Yoav Einav, Guidde: AI How-To Videos Beyond Loom

Guidde CEO Yoav Einav explains how AI made how-to videos a product, helping support teams cut tickets 20-30% and lifting signups to 1,000 a week.

Yoav Einav, CEO and co-founder at Guidde, on The AI Report podcast
November 14, 2024

Watch on YouTube · Listen on Spotify or Apple Podcasts

Yoav Einav, CEO and co-founder of Guidde, says generative AI only earns its place when it moves a business number, such as the 20 to 30% of support tickets his customers avoid with how-to videos. He tells host Liam Lawson how Guidde grew out of what Loom users struggled with, writing scripts, editing and keeping videos up to date, and how adding AI features took signups from 50 a week to 1,000 a week.

Key takeaways

  • AI went from a feature to the product. Guidde launched in November 2022 and took about six months to find product-market fit. As AI cut the time spent creating and editing videos, weekly signups jumped from 50 to 1,000. (10:05)
  • Sell business results, not time saved. Yoav says saving time is great but not something people tend to pay for, and not sticky. Guidde instead aims to quantify the business KPIs it improves. (07:03)
  • Support teams are the biggest users. Customer support teams using Guidde reduce 20 to 30% of their tickets, and tickets that do come in close faster when the answer is a video. (13:39)
  • Recording is easy; updating is the hard part. Editing in Loom usually means re-recording from scratch. Guidde set out to make updating a video as easy as updating a PowerPoint. (18:02)
  • Build or buy each piece of the AI stack. Guidde buys text to speech from two vendors but built its own algorithm to learn each customer's product names, and uses different language models for different jobs. (19:39)
  • Saying no is part of the strategy. Guidde turned down requests to become an LMS and to prioritize mobile, and focused on desktop applications instead. (34:23)
  • AI makes people more efficient, it does not replace them. Yoav expects AI to take over repetitive tasks so people can focus on their added value. (47:56)

What is Guidde, and why did Yoav Einav start it?

Guidde is an AI tool for creating how-to videos for software applications, and Yoav Einav started it after years of making videos the hard way. He began his career as a software engineer and led engineering teams focused on AI and video: streaming, encoding and natural language processing. He met his co-founder in 2010 and joined him as one of the first employees of a video delivery network startup, where he worked for about seven years.

As VP Product at another early-stage startup, he used a lot of Loom videos for pre-sale, post-sale and onboarding, which became part of the vision for Guidde. When COVID pushed work remote first, the two co-founders reconnected and saw an opening to reduce Zoom sessions and make video creation accessible to everyone in the workplace.

How did Guidde find product-market fit?

Yoav Einav describes Guidde's start as an evolution. The first idea was something like Uber for questions: click a button and someone who knows Excel jumps on a call with you. Customer conversations showed that most answers are specific to each organization, so the team tried building an FAQ library from Zoom and Teams recordings. Much of that turned out to be garbage in, garbage out: out of date, or not the answer you would want to pass along.

The key insight was that people need to create content, but nobody likes creating documentation unless their boss tells them to. Guidde automated the manual parts people complained about, such as writing the script, editing and applying brand guidelines. Yoav recalls people blocking three hours in their calendar to record one video, only for it to go out of date two or three weeks later.

The product launched officially in November 2022. It took about six months to find product-market fit, and then growth took off. The more AI functionality Guidde added, the less time users spent creating and editing, and weekly signups jumped from 50 to 1,000.

How is Guidde different from Loom?

Yoav Einav calls Loom probably one of the best products he ever used, but says it leaves the hardest parts of video creation to the user. In his VP Product days, producing a handful of customer videos meant one person to record, another to narrate, marketing to approve the brand, enablement to write the script, and an editor to sync visuals and audio. It took about three months to get four or five videos ready, and they went out of date within six months to a year.

Most people creating videos do not know how to write the script, he says, and editing in Loom usually means re-recording from scratch. Customers also wanted analytics on which segments watched which parts of which videos, connected to marketing automation, sales automation and CRM tools. Guidde's focus is narrower than a generic video tool: how-to content for software such as Excel, Salesforce, HR systems, CRMs and ERPs.

Who gets the most value from Guidde?

Customer support teams are the most common Guidde users, says Yoav Einav. They reduce 20 to 30% of their support tickets, and when a customer does reach out, a video replaces the usual four or five back-and-forth questions, so tickets close faster.

The second group is customer education and learning and development teams, which create training and onboarding content for other departments. The third is product marketing, for feature announcements and how-to questions, such as how to set up one of hundreds of integrations. Yoav argues that capturing this tribal knowledge is critical: often only one person on a team knows the answer, and if they leave, the knowledge goes with them.

How does Guidde use AI under the hood?

Yoav Einav says the biggest question with every AI component is build or buy, and the answer depends on where your edge is. For text to speech, Guidde picked two of the best tools on the market but built its own algorithm that learns each customer's dictionary, so product, feature and customer names are pronounced correctly and improve as users edit.

For scripts, Guidde uses several models, including Claude, OpenAI and Llama, depending on the job: turning a PDF or PowerPoint into a video needs visual understanding, which calls for a different model than text to text. It also fine-tunes and enriches the context, for example with a software vendor's help center or a customer's own knowledge base, and feeds the model the account's best edited videos as examples so output improves over time.

