button.
The screen refreshes as most of you are probably well familiar with at this point. And so once this comes back up, just to call out the very obvious, this is still native Salesforce.
We’re not on rails. We haven’t changed to a simulation.
Still the native product. The only difference now is that Saleo demo is running, and so we are injecting data where applicable.
So we will bring in the sidebar again as the exact same prompt. How is loyalty trending across our data streams? Again, maintaining that native, effect, thinking, processing, and then we get a response, which, in this case, is formatted exactly in the way a native Salesforce response formats, and I do believe there’s going to be a graph that shows up here.
Yes. There we go.
Sometimes the load time can be a challenge in itself. So, we’ve got a positive trend which corresponds to what that text output is, all of which is being controlled by the Saleo demo.
And I’ll show you where and how that’s working here in just a moment. But just to show a second example here, we’re going to ask about which of our segments need the most attention.
So similar prompt, looking for some general insight. And, again, we have a custom, well crafted response with a visualization to accompany that.
And what I have found is it’s often really valuable to use AI to give you a realistic response. Maybe just tweak the prompt a little bit to make sure it’s going to align both with the story, and to give you the output that you’re hoping for.
Now the way this works, I’ll stop the demo, and the screen, of course, refreshes, goes back to its original state. And now we’ll open the extension again and drill into this Agentforce demo.
And so we have a table here that controls the Agentforce interactions, and it’s really quite simple. So the message column here is the prompt or the keywords of the prompt that we’re looking for.
And then in this case, we’ve got a markdown response so we can include formatting, bold text, bullets, all of that, and then even a graph designer built right into this table to control the interactions. So just like any graph designer you’d use in Saleo, you have complete control over exactly what trend this shows and also over what time period.
You can label the different axes so that you get a really relevant output. And then since Agentforce includes sources, it wants to be able to attribute that insight to where it came from, we can include that as well.
So, truly, every piece of that native response can be replicated within the Saleo extension, and you can create as many prompts and responses as you need. What’s also important to call out is that we can essentially set either a default response.
So if the prompt entered doesn’t match an item from our table here, we can provide a default response. But in many cases, it’s preferable to use the native output.
So if the prompt doesn’t match any of the values here in our table, we can actually, in most cases, pass that through and prompt an authentic response from that AI bot. So just depends on what the best use case is and how you want to show that during a live demo.
But this can remove that unpredictability, that variability quickly and easily so that you know every time what you’re going to get, and you can tell a consistent coherent story around that. Alright?
Awesome. Thank you, Jason.
And I’ll say with this and everything we talk about today, I’ll make a couple plugs. One is our knowledge base and help center has quite a bit of resources there.
So they’ll be probably most applicable to this next section that we go into, but want to make sure everyone’s aware of being able to go there and check out our articles, how tos, screenshots, videos, and then, of course, your CSM. So, depending on your relationship to Saleo, you may need to work with one of your admins or the champion of Saleo on your side.
But every one of our customers has a dedicated CSM, and they are always happy to host office hours to talk to you about strategy. So if you are curious, you have an AI feature that you are curious about how Saleo could support, don’t hesitate to reach out, and they can help talk you through that and and dig into the details.
Alright. I will get us moving on to the next section here.
Okay. Switching gears. We’ll start talking about Saleo’s own AI features and how they can drive some additional value for you all.
And just want to set a little bit of a stage first with two different ways that we’re able to support some of your data customization needs. The first is what we call overlays or you may have heard it referred to as text replacement.
So this is a great way to make quick changes to either text or tokens. It’s out of the box.
It’s really powerful, actually, and I think customers sleep on this sometimes. So I’m also going to advocate that you explore ways that this could do more in your product, and in your Saleo demos.
But we’ll talk a little bit about how Saleo AI can really amplify this overlay feature. And I think you’ll see how powerful it can be in the next slide demo that Jason ends up doing.
