AI Strategy, Leadership & Future of Work with Tamar Yehoshua, Advisor & Former President at Glean

0:00I meet with a lot of CIOs and the CIOs are, so many of them are like, I have an edict from the board to have an AI strategy. Help me. What should my AI strategy be? And that's how I see the board's role, not as somebody who knows the strategy, but is keeping people honest of making sure they are doing it and that they're seeing tangible results.

0:28Welcome to Data & AI Mastery, the podcast where we bring you cutting-edge insights, practical advice, and inspiring stories from the leaders shaping the future of data and AI across the globe. I'm your host, Raoul Gabriel-Urmer, founder of Cambridge Spark, the leader in transformational data and AI upskilling, career development, and progression. In each episode, I will be diving into real-world case studies of companies harnessing the power of AI to drive innovation, reduce costs, and create new business opportunities. So whether you are an aspiring data scientist, AI engineer, or seasoned executive, this show is designed to give you the tools and knowledge to stay ahead in a world where data is transforming every aspect of business.

1:13Stay ahead, stay inspired, stay masterful. Welcome to Data and AI Mastery.

1:22Hi, Tamar. Good to see you. How are you doing? I'm doing great. Thanks so much for having me. Well, what a delight. I think we're in for an absolute treat. You have one of the most impressive backgrounds I've seen, so I can't wait for a discussion. Oh, thank you.

1:43I'd love to trace your background because you worked at some really impressive companies in a really impactful role so you were a VP at Google Chief Products at Slack you were a board member for several big companies as well like SNCC, ServiceNow and so on so I was going to ask you for those that have the ambition to have a successful care industry what sort of advice do you wish you had received earlier or would you give them today? The advice that I would give is definitely to follow people who you're going to learn from and follow people that where you're learning from the people who are the best at what they do.

2:23So if you want to learn about some kind of infrastructure engineering or sales, like go to the places where there's the best salespeople or the best infrastructure engineers, and then you're going to learn the techniques that are going to serve you well, whether or not the company succeeds. Not every company I went to did well, but a lot of them did. And the ones who did were like you were really learning from the best. And so that's what I encourage people to try and do. And companies that are aggregating good people, you also get a good network. My network is full of ex-Google people. And so all of us who work together, we're now all over the valley and we keep in touch all the time.

3:04We share learnings. And so that's a really powerful thing, the network that you're building throughout your career. And then also see where the trends are taking. So after Slack, when I went to Glean, part of it was, wow, AI is going to change how development is going. So I need to be in a company to experience that. So what are the things that are going to be important for you, like the Internet or Java programming or, you know, working with LLMs, whatever it is, like seeing where that's going to be relevant to where the industry is going. So in terms of your leadership experience, so, you know, VP at Google, CP at Slack and so on, what are sort of leadership principles you follow that maybe others can learn from?

3:56One is there's lots of different ways that people lead and be authentic to who you are. You don't have to say, oh, this is a good leader. I'm going to mimic them because if it's not authentic to who you are, it's not going to come across well. So be who you are. Be transparent so that people understand why you make the decisions you make. Because I think that's really important for people to follow you. They need to understand you. And so you have to communicate effectively and you have to be consistent in your principles of what's important to you. So when I was at Amazon, I was at Amazon when it was much smaller and I was very lucky to meet, have like quarterly reviews with Jeff Bezos.

4:39And one of the things that was really, yeah, it was amazing. I learned a lot from those meetings. And one of the things about him that really stood out to me was his consistency. Like he had his principles of what was important to him and you knew what he was going to ask you. And you knew what was important to him. So like the customers were very important. Like how is this customer, our customers going to view it? You knew you'd always be asked, like, what is the value to the customer? Every time. You knew that he didn't like icons. And it's like, what does this icon do? Because it wasn't very explanatory.

5:11You knew that if something wasn't relevant, he was going to get upset. And so I always knew how to prepare. And so then you know what the expectations are. And when your organization gets larger, if those principles are consistent, people can follow them more effectively. And then you don't surprise them. And with Stuart Butterfield, the CEO and founder of Slack, that was another thing when I got there. I spent a lot of time on the product principles to try and pull them out of his head and say, what are the things that are important to you? And we came up with our product principles that then we really kind of talked about all the time we made.

5:45special icons for them that represented each principle that we could use all the time. Every product review used the same principles. So that consistency helps drive whatever culture it is, whether it's frugality and customer focus at Amazon or technical depth at Google, then people can operate as the company scales. So that kind of transparency together with consistency, I think is really important.