Yoav also says the barrier has fallen sharply. Five years ago, a call center routing project for one of the largest US banks took two years and five data scientists to handle 5% of cases. Today one or two people with an LLM could build an initial solution in two weeks that handles 20 to 30%. What matters now is the business use case and building a moat that cannot be copied with a prompt to an API.

How should AI products prove their value?

Yoav Einav says most people treat generative AI as a productivity tool, but time saved is not something people tend to pay for, and it is not sticky. Guidde instead tries to quantify the business KPIs it improves. A support rep does not create videos for their own sake: if videos let end users solve problems without contacting support, that is the value.

Inside a new account, Guidde introduces that value step by step, like a nurture campaign: create content on day one, set up branding on day two, then editing features and analytics. Based on a user's first 48 hours, it segments them by persona, such as viewer, creator or manager, and adjusts the journey. Yoav's reasoning is that video is the closest thing to a human conversation, and far more effective than sending someone an article or a PDF.

Where is AI video headed in the next five years?

Yoav Einav sees two main trends: personalization and automatic updating. Language models will answer the same question differently for different people, and Guidde could produce variations of the same video for different recipients, changing the voice, tone, logo or level of depth. He also expects a copilot that detects changes in software and updates videos, first with human review, like a semi-automated car, before becoming fully automated. At the time of recording, he said Guidde had big announcements coming at Microsoft Ignite at the end of November.

Looking wider, he expects voice cloning, emotional text to speech and translation to remove cultural, time zone and language barriers. But he flags unresolved questions: who owns a cloned voice when an employee leaves, and who owns the IP when a model is enriched with other companies' content. For voice cloning and deepfakes, he thinks regulation from the US and the EU is needed.

How do Yoav Einav and his co-founder run Guidde?

Yoav Einav says the key to founding a company is finding the right partner, before you even have the idea. He and his co-founder have worked together since 2010. He also stresses founder-market fit: they picked a problem they had felt firsthand, since in his career the only thing that really improved user engagement was Loom videos, which were hard to create and keep up to date.

The name and color came from research. People remember names with a typo or a double letter, like Zoom or Fiverr, so Guidde has a typo and a double D. Red was chosen because it stands out among 20 logos at an event or on LinkedIn.

The team works across several locations, including San Francisco, New York, Kyiv and Tel Aviv, and has been remote first since its COVID-era start. Everyone, from junior QA to sales leadership, is expected to put customer needs first, and new employees join customer calls. At the same time, at least 20% of engineering effort goes to the founders' strategic direction, because users can describe their problems but not necessarily how to solve them.

Saying no matters too. Guidde declined requests to add LMS features like quizzes and courses, and chose desktop over mobile because most customers work in desktop apps like Excel, SAP and Tableau. Yoav's goal is not to make every customer happy, but to focus on the users who rely on the product daily.

How has Yoav Einav's role changed as CEO?

Yoav Einav says a founder should do whatever brings the most value at each stage. Early on he did QA, support shifts and ran R&D, then moved to product, then sales and marketing, and now customer success and expansion. He is still involved in product, but about 80% of his attention goes to sales, marketing, customers and people, building a structure that can support growth across time zones and cultures.

He has three kids, blocks Monday afternoons for family, and removed Slack, email and WhatsApp from his Apple Watch. He calls Guidde his fourth kid. He says he would probably earn more at a big tech company, but would not enjoy it, and that great companies typically take eight to 10 years: it is a marathon. His north star is for Guidde to become an everyday workplace tool alongside PowerPoint and Zoom.

What is Yoav Einav's advice for AI founders?

Yoav Einav's advice is simple: if you have a dream, go for it. To the common question of whether AI will replace humans, he answers that it will make people more efficient by taking over repetitive tasks, such as cutting highlights from a webinar or turning a transcript into LinkedIn posts, so they can focus on their own added value.

In his words

"Creating content documentation processes is something that no one really likes doing unless your boss tells you to do it" (11:50)
"A video in that sense is the closest thing to a human interaction" (24:57)
"Our goal is not to make all the customers happy but to make sure the majority of our focus is on the right user" (35:23)
"Being open to feedback and criticism is really key" (36:13)
"It will just make you focus on your added value and not on stuff that anyone can do" (48:18)

Chapters

00:00 Introduction
01:03 Yoav's background in AI and video
03:09 From engineer and product leader to founder
05:08 Customer conversations vs data
07:18 How AI has changed in five years
09:12 Launch, product-market fit and signup growth
10:17 The genesis of Guidde
13:05 Who uses Guidde
15:29 Guidde vs Loom
19:17 Build or buy: the AI under the hood
22:12 Showing enterprise clients the value
25:31 The future: personalization and auto-updating videos
27:54 Why red, and why the name Guidde
29:29 Co-founding and founder-market fit
31:33 Team, culture and staying customer-centric
33:52 Lessons: knowing what not to build
36:20 How Yoav's role has changed
38:22 Work-life balance as a CEO
39:49 What keeps him going
41:30 Guidde's north star
42:34 Predictions: personalization and voice cloning
46:02 Deepfakes and regulation
47:16 Advice for aspiring founders
49:13 How to reach Yoav

About the guest

Yoav Einav is CEO and co-founder of Guidde, an AI platform for creating how-to videos for software. Based in Israel, he joined his first startup in the summer of 9th grade, studied computer science, and worked as a software engineer and engineering leader in AI and video before serving as VP Product at an early-stage startup. Yoav on LinkedIn

About the host

Liam Lawson is CEO of The AI Report and host of The AI Report Podcast. Meet the team

Full transcript

Edited for readability. Quotes above are word for word from the recording.