But the thing to call it here is it is more, you know, on that, superficial layer. So we’re making changes to the text and the HTML, but not that deeper level of the actual API, which is what we’ll talk about next with data injection.
So this is what’s powered by our modeling engine. This is actually injecting live data from your Saleo tables into your product when you hit play on a Saleo demo.
So this gives you a bit more robust way to tell those more complex data stories where we actually can do more with interconnected data and have that deeper injection at the API level, rather than at the surface layer like the overlay. Both are really powerful.
Both have fantastic use cases. And this is another place where your CSM, your implementation manager, they’re here to help guide you in the best times to deploy this depending on your needs.
Alright. So we’re going to talk a little bit about our Demo Data Agent.
So we know it’s really important to have great data in your demos, but we also know it can be really time consuming to populate that data, especially if you need to customize stories to really connect to your prospects or customers. Our Demo Data Agent allows you to instantly create demo data.
It will allow you to scale that data in seconds and then even do things like translate it into different languages or different currencies if you need to speak to prospects based on their geography. It’s a really powerful feature.
And it works by essentially you input a prompt into the demo data agent, and then it does the rest. It will fill out your Saleo tables or tokens.
And, as you’ll see in what Jason shares, it’s literally does it in under a minute. I will set the stage quickly too just so if you are either following along in Saleo or you want to go explore this, you’ll see this little blue kind of the blue stars.
That’s one of our Saleo AI icons. You’ll see this for a blue button with Saleo AI.
And when you see that in a table or in the tokens, you’ll be able to click that and open up AI, enter your prompt, generate and review your data, and then you’re good to go. You’ll have a new customized demo that you could save, you can share, and then hit play on.
Alright. The next thing that I’m going to talk about is our tokens.
So tokens, we often kind of refer to these, on our team as Saleo Mad Libs, which is a fun way to think about them. But they’re smart reusable placeholders that allow you to personalize text within your Saleo demos.
So they’re great to customize things for maybe a specific account, for different verticals, company size, geography, persona. There’s a ton of great use cases that you can leverage tokens for.
And there’s a couple things about tokens that are really great that I want to highlight. The first is that admins, demo creators, and sales seats can all customize and edit tokens.
So, it really gives you a great way to scale customization across all your different types of users, but also maintain a certain level of control. So as an admin or demo creator, you can set up tokens, and then your sales users would have the ability to edit them, but they couldn’t create net new.
So you do have this great balance of flexibility and control in being able to edit data for different demos. The second thing I’ll highlight is that tokens are out of the box.
So, they don’t require any elements, if you’re familiar with those, to support them. You don’t need any support from Saleo to set them up.
You could go open up Saleo after this webinar and start creating tokens, on your own. So, really great to be able to just dive in and get rolling with them.
And I’m going to turn it back over to Jason, and he’s going to walk us through some examples of this live.
Absolutely. Let’s take a look. We’re going to use HubSpot for this example.
And, right now, we’re looking at a product’s page, which obviously is not showing any real or compelling data. Essentially, it’s placeholders right now.
So, we’ve got the name of the products. We’ve got SKUs, prices in euros and dollars, as well as descriptions.
And what I’ll do is go ahead and open the extension again, and I will bring in my token demo, and just hit play to get started. And as soon as I do that, of course, the screen refreshes, and we are now going to look at a list of, in this case, consumer electronics products.
Everything you’re seeing now is being controlled by Saleo, and it’s all using both, the Demo Data Agent and tokens. And I’ll show you how easy it is to control this.
But what’s great is that it’s all relevant. It’s not just close enough kind of data.
It’s calling out actual products. It’s giving us relevant SKUs.
And the best part is the descriptions are detailed. They’re applicable to each item.
So it’s not just an example or sample, data. This is real and very, very applicable and easy to repurpose for each audience, whether it’s different vertical, industry, persona, however you want to slice and dice that demo data.