6:17Now, I'd love to talk about AI in action. I mean, to kick us off, what are you hearing in the boardroom right now with AI? How, you know, Ned's approaching it. How does that look like? People are stressed. People are feeling like, OK, AI is coming and we need to make sure that we're not going to get become irrelevant. And we as in every company. And board members don't have any information that other people don't have. They read the same articles. Their lens is more of a long-term lens and a strategic lens. So that's really the big difference. But they don't necessarily have the answers. Board members don't say you should do this.

7:02A board member is like, how are you reacting to the new AI trends? They want to see that a company is taking it seriously. They're the push to say, wait, you say you're doing this, but show me what are you actually doing. And so in a company, it's easy to say, oh, I tried this. I tried that. It didn't work. I'm putting that aside. I've talked to many people who are like, yeah, you know, AI doesn't really work. I tried it for this one thing. it hallucinated. And so, yeah, it's all overblown. And they don't really try because they don't, they're scared. They're scared for their job. They're scared for the organization.

7:37They don't know what to do. So the board is there to say, no, you need to take this seriously. So in my job at Glean, I meet with a lot of CIOs and the CIOs are, so many of them are like, I have an edict from the board to have an AI strategy. Help me. What should my AI strategy be? And that's how I see the board's role, not as somebody who knows the strategy, but is keeping people honest of making sure they are doing it and that they're seeing tangible results. Like the first year of AI was just do something, like use a tool, say you're doing something. And it was a lot of prototyping and people weren't really looking at the value and the results.

8:22And now 2025, 25, it's much more people are demanding, well, you're spending this much money, where is the value? So now the board is saying, show me the ROI, show me the, how you, how you, it's not dictating how you measure, but they're saying, how are you measuring it? And show me that you're measuring it and show me that you're getting value from it. And they often also, most board members sit on multiple boards. So they often will bring learnings from other boards and they'll say, well, this board I'm at, I saw that they're doing it this way. X percent of their code is being generated. What about you?

8:59Super interesting. So let's dig into this a little further. How important is it for board members to be technical or illiterate? Because there's nothing worse for an exec or a CEO in this case to kind of get challenged and like, you need to do this, you need your strategy, but you have no clue what you're talking about. So how do you manage that? How can exec, I guess, magic expectation of the board and how accountable should board members be to actually understand the space? What's your view on that? So boards are usually made up of people with different expertise. So also it's very different between a public company and a private company.

9:35Public companies are much more fiduciary responsibilities, governance, where private companies is much more building of the strategy and trying to get to to be a public company. So the makeups of boards in Private companies tend to be more VC heavy and don't always have independence, whereas public boards, it's mostly independence. So you usually have one finance person, a legal person, a go-to-market person, a product person. So you try and get that set. So I absolutely wouldn't say that everyone needs to be AI literate on a board. But what they have to do is they have to be aware of the industry and where it's going, whatever your industry is in.

10:11And they have to be educated about that industry. They have to see what the peers are doing so that they can know that you're going to win and you're going to be ahead. I wouldn't say that the board numbers need to understand how foundation models are built or weights of different models or open source versus closed source. But they have to be versant enough to say what is your strategy and then to understand it and how it impacts the business. So are you using AI for customer support? If you are, they have to know to say, well, what's your resolution time and is it going down? So they shouldn't, again, dictate which metric you're using, but they should know to ask for a metric.

10:55And then I would say every board should at least have one board member who's AI, you know, knowledgeable and who can talk to the trends and who can know what the latest is to be asking how they're dealing with the latest. That's super insightful. So in terms of metrics, what are the type of metrics you'd expect your CIO and CEO to report back to the board around AI? You know, I'm hearing companies where they have a simple metric, which is what's the P &L improvements where, you know, AI has played a part where there might be some efficiency gain or some revenue creation. But, you know, as simple as that, do you think they should be more in-depth metrics or what would you expect to see in your board pack?

11:44For the board, it really is the same metrics that are your business metrics. What is what's the efficiency that you're using? Like how are you spending and what are you getting out of it? So I don't necessarily see like what's your go-to-market, what tools are you using for go-to-market, what's your close rate, what's the time to close because AI can improve all of these metrics. If you're using the BDR tools, like you can reduce your cost for getting leads or you can get more of them. So I think the business metrics don't necessarily change. It's just that you can radically shift the input to those metrics if you're doing AI well.

12:23Like, again, in customer support, in the resolution time, in code, how much code are you generating? How many engineers do you have? What is your efficiency that you're getting? But, you know, the most important thing is your cost, your margins, your cost for what you're accomplishing. So I think that board members might want to double-click to see, well, are you, like, if these numbers aren't where you think they should be or they're not improving from previous pre-AI eras, then to double-click and say, well, how are you measuring what's getting you there? How are you, what tools are you using?