Liam Lawson (00:12): All right, we are back for another episode of AI Tool Report Live. Today we have Yoav, who is Guidde's CEO and co-founder, on the podcast. Welcome.

Yoav Einav (00:23): Thank you, Liam. Great being here.

Liam Lawson (00:25): Yeah, I'm really excited about this one. We've obviously worked with Guidde for a long time now, over a year, on the newsletter front. Initially we were in contact with Moran, then we've done some stuff with Dan, and now we've got you on the podcast as well, so we're slowly working our way into the Guidde system as it stands.

Something that I really want to know about Guidde and about yourself is how you even got here in the first place. Before Guidde, what was the case? I was wondering if you could share a little bit about your professional background before Guidde even existed.

Yoav Einav (01:03): For sure. So first of all, I started my career as a software engineer, and I led engineering teams focused primarily, and not surprisingly, on AI and video: video streaming, video encoding, natural language processing and understanding. So a lot of the stuff we're doing now, but with LLMs and all the recent advancements, the whole ecosystem kind of changed. That's how I started.

That's also how I met Dan, my co-founder, back in 2010, when he co-founded a different company called Qwilt, with a Q. Qwilt is a video delivery network at the edge. Basically it serves all the online video in the US and in the EU, whether it's Disney Plus, Amazon Prime, ESPN Live or HBO Go, across the Verizon network, the Comcast network, the British Telecom network. What we did back then was distribute the content based on the watching characteristics of each household. So if you're a Liverpool fan and I like Game of Thrones, based on that we'd distribute the content to make sure you get the HD stream at home. That's how we met. I joined Dan as one of their first employees, and I worked there for about seven years.

After Qwilt, I was VP Product at a different early-stage startup, where I would use a lot of Loom videos for pre-sale, post-sale and onboarding. That was part of the vision for Guidde. Then when COVID started, Dan and I reconnected. When everyone was really stressed about the vaccine and the lockdown and everything around COVID, we saw a lot of opportunities in the market: how everything shifted to a remote-first workplace, and how we wanted to reduce Zoom sessions and make video creation more accessible to everyone in the workplace. I think that was the trigger for what we're going to talk about today.

Liam Lawson (03:09): At that point, when you're meeting Dan, I can see from your professional history that a lot of your experience is in product. I'm curious what made you switch out of being on the product and marketing teams and delve into that entrepreneurial journey. Or has that always been something you've been interested in?

Yoav Einav (03:27): So first of all, I'm based in Israel. Israel is called the startup nation, because that's what we do: we like to build stuff, we have a lot of ideas and we want to implement them. My first startup was actually in the summer of 9th grade. So I started my journey in the startup ecosystem really early on. I just joined a startup. I was on the support side, then they asked me to do some QA, then I started to learn how to code and started writing part of the features. So I taught myself into the engineering space, and then I also went to university and did my computer science degree.

I think a lot of what I'm doing right now is a combination of technology and customers, because when you join an early-stage startup, even if you're an engineer or QA or even a CEO, everyone is covering all the aspects. You can't be the second employee if you're not in touch with the customer. So a customer-centric or customer-first approach is something I'm used to from my engineering days and then my product days, and that's something that at Guidde is also part of our slogan: we're always talking to our customers and listening to their needs, and based on that we're building the roadmap and what's coming in the next quarter. I think that's part of what made me shift to product, and later on to found a company that is all about communication and video and improving customer communication.

Liam Lawson (05:08): I'd love to delve into the product side. You said that you have a lot of conversations with your customers to find out what they like, what they don't like and so on. From the product side of things, how important is quantitative versus qualitative information? How do those interviews compare against the statistics you're measuring? What's the split you take into account when you're shaping the product?

Yoav Einav (05:36): Cool. So I think at an early stage, in-depth conversations have a lot of value. As you start growing, you have more data, and the data also has a lot of insights. I think the right approach is a mix of both, because there's a lot of stuff you can see in the data but can't necessarily understand the reason for. Let's say you have churn. You can talk with the customer: maybe it's because of the pricing, maybe they didn't get the product, maybe they were looking for a feature that wasn't intuitive. So a lot of it is doing both. That's one thing.

The second part is that there's been a lot of buzz around GenAI in the last year, probably a bit more, and in most cases people look at GenAI as a productivity tool: it saves time. From the Guidde perspective, what we're trying to do is really quantify the ROI, the business KPIs that we're improving. We'll probably talk about it later on, but if I'm a support rep, I don't create videos for the sake of creating videos. I have a load of support tickets which I need to address, and if I'm able to reduce 20% of them by having my end users watch the content, understand how to solve the ticket and then not approach support at all, that's my angle. So I think it's always about how you use GenAI to really meet a business agenda, and not just save time, because saving time is great, but it's not something people tend to pay for, and it's not really sticky.

Liam Lawson (07:18): I think you have a really unique perspective, because you've worked in AI and machine learning for a long time, and most people in the AI space have only been made aware of it since ChatGPT came along a couple of years ago. Could you take me back to the first time you heard about AI, or the first time you were working with AI? What did it look like at that point, and what were the first applications you were working with?