So what I’ll do next is I will, again, stop the demo, at which point everything resets back to its original state. And now we’re going to open this back up, and we will, again, drill into how this demo is configured.
And I’ll go straight to the token library. So since we had set up tokens to control truly every placeholder value on that product list, that’s what you’re seeing right here.
The descriptions, the SKUs, the product names, the prices, all of that has been tokenized, and you see the values that are applicable for this electronics demo. But now I can use Saleo AI, and I’ll just say, to update all rows to reflect data.
Let’s say: reflect products sold by a, let’s say, clothing retailer. That’s an easy one.
Start with that. And you can get as general or as specific as you want.
If you want to get down to the brand level and talk about products geared at a specific audience, AI is shockingly good at getting really, really nuanced. And so the more fields you have, the more detailed your prompt, the longer it takes normally.
As you can see, though, it’s going through and it’s selecting specific data points to update. And now you don’t have to accept its output.
You can go through and make any changes to any of these values that you’d like. But I’ll go ahead and accept these and close that out.
Hit the play button again, although you can already see most of it. And in a matter of, what was that, forty five seconds or so, we have just changed this electronics demo into a clothing retailer demo.
And that’s a good example to show the applicability, but also the speed. As Christina mentioned, you don’t have to have access to configure tables or update graph designers.
This is not a demo creator function exclusively. Anybody who has access to the Saleo extension is able to go through and apply this tokenization and the data, agent to update that content.
And just to kind of expand a little bit further here, I’m going to go one level deeper, and we’re going to open that extension back up, go into the token library one more time, and let’s get even more specific. So we will use AI one more time, and we’ll say update all rows to reflect products sold by Nike.
A good example account everyone’s familiar with, and this is just to highlight that this isn’t just going to talk about generic shoes. This is going to call out Nike specific products.
So you can get very, very nuanced, very granular. We’ll accept those changes.
It also applies relevant prices that are specific to the item and also roughly appropriate for the conversion rate between euros and dollars. So, really, you can control a massive quantity of data this easily.
So I’m showing, I think it’s 20 products at a time. You can do this for 20,000 if you want to.
It might take a few more minutes, but you can do this, for an entire demo’s worth of data in a matter of a couple of minutes. This, as somebody who is building demos and delivering demos literally all week, has been a game changer.
This has saved me hours and hours every single week, and would love the chance to talk in more depth and and help the rest of the audience out there figure out how to make this applicable for what you’re doing.
Awesome. Thank you.
The thrill never leaves me when I see Saleo.
Okay. Well, let me get into some best practices.
So, we want to give you some things that are a bit more actionable to think about as you go back and consider, either deploying some of this for the first time or maybe just taking a second look at some of what you’re using today and your strategy. These, once again, are things that your CSM would be happy to do a deep dive with you on, but wanted to share some highlights on best practices that we’ve learned from our own testing and from working with customers around these things.
So this is really specific to the Demo Data Agent and then using it more specifically for tokens. So first, it’s best to keep your prompts for the Demo Data Agent short.
Be being really kind of simple, focused, one prompt at a time. You can see how quick it is.
It’s very easy if you need to do additional prompts and responses and updates. You can kind of do those in order, but try to keep each individual prompt a bit more simple so that it can create a great output for you, and then you can move on to the next thing that you need to update.
Another thing. Our AI will look at the descriptions that are provided for the tokens in the Saleo product.
And we have a recommendation to, in those descriptions, while you might provide some insight for your own teams about what that token is or what it’s updating, it’s also really important to provide some insight for Saleo AI in those descriptions. So we try to include two things, context and guardrails.
So context that’s clear and concise, and it will help Saleo AI understand a bit more about what those token values should be and then any guardrails, like a field that is a dollar amount, but you always want that dollar amount to be between a 100,000 and a million dollars. So you can put that kind of context within the descriptions.