12:59How are you shifting what tools you're using? Are you experimenting with new things? Do you have anything that you're experimenting with on a small scale that if it works and you can expand it, it's going to improve the margins over time? So the way that I look at the AI, there's the cost savings and then there's the revenue growth. Because there's a lot of things on the cost savings, but there's also new product lines that you could be developing with AI. So your company could be actually developing AI products, which then has more impact on your top line. So when you meet, you know, execs in the industry, you mentioned a few CIOs.

13:39What are the challenges that those individuals have today? One executive at a large bank that I, we were having a presentation on Glean, she said to me, I get so much inbound from AI startups. I'm inundated with these AI startups. They come to me and they say, we're going to do this for you. We're going to do that for you. She's like, they all sound the same. They're all saying the same thing. Like, how do I decide between them? And so that's right now with the overwhelming, every large company has an AI offering and says they do everything. So I think the biggest challenge is deciphering what's real.

14:16And so there's a pretty marketing slide or a pretty website that says, we will save you this much, or we were, you know, like we're going to be your automated customer support rep or BDR. And then you talk to somebody else and you're like, we tried it, but it didn't work and it had lots of issues or it didn't have the ROI that it had. So I think that's the biggest challenge. It's not that there's a lack of products out there. It's that there are too many and a lot of them are really early. And if they're early, it's hard to get validation. So what I told her and what I tell people is really what you need to do is talk to your peers.

14:51So there's a lot of forums right now of CIOs and CTOs or whatever function talking as peer groups to discuss their AI tools. And you want to hear from one of your peers, oh, we tried these 10 tools and this one really worked and we've implemented it. We've seen the ROI because now it's been enough time that there are products that have broken out because they're showing value. And so you want that proof. If you're like a large bank, you don't want to be like testing out like the first testers, the design partners for that. Maybe you have a like small team that's doing it. You can save that for like the more progressive smaller companies to be the design partners.

15:31So now if you are one of those smaller companies and you want to be a design partner, if you've got a big problem you're trying to solve, that can actually – so I know one, not a small tech company that wanted to solve a problem and was a design partner to a startup with four people. And then that tool became almost part of their extension of their development team. And then they developed it together and then the tool ended up doing well. And so that's the other extreme. So you can be a design partner to solve something that's really important to you and work with a startup to help them be successful, to show if it's a domain that there really isn't a great tool out there, but you know that something can be developed.

16:13So those would be my two ends of the spectrum. Invest in it and really tie yourself to someone or let other people do it and ask your peers and go with the tried and true. What other use cases are you excited about before we talk about Glean? I think there's, with new technology, it usually starts by automating existing processes, and then it goes to fully changing how you work. I think we're still in the first phase of, like, I used to write this spreadsheet, and now AI can write the spreadsheet for me. Or I would answer calls, a human would answer a call, now an AI can answer a call. Well, how do you get to the point where you're not even getting those calls in the first place?

16:54Or you don't even have to build those spreadsheets in the first place? and that's like the second phase. The second phase, like the first phase of mobile was we're going to take a desktop website and we're going to make it really small and put it on a phone, right? It's going to do the same thing. It's just going to be a different form factor. And then you got applications like Uber who wouldn't exist or Instagram that wouldn't exist without a phone. And we're taking advantage of that native capability. So I think that's where we're going with agents, that agents are now able to automate work.

17:28And that's what I think is going to be really interesting is instead of saying, I'm going to make BDRs better, I'm just going to skip the whole process. The sales process will change because I'm going to be able to build an agent that gets you what you need directly without going through all these multiple hops. So I think we're going to see some real changes in how organizations work. And that's harder for people to get their head around. And the newer companies are going to do it instinctively. And because they're starting out and they might, especially if they're like first time founders and they've never seen anything else, they're going to be like, oh, that seems better.

18:08I read a post by a guy I know from Google who is a who has a startup. He's like, I have a team. My marketing team is 100 AI agents. I have no one in my marketing team. And here's how I manage my 100 AI agents. Yeah. IT is the new HR. I think that's kind of like the digital idea, right? Literally managing agents. So I love your analogy. We don't even need the spreadsheet. We don't even need the phone call. So you remind me of a conversation I had with Renil. He's the chief data officer at Lloyd's Banking Group. He advocates something similar to what he said. There is kind of like incremental improvements, but there's also reimagination.