Yoav Einav (07:45): So if you look at stuff we did five years ago, not even talking about 10 years ago, five years ago even something like transcription of a Zoom session was cutting-edge technology. Nowadays it's a commodity. I don't need to write a line of code to transcribe a Zoom call. Same goes for one of the projects I did with one of the largest banks in the US, which was automatic routing for the call center: based on the question, route it to the agent that can help you. Back then it was a two-year project with five data scientists, and you ended up addressing 5% of the use cases. Now, with one or two people within two weeks, using an LLM, if you know the domain, you can really provide an initial solution that will probably address 20 to 30% of the use cases.

So I would say the technology barrier is much lower and the time to value is much faster, and now it's mostly about the business use case: how you put it into action, how you use it, and how you create some kind of moat, so not everyone can copy and paste your solution in a second with just a nice prompt to an OpenAI API call.

Liam Lawson (09:12): And how long has Guidde been around for exactly? How many years?

Yoav Einav (09:18): We launched the product officially in November 2022, so basically next month the product is two years old. But I would say it took us about six months until we really figured out product-market fit, and then it started exploding and growing exponentially. Part of it was the AI features, which back then were a feature and now are the product. The more we incorporated AI functionality and reduced the time to create, the time you need to spend on editing videos, and really streamlined the entire process, brand, music, motion, suddenly the value was so big. We jumped from 50 signups a week to 1,000 signups a week, and it kept going like that, on this trajectory, month over month over month.

Liam Lawson (10:17): Got you. I was wondering if you could take me back to that time. You said that at the time it was a feature, not a product. So when you and Dan are sitting there with your whiteboards out, trying to decide what Guidde will be, what was the genesis of the idea? What was at the forefront of what this product or service would do?

Yoav Einav (10:36): Cool. So first of all, it's an evolution. Where we started is not where we are now. In the early days we thought about something like Uber for questions: you have a question, you click on a button, and someone who knows Excel jumps on a call and gives you an answer. Then we started talking to customers, and we understood that a lot of the answers are really specific to your business and organization. If I'm a sales rep at Check Point and I have a question on Salesforce, I need someone at Check Point to address that question. I can't ask someone from Palo Alto Networks.

So then we said, where can you find the answers to all those questions? In all the Zoom and Teams recordings. So we started seeing if we could build an FAQ library by extracting the information from those recordings, and we ended up seeing that a lot of it is junk. Garbage in, garbage out. A lot of it is not up to date, and the answer in the Zoom session is not necessarily the right answer you want to pass along to other team members.

So the key understanding was that people want and need to create content, but creating content, documentation and processes is something that no one really likes doing unless your boss tells you to do it. We wanted to make the process as simple as possible and remove all the friction: the manual edits, writing the script, making sure it's based on the company brand guidelines and the templates. So we took all the manual stuff people were complaining about and automated everything using AI.

One of the common complaints we heard about screen capture tools like Loom or Snagit is: I need to create a video, and it takes me at least five or six takes until I get it right, because in one take the page didn't load, in the other I mumbled a bit. People told us they actually blocked three hours in their calendar to record a single video, and then two or three weeks later the video is not up to date and they need to redo it. So we wanted to reduce that time, and make updating content as easy as updating a PowerPoint, and creating content as easy as taking a screenshot. That was our main goal.

Liam Lawson (13:05): With that, what customers do you see get the most value out of Guidde? Obviously you have some balance between enterprise customers and individual users. Are most people using it for internal team communication? What are some other use cases people get from Guidde?

Yoav Einav (13:19): Cool. So first of all, we see it across different geos, different company sizes and different verticals, so it addresses a lot of different challenges. But the common denominator, or the typical roles we're seeing use Guidde a lot, is first of all customer support teams. With Guidde they're able to reduce 20 to 30% of their support tickets. And if you do reach out to support, usually there are four or five back-and-forth questions, versus when they send the video: you watch it, you understand it and you close the ticket. So the time to resolve a ticket is much shorter. Customer support on software applications is a very common use case.

Second is customer education or L&D teams, for training purposes. Think about DocuSign University or Check Point Academy. In large organizations you also have learning and development teams whose whole job is to create content and help different departments with onboarding, with training, with adoption of new processes. So that's very common.

The last one is probably product marketing: new feature announcements, and a lot of how-to questions. If it's a software with hundreds of different integrations: how do I integrate Stripe with Microsoft Defender, how do I integrate Stripe with NetSuite. In many cases even the actual agents don't necessarily know the answer. They have thousands of integrations, most agents know five or 10 of those, and for the rest maybe you have one guy on the team who knows the answer, and if he leaves tomorrow there's zero knowledge. So capturing the tribal knowledge, as I call it, is critical, and being able to reduce the blockers to doing it is key here.

Liam Lawson (15:29): So it's about streamlining that knowledge creation and consumption process internally within these teams. By the sounds of it, you're more enterprise focused, with a lot of these teams using it. When people think about this sort of software, one that comes to mind is Loom, which a lot of individual people use. Could you go over the differences between your product and Loom, and where you excel? I realized you'd written a blog recently about some of the security issues that come with that, and how Guidde is very enterprise focused. I was wondering if you could cover that for me.