The next thing, if you’re thinking about rolling this out to your teams, is we often recommend setting up an MVP of tokens to start. We’ve got customers that utilize hundreds of tokens.
They’ve really mastered doing this at scale and being really thoughtful about how they create tokens to enable all of their users to make updates. But I think a great way to start is to think of some areas that would have high impact for your users to be able to update.
And maybe you start with 10 to 15 tokens that they can kind of get their hands on and start to leverage. And then let demand from your team drive the additional tokens you create.
So it can be overwhelming if you just turn over 50 tokens and people are trying to figure out what makes the most sense to update for different types of conversations. But if you start with a small subset, they see value and then you’re able to have a team or end users start to ask for more.
That’s a great way to kind of organically grow your token usage where you know it’s having an impact with your team. And, along those lines, as you start to scale out your token usage, we have some different ways to organize tokens within Saleo, both our folders and our tags.
These are great ways to organize tokens, maybe it’s all tokens that need to be updated for a specific vertical. Or some customers look at tier one tokens, which are if you just want to update 10 things for an early stage combo, filter by your tier one tokens.
Maybe you have a later stage demo where you want to do deeper customization, filter by tier two tokens and update more. But there’s lots of different ways you could think about organizing them, and that’s something else that CSM can help you talk through, and also share examples of things that they’ve seen other customers do.
Alright. We will do a quick sneak peek now of a couple more things that are coming on the road map, just kind of related to this, and then we can see if we’ve got any questions to talk through.
One, if any of you are using Saleo tables and you have sub tables, we know that that can be a challenge as well to get data updates to kind of flow through and connect. So we are currently working on AI for subtables, which we’re super excited about.
So I know this for any demo creators out there using subtables, I hope this is music to your ears. And we are really excited that this will be just another way that we’re able to leverage AI within the product.
And we are also working on image tokens. So today, you can create tokens to update text within your platform.
We are also going to be releasing the ability to create image tokens. So places where you want to be able to have image replacements kind of scale across your demo where the same image may appear in multiple places.
You can also organize things in the token library in a much more meaningful way. And then you’ll also have the same ability to to draft, descriptions and organize these in tags and folders.
So just a little sneak peek there, but wanted to share a couple things we’re excited about, that relates to some of the things we talked about today. And I think with that, that is the majority of our slides and content, but would love to see if we’ve got any questions or any burning things from you all we can address.
Yes. We definitely have some questions from our audience.
So thank you all for sending those in. I think we can start with can you please expand or clarify on Saleo never touching the back end data?
We sure can. There are several ways that this can be set up. But for the most part, Saleo is never going to actually push data to the back end.
So nothing that’s done during a demo will ever persist beyond the demo. It will never impact the actual tenant or instance that you’re using Saleo on top of.
What we can do, through what we call cache blocking, is we can essentially simulate what that would look like. We can use a workflow to go end to end to create, update, delete actual records.
But as soon as that demo is turned off, it all resets right back to its original state. So the actual date that that’s being changed is cached within Saleo.
So it appears as though it’s real back end data. We can give you the entire experience, but it won’t persist.
So there we are, built to step between the front end and the back end and just return a different dataset based on what API calls are being made, what data is being queried. So we stop anything appropriate from actually hitting the back end while we can also work as a pass through or a hybrid function and allow certain data from the back end to populate on the front end.
But that’s one of the beauties of how Saleo was built is that you don’t have to worry about actions being taken during a demo actually changing anything for anybody else using that instance. And and happy to speak in more detail if there are questions more specific about how that would work.
Awesome. Thank you.
Another one here, I see, image tokens. Are those going to have an AI feature once we get to that rollout? Do we have details on that at this point? I don’t think in version one, but we can find out what may be road mapped beyond that.
So first, we’ll just be the ability to tokenize images and organize them, And we can confirm if AI is to come later. Thank you.
Speaking of tokens, do they work with live only, or are they also available on capture? Today, they work just with live. They are on the road map for capture.