18:48So that's kind of like the phase two that you're talking about. So I guess if we continue in this phase of there's a new vision, we can reimagine how things work and agents play a big part to that. Can you tell us a little bit about, you know, Glean, you know, what's Glean about? Maybe we can start with that. Yeah, I'd love to. So Glean is a working AI platform. I'll tell you a little bit of the history of Glean because it gets you to where it is now. The founder and CEO of Glean is Arvin Jane. I knew him back in our early Google days, and then he left Google and started Rubric, which is a data security company.

19:24And when he was there, he found it really hard to find information. And as the company grew, it became very hard to find information. And so he's like, I know about search. I should build an enterprise search engine. And so he left Rubric and started Glean in 2019. So it started as enterprise search. crawl and index, like read all of your documents from your SaaS application. So Google Drive, Microsoft, Slack, Jira, Confluent, Salesforce, any product that you use in your organization, Glean would read those and then understand them and be able to find information. So think of Google search, but within your organization.

20:06So Glean spent like five years building a very robust enterprise search engine to be able to find information. Then when GPT-3 came along, ChatGPT added an assistant interface. So now you could, just like ChatGPT, ask a question, but of your enterprise. And think about if you're like a large enterprise with tens of thousands of people and you have all of this information, you might have acquired 100 companies over your lifetime and they're all in different places. And so now you can ask a question and find the most relevant information for your entire organization. So that's the assistant interface for Glean.

20:45And then add on agents, which we launched this year, which is now, think of Glean as the context of your organization and a knowledge graph of your organization. Now, what is an agent? An agent is something that has multiple calls to an LLM, multiple tools, and can take action. So now you can build a workflow that understands your company to take action on something that you need done. So that's the evolution of Glean. And the result is a work AI platform that understands the context of your organization and can help you be productive. Great, great. So if you're buying this vision of AI agents within an organization, then you need to really know how the organization works, how it operates, and all the internal knowledge.

21:32It sounds like that problem was solved beforehand, so now it's a great platform to build on top of. So I love that. So what are examples of how AI agents are being used in practice and what sort of problems can it solve for organizations? So it's a very horizontal platform. so you can build whatever agent you want. You can use the prompting for your assistant. So I'll go give examples kind of across the gamut. But, you know, engineers can use Glean for incident management. So you get an incident. How do you debug it? How do you find all the information that you need? Or you need to write a function and, like, find all of the design docs about this feature and function plus all the code that was previously written.

22:19And so it helps give you input to what you need and find all the customers that had issues or requests around that. So that's kind of for engineers. For product managers, a product manager needs to understand the context of the customer and customer requests. So it used to be that you had to talk to 100 customers. Now we still talk to customers. Like that's important to talk to customers. But it used to be that was the only way you had input. But then there are tools like Gong that record calls, and then you could listen to the Gong calls. Well, now you can write an agent that will go over all of the customer calls that you have recorded for the last year and say, what are all the features that customers have requested?

23:03And who has requested multiple features? And who is the person in your company that could tell you the most about this customer and put it in a spreadsheet? And now in like five minutes of building this agent and running it, you can get all of the information that would have taken you maybe quarters to go through and either listen to all these calls or talk to all these customers. So things like that are really changing how we do development. Another good example is product managers hate writing help documentation. So you can build an agent that will say, take all the design docs for this feature.

23:37Look at all the help documentation on our website. Write a help document in the format of the documentation on our website for this feature. Now, is it 100 % perfect? No. It gets you 80 % of the way there. And then you review it and you make small changes. So you're not going to, like, publish it directly. You're going to want to read it and review it. And then sales teams prepare me for a meeting. I'm going to meet with this customer. Tell me everything you know about this customer. Like find out what they talk about in the news. What are their biggest issues? What is their AI strategy? Who from our company has met with them before?

24:10How did the meetings go? Do we have anybody employed at the company who worked at this company? And prepare a briefing to brief me. I use Glean before I meet with any customer to say, to do exactly this, prepare me for the meeting. And if it's a prospect, it's more what I said. If it's a current customer, what issues do they have? Have they had any incidents recently? Are they waiting on any feature requests? So I have these agents that I've already built. And all I have to do is input the name of the customer and it'll brief me. So I used to have to like talk to the salesperson or the customer success manager and ask them all this information.

24:45And now you get like this really robust briefing. And I could, so let's say somebody else in the company, one of my peers wrote an agent to do that. I can use their agent or I can copy it and change it. And like maybe their agent, they're a marketing person and I'm a product person and I want different information. So then I can just copy it and edit it for me. And then I can share it with my team. So you can build agents and share it so that they could use it. So this is just kind of gives you an idea of the agent can be searching the web. It can be searching your internal files. It can be looking at tickets.