Yoav Einav (16:12): For sure. So first of all, I think Loom is probably one of the best products I ever used, and part of the vision for Guidde came from me as a Loom user. I used to create a lot of videos in my VP Product days, for the trial period, for onboarding purposes, for new features. It was always really challenging for the team to create the content. Sometimes the guy who created the video is not a native speaker, or maybe he's based in Texas with a heavy Texan accent. So we needed some guy in Israel or in Ukraine to create the content, someone else to do the narration, someone from marketing to approve the brand kit and the brand guidelines, like the fonts, the colors, the edit, someone from the enablement team to write the script, and then some sort of video editing tool, could be Camtasia or other tools, to synchronize the visuals with the audio stream, and then upload it to [name unclear] or YouTube, where usually only the marketing team had access.

So it became really tedious and lengthy. It took us about three months to have four or five videos ready for our customers, and then six months or a year later you're using the same videos, which are not up to date, because it's really hard to update the videos.

So in essence, most people creating content or video don't know how to write the script. They need help, and they're not necessarily the ones doing it. Then editing is another challenge by itself, because editing in Loom usually means re-recording the Loom from scratch. There's no real way to update part of the video. And on top of that, there's stuff customers were missing, like analytics. If I have a campaign, I want to know exactly which customer segments watched which part of the video, which videos, based on which features, and I want to connect it to my marketing automation tools like HubSpot, sales automation tools like Outreach, my CRM.

So making the creation easy, making the editing easy, integrating it into all the workflows with the analytics and the branding, that's what made a real change. And in our case, it's not a generic video tool. It's for how-tos on software applications: explaining to people how to do stuff in Excel, in Salesforce, in HR systems, in CRMs, in ERPs. That's why I think we're more laser focused on that, versus being a generic video creation tool.

Liam Lawson (19:11): Got you. So it's a lot more integrated into these different tools that everyone's using anyway, such as your CRM.

Yoav Einav (19:17): Exactly.

Liam Lawson (19:17): You'd also mentioned that AI quickly went from being a feature to the product itself. Could you delve into that? Do you use proprietary AI, or are you calling an API such as ChatGPT? What does the underlying software that runs Guidde look like?

Yoav Einav (19:36): For sure. So I think the biggest question is always build or buy. For everything you do with AI, does it make sense to build it, or is there an open-source or off-the-shelf solution that can provide it? The answer always depends on where your edge is. For example, for text to speech there are plenty of solutions in the market, from Google and GCP to ElevenLabs and others. So we picked two of the best text-to-speech tools there are, but we built our own algorithm, which learns the customer's dictionary and adjusts the text to speech for terminology like product names, feature names and customer names, based on your needs and based on your edits. As you edit the video, we keep improving the pronunciation and the way it's pronounced in different languages. That's a case where we picked something out of the box and improved it for our own use case.

Same goes for writing the script. We have different models we use under the hood, Claude and OpenAI and Llama, for different purposes depending on the use case. For example, if you want to convert a PDF or a PowerPoint to a video, you need visual understanding, so an image-to-text model is different from a text-to-text model, and so on.

And even if we take something like OpenAI, we do fine-tuning and we enrich the context. For example, if you create a video on DocuSign, we enrich it with the DocuSign help center, so it's not just what the LLM was trained on, it also has additional context. If you create it on Salesforce, we'll enrich it with your own internal content in your own knowledge base or LMS, so it's context specific to your own business. We'll also provide examples to the LLM. Let's say you have five videos you created on top of stuff, and you edited them and invested time. Now we can tell the LLM: those are the best examples, when you create the script, base it on those examples. So it improves over time and tunes itself to each individual account.

Liam Lawson (22:12): Got you. So it's very individualized as well. You're going for that level of personalization and understanding of each individual team and organization, which for someone with a Scottish accent would probably help massively, because communicating some of the words that I use is definitely a challenge at times, as people who listen to the podcast know.

Something I have noticed about Guidde as a product is that your product team is obviously on top of things, and this is probably credit to you: you're always bringing out new features, new AI features, different translations, stuff like that. With all these different features, do you ever find it hard to convey all the value Guidde offers? When you're talking to an enterprise client, what are the main things you speak to them about and highlight as the biggest needle movers for their internal organization?

Yoav Einav (23:14): I would split the answer in two. One, I think it's really hard to convey all the value in the first two days, so it's always a journey, like a marketing nurture campaign, where you give the user bits and bites of additional value. On the first day you just want the user to create content. On the second day you want them to invest a bit in the branding, because it will make their life easier and they'll probably share it with the team. Then you want them to see all the editing capabilities, like blurring private information or adding additional sources like PowerPoints and PDFs, and then you want them to see the analytics.

So you build those journeys based on your goals. Is your goal for the user to pay, to create as many videos as they want, to share them with as many people as they want? Based on the goals, you need to walk them through a journey. Usually, based on your first 48 hours, we segment you by what type of persona you are: are you more of a viewer or a creator, are you a manager or not. Based on that we define the journey that's the best fit for you, and over time it may change based on your activity. So it's something you learn on the go.

The other aspect is that the reason we created Guidde is that the best way to transfer knowledge is a chat. If I could talk with every user, I would jump on a Zoom call with everyone, they'd ask me questions and I'd reply specifically to their questions. So I think a video, in that sense, is the closest thing to a human interaction like we're doing now: someone explaining and showing, which is much more effective than written documentation, like sending someone an article or a PDF or a PowerPoint. So we use Guidde for our own feature announcements, in our support chatbot, for the onboarding process. Basically everything we do, we test on ourselves, and then you also get a sense of what works and what doesn't, and you can feed it back to the product team on our side.