So while you can do text replacement now with capture and some post capture editing, we don’t yet support tokens, but it is on the road map just to do tokens with capture. Wonderful.
Another tokens related question. What should we prioritize to help empower our sales seats in terms of tokens? That’s a great one.
I think if you can start with things that would really have the most impact with your prospects. So if the stories your AEs need to tell, maybe before they’re bringing in an SE or where they’re demoing on their own, if it makes the most sense to be able to customize it for a specific vertical.
Or sometimes it is that company size, you’re going to show a totally different set of data or volume for an enterprise customer versus a SMB, or the geography. So things like the currency, or even the language.
That’s the great thing about Demo Data Agent is it can help update language. I think those types of things where you know it would have impact with your prospect.
If you were able to customize it, then that’s going to be probably the most powerful for your demo creators or your sales seats, but really sales seats for those tokens. Awesome.
Thank you. What’s the best way to structure tokens for big teams to help make it manageable? Great question.
Okay. We have some customers doing some really cool things.
I actually typed out a couple examples. I don’t want to forget these.
One of them I was speaking to a little bit was that they have tokens. They’ve got them defined by tier priority.
So if you want, a small set of tokens for quick customization wins, but it’s not going that deep, you just filter by your tokens that are tagged tier one. I update maybe these 15 tokens, that demo’s good to go.
Maybe I’m middle of the funnel, I want to be able to customize a bit more deeply or I have a higher ARR prospect where I want to spend more time, I might jump to the tokens that are classified as tier priority two or three. It’s going to be a longer set of tokens, but it’s going to have greater customization and give you more options to kind of craft that story to have impact with the prospect.
Others do things by different stories or demos. So they use the tags to, you know, once again allow a rep to filter and just look at what they need to update for that specific demo story.
And, that’s where I think developing some kind of nomenclature or system of organization is just half the battle, like deciding upfront how you want to structure things, and then you can really kind of scale out from there. But I hope that that gives some kind of relevant examples.
No. I think that’s great.
Looks like the last question we have so far, what are some of the bigger mistakes you see organizations make when trying to show new AI capabilities live? I know I can speak to a couple of Jason. They have some you’ve seen out in the field.
Not testing beforehand. Like, have your prompts tested.
Now it can be tough if a prospect asks on the fly, like, what happens if you type in this? But, ideally, you’ve got a scenario and a strong story and kind of a bit of a script you want to tell, and you’ve tested out so you know what that prompt and response or what that interaction is going to look like. And then I think the other, and this is where we can help, is if you don’t have a relevant example.
So, maybe you’re demoing something with an AI feature and it’s a good prompt and response or or good scenario, but it’s not as relevant to the prospect that you’re speaking to. It’s tough.
Sometimes we wish they could use their imagination a bit more, but some things can fall flat if it’s not really tailored to an example that would resonate with them. So I think being able to provide a bit more personalization in the example you’re showing is great if you can do that.
Jason, anything else you would add?
I would just say that I think and I’ve learned this the hard way is, it’s risky to rely on what you think is going to be a reliable, predictable answer. Even if you have queried AI repeatedly with the same prompt and gotten an appropriate answer each time, that doesn’t mean in fact, it almost ensures that when you do that live on a demo, you’re going to get something completely different.
So, don’t get too comfortable with your native AI interactions because those can change quickly without any warning at all, which is why having a predictable, you know, scripted interaction can be so valuable. But, also, you gotta show it.
I mean, that’s a huge portion of the value that your product’s going to offer, so you can’t be scared to show it. You’ve gotta find a viable way to really highlight that capability and the value you get from it.
So I think we’re all diving in headfirst, but you have to do it cautiously and with some solid control. And that’s a fine balance to strike that we’re all still coming to terms with right now.
Thank you again, everyone who’s joined us today. Thank you, Christina and Jason. This was truly a great session.