Liam Lawson (25:31): With that, you're again saying that the north star you're looking at is streamlining communication. So my brain leads me to the question: where does the future of Guidde lie in the next three to five years? Is that dependent on how AI advances? Do you have a vision of where this will be in five years? Can we get any better than what we've currently got?

Yoav Einav (25:56): We can. I can tell you there are a few big announcements coming at the end of November at Microsoft Ignite that we'll be part of. It will blow your mind, what we can do. And I think there's still a long way to go.

If you ask me about three to five years into the future, there are probably two main trends I can see. First of all, personalization. You can already see the partnership between OpenAI and Apple starting to work on that. So you have a Liam LLM model and you have a Yoav LLM model, and for the same question you get a different response based on your characteristics, your past questions, who you are and what you've been doing. The more we advance, the more personal it is. In Guidde's case, we could create the same video in different variations based on who you send it to. For example, we can change the voice, we can change the tone of voice, we can change the logo, and we can change the level of depth in the video. So personalizing based on the creator, but also based on the recipient. That's one thing.

The second one: you see it a lot in R&D and engineering teams, with all the copilots and GitHub helping you create code. In our case, I think a copilot to update the content: automatically detect changes and update the video for you. And again, the question is whether it will be fully automated. I assume it will take time until we get there. It's like a self-driving car: you'll have a semi-automated car first, so it will update the content and ask you for review, and then the end game will probably be fully automated.

Liam Lawson (27:48): So with these new announcements coming out, when exactly are they coming out?

Yoav Einav (27:52): It's part of Microsoft Ignite, end of November.

Liam Lawson (27:54): End of November. All right, we'll have to release this podcast exactly at the end of November.

With that, I want to take things outside of Guidde the product for a second and talk about Guidde the journey. A question I always love to ask, because it's at the forefront, I can see it right now: why the color red, and why the name Guidde? Why was that what you went for?

Yoav Einav (28:18): Cool, that's a good question. I get it a lot. When Dan and I started, it was before ChatGPT, so we had to do it ourselves. We did huge research about what type of names people actually remember, and we saw that people remember names with a typo, and names with a double letter, like Zoom has a double O, or Fiverr has a double R. So we picked a name that is in our own domain, with a typo and a double D. And red is just because it stands out. If you go to an event, or go to your LinkedIn and see 20 logos, the red one, and sometimes the purple one, will stand out in the crowd. You can't miss it when you see 20 different logos. That was the main reasoning for it.

Liam Lawson (29:14): Yeah, it's interesting. Now that you say it, I think about a lot of AI companies and their logos and colors. There's a lot of green and blue. Those seem to be synonymous with the AI space in general, the colors people will opt for.

Yoav Einav (29:28): Yeah, for sure.

Liam Lawson (29:29): With those early stages of founding Guidde: you'd worked in product teams, and you'd been part of entrepreneurial startups from a young age. What was the process like of co-founding a company, deciding your different areas of expertise, and managing that relationship while actually running the company?

Yoav Einav (29:54): So first of all, I think the key in founding a company is finding the right partner. That's before you even have the idea itself. Dan and I go way back, since 2010. We worked together, we know each other, and in many cases we talk to one another more than we talk to our wives. So the relationship of the two founders is always critical.

Second, there's always the question of founder-market fit. Not product-market fit, founder-market fit. If you take companies that make DevOps software, probably one of the founders was a DevOps engineer. In this case, we picked a problem to solve that we had firsthand. In my entire career I used product tours and popups and different checklists and different ways to improve user engagement, and the only thing that really worked was videos with Loom. But creating the Looms and making sure they're up to date was really challenging. So we picked something we felt firsthand.

Then the next step is always finding the right team, and the continuous search for product-market fit until you find it. Which is sometimes a bit misleading, because it's a moving target. It's not a one and a zero, where you found it and then you zoom in. It keeps changing, and you need to adjust based on the market dynamics, the competition and the user feedback. So it's an ongoing task.

Liam Lawson (31:33): With that, it sounds like there are a lot of things changing. You're trying to find product-market fit, you've got two co-founders who know each other very well, and you have this idea in your mind of what you want the company and the internal team culture to be. How do you maintain that spirit of entrepreneurship and development as you start to build? And as part of that question, how big is the team now?

Yoav Einav (31:59): So the team is spread around. We have a team in San Francisco, New York, Kyiv, [city unclear] and Tel Aviv in Israel. Thanks to the fact that we started during COVID and were remote first in the early days, everything about the culture and communication uses video and Slack on a daily basis, so that works pretty well.

I think part of maintaining the flair is, first of all, having a really customer-centric, customer-experience state of mind. Today everyone at Guidde, from a junior QA to the VP Sales, puts their customers' use cases and needs first and foremost. For any new employee who joins, we try to have them join customer calls to get feedback, and to have the engineers talking to customers and getting feedback. The more customer-centric you are, probably the better fit you'll have. That's on the one hand.

The second part is that Dan and I always have the north compass of where we want to go, and we always make sure that at least 20% of our engineering effort is focused on the strategic direction that we think will bring the company to where we want it to be, even if it's not necessarily feedback you get from users. Users can talk about their problems, but they can't necessarily explain how to solve them. So it's mostly that you know where you want to go, you have the goals in mind, and you have the right team in place to make sure they both execute and have the bigger picture in front of them.

Liam Lawson (33:52): And it's difficult to stay on task when there are so many opportunities, especially in this AI and machine learning software development space. We've realized that as well. We've launched a million different features, some things don't perform as well as others, and some projects end up in the wastebasket and you'll never see them again. What are some lessons you learned from early challenges? Is there anything that comes to mind as a lesson, never do this again, never do that again?

Yoav Einav (34:23): Yeah, I think part of the decision-making process is also saying what you don't want to do, or what you're not. For example, sometimes customers try to push us to be more of an LMS: we need quizzes, we need courses. And we're saying there are plenty of LMS solutions in the market. That's not us, we don't have any advantage there. You can use Guidde and embed it in any LMS you have today, and we can recommend one based on our customers. Or for example, there was a customer that wanted us to support mobile. There might be a market there, but most of our customer base wanted us to support desktop, where they have Excel and SAP applications and Tableau Desktop and others. So we focused our agenda on desktop, and we just launched it last week. I think it's a big one for our customers.

So our goal is not to make all the customers happy, but to make sure the majority of our focus is on the right users, who really use the product on a daily basis. It's always the data, the feedback and the goal you want to achieve, and that always makes the decision process clearer.

Maybe another aspect is that sometimes founders are so in love with their own concept that they're not listening to feedback. I think something that's really part of Dan's and my DNA is that we don't have all the answers. Sometimes the best feature could come from feedback from our QA team or someone in support, not necessarily from us. Sometimes it could be a user who writes us feedback in a G2 review. So being open to feedback and criticism is really a key part of growing the product where you want it to be.

Liam Lawson (36:20): Absolutely, it's instrumental, and we've found that as well. Our readers at The AI Report have helped shape how we deliver that product, that newsletter, on a daily basis. It's been absolutely paramount.

I'm really curious about how your role has changed over time at the company. I can see that initially you were very involved in product. I think at one point your title was CPO. Recently you're doing a lot of public appearances. You did the live event with [name unclear] last week, and last month, and it seems like you're doing more public interviews. Could you walk me through the transition of your role in the company?

Yoav Einav (37:03): Cool. So as a founder and as a CEO, I usually do what brings the most value to the company. In the early days I was an all-around player. I was QA, I did support shifts, I managed the R&D team. Then as we started growing I moved to the product team, and when we grew further we brought in someone to manage the product team, and I focused on sales and marketing, and now on customer success and expansion. So at every given point in time it's based on the maturity of the company and where my focus is most critical.

Nowadays I'm still involved in the product, but I would say 80% of my attention goes to sales, marketing, customers and people: how to scale the company, how to build the people, how to make sure the company structure can support the growth we're at, especially with spread-out geos, different time zones, different cultures and different languages. You need to have the right people, experienced people with the right DNA, who are the right fit for us. That's where I'm mostly focused today.

Liam Lawson (38:22): For anyone listening, I assume they'll get the same sense that I do: you're deeply involved in the company, you live and breathe it. You're wearing the T-shirt right now, for anyone listening on audio only. How do you manage a work-life balance as a co-founder and CEO? Does that exist? How do you do that?

Yoav Einav (38:46): Yeah. If you ask most of the employees, they'll probably tell you I'm available 24/7. I have three kids at home and my wife, so I'm trying to balance it as much as I can. Usually I block my Monday afternoons to make sure I have time with the kids at least once a week, and on weekends. But again, it's a challenge for sure. When I bought my Apple Watch, the first thing I did was remove Slack and email and WhatsApp from the watch, because I said that would kill [unclear] for sure.

So it's a continuous target, but it's a good problem to have. I enjoy what I do at work, I enjoy the company, I love my kids, and Guidde is like my fourth kid. So I have four kids: three kids, and the startup that I started two years ago. That's how I look at it.

Liam Lawson (39:49): I feel as if you've already answered this in what you just said, but is that what keeps you going through all of this? Building a startup isn't easy. It's hard, there are loads of challenges, and there has to be a motivation behind why you continue to do what you do. Is that very much being family oriented, or are there other factors that keep you moving?

Yoav Einav (40:08): So, as I said at the beginning of the session, I've been working in startups since I was really young. If I worked at a big company like Google, Facebook, Cisco, whatever you name, I probably wouldn't enjoy my time. I'd probably work half as much, if not less, and probably earn 4x more, but I wouldn't enjoy the day to day. I always say to our employees as well that my goal is to have everyone in the company come to work with a smile and enjoy what they do. It doesn't necessarily mean every day is a great day. It's a roller coaster, there are good days and bad days, but as long as you believe in the angle and you believe in the people around you and the team, I think that's what makes me wake up in the morning and do what I do.

And for Dan and me, our goal is that we're building something big, and it takes time. With the best companies you know, there's no overnight success. The best ones take eight to 10 years, that's on average. So it's a marathon, and you need to have the time and the patience to do it right. That's a process you go through with the company.

Liam Lawson (41:30): You've obviously got an innate trust in that process, and you have a grand vision. You talked about this being a marathon. What does the finish line of that marathon look like for you? Where do you imagine Guidde being positioned in the global SaaS AI space in the next five to 10 years? What's the north star?

Yoav Einav (41:50): Cool. So for us, I would say the north star is to have Guidde as an everyday tool that every employee uses, in between PowerPoint and Zoom. You have PowerPoint and Zoom, or Teams, it doesn't really matter, and Guidde as the tool to create content in the business workplace. That's our goal. Today we're at an early stage, with early adopters, and we're growing fast. The day my mother and my kids use Guidde will probably be the day it's like ChatGPT. A year ago only a few people used it, and once our parents and kids use ChatGPT, you know it's fully adopted. So that's our angle with Guidde.

Liam Lawson (42:34): Got you. Being an absolute household name. With that, I want to take a macro focus to finish off. At one point during the podcast you spoke about a level of personalization: these fine-tuned models, where Liam's model might be different from your model. Do you have any predictions for the AI and machine learning space in general over the next five to 10 years? Where is this space moving, and what's going to succeed in the future?

Yoav Einav (43:09): Yeah. So if you look at all the new plugins and applications, I think we're starting to see a trend. You have the App Store, and now you have a marketplace for applications on top of ChatGPT, and that will become the new norm. There's a new ecosystem that is becoming huge every day, and the personalization aspect is key to it. If I live in Israel and Dan lives in the US, and I'm 40 years old and he's 50 years old, and I have two kids and he has two dogs, then based on who you are, the LLM should adjust. That's really different from today, where if you go to Google and search for something about security, whether you're an engineer or a sales rep you'll get the same answer. I'm not even talking about hyper-personalization. Today the same search term will mostly get the same response, besides some characteristics like your geolocation, your device ID and stuff like that. So the context itself, how much information it has on the person asking the question, will be key.

With that, we also see the advancement in voice cloning, and in text-to-voice models in general. With the new models you can actually adopt emotions: you can laugh, you can be angry, you can be sad. So combine voice cloning, where I can clone your voice with a text-to-speech model, use emotion, and now bundle it with translation, and you can really remove all the cultural, time zone and language barriers you have today when you need to talk with someone in Asia or France or Germany. It will open up a whole new kind of communication that we were blocked from until now.

But it also has a lot of compliance and legal issues as part of it. If I created videos with a Liam voice-cloning model and Liam left the company, can I still use it? Who owns your voice, you or I? So there are a lot of compliance and legal questions. Same goes for who owns the IP. I created content, and I enriched the LLM with content from YouTube and content from Intercom's help center. So who owns it: Intercom, Google, YouTube, or I? There are a lot of issues about copyright, ownership, who owns the IP and stuff like that, which are still not really resolved.

Liam Lawson (46:02): Do you foresee any solutions to these problems? Aside from those ethical considerations, we're also seeing an influx of deepfakes, for example Joe Rogan somehow promoting random supplements that don't exist, or someone starting a Shopify store. Is there anything on that side of things we can solve, or is that just something that's yet to come?

Yoav Einav (46:26): There are some tactical solutions. For example, for images there are solutions that make sure the LLM is only trained on images with copyright, and if you do use them, you pay the fees. But for stuff like voice cloning, I think it's mostly about regulation. The US and the EU need to have some sort of regulation around it, otherwise people can use it for the wrong reasons, not for the good of it. It's really a challenge, because you want to make sure people have access to the technology on one hand, but you want to make sure they don't use it for the wrong purposes.

Liam Lawson (47:16): There are a lot of challenges and a lot of opportunities, and that leads me to my final question. If you were going to give advice to a young person today who wanted to start their own AI company or SaaS company, what would you say? Would you tell them to start it or not, and if they were going to start it, what advice would you give them?

Yoav Einav (47:36): In my case the answer is easy: if you have a dream, go for it. That's a no-brainer. It really depends on what you want to do. I think the common question I get asked a lot is, will AI replace humans? And I keep telling people that AI will make your time more efficient. It will help you reduce repetitive tasks, like creating highlights from this webinar, or taking the transcript of this session and creating a series of posts for LinkedIn. So it will make your life more efficient. It won't replace humans, because it will just make you focus on your added value and not on stuff that anyone can do, like designing a logo or building a website. You can focus on the actual content and messaging, and where you bring value. That's where I see AI going.

Liam Lawson (48:34): I'm glad. It's a very optimistic message, and it's similar to what I tell people as well. I think it's never been easier to start a company, so just start. It doesn't matter what it is, whether you're writing, building a company or posting YouTube videos. I think the entrepreneurial journey is very rewarding and fulfilling, and that creation of value is something to aspire to.

So I want to firstly thank you for your time. I know you're a very busy person. And second, if someone wants to get in contact with you, or with people on your team, how would they do that?

Yoav Einav (49:13): Cool. So first of all, I'm available on LinkedIn. Whoever wants to connect and reach out on LinkedIn, I'm happy to. And second, I always tell people they can use the product and give it a try. We have a freemium version with the first 25 videos free of charge, and you can try out the paid version for two weeks free of charge. So go to our website, sign up, try the product, and we'd love to get your feedback.

Liam Lawson (49:47): Awesome. Thank you so much.

Get the five-minute AI briefing

Join 400,000+ business leaders who start their weekday with The AI Report. Subscribe free

Join the Newsletter
Inchide fereastra