Good morning. Good afternoon. Good evening, everyone. Thank you for joining me today. My name is Soumya Sundaragavan.I am a senior director of product at Freshworks, where I lead the product strategy and vision for IT asset management platform under the fresh service suite of products. As a product leader in SaaS data and AI in the past two decades, I have navigated a complex journey of building and scaling complex AI driven solutions. Today, the topic that we are gonna look at is business value realization through AI product strategy. I'll cover how you set AI product goals, assess risks, define success metrics, and secure business commitment, all with real world examples that bring these concepts to life. So let's talk about how the AI market has been booming. Right? If you see, I'm showing in the screen here. I hope you guys can see it.

AI market is essentially booming at the rate of 35% CAGR. The AI market is projected to grow significantly in the coming years with a compound annual growth rate of 30 to 35%, estimated market to expand from $371,000,000,000 to over 2,000,000,000,000 by 2032. This really paints a picture of how AI product strategy will make all of us rethink and reshape our North Star goals, one that is aligned with business objectives, is ethical, and as well as value based. When we look at business value realization through product, these are the four things that I want you to keep in mind, clear AI goals that drive value and business impact, aligning the AI product goals with business objectives, assessing key risks and define the success metrics, and drive your organizational change by commitment to a long term investment.

When you build a goal driven AI product, how do you go about doing that? When it comes to AI product development, setting clear, aligned goals is paramount to ensuring that our efforts are focused and it can lead to tangible outcomes, so AI products should be directly tied to business outcomes. Ask yourself, how will this AI product drive business growth? How will it improve customer experience? How is it gonna enhance operational efficiency? I'm gonna give some real life examples to showcase this. Right? So at Spotify, a lot of you guys must have used Spotify, the goal of their AI powered recommendation engine wasn't simply to create a better algorithm. It was to drive the user engagement and revenue growth by personalizing listening experience. This alignment resulted in a 20% increase in user retention. Next is focusing on the problems.

AI is a tool. It's not the goal in itself, right, so AI is powerful, but it's important not to get lost in the technology. Focus on the problem you are trying to solve and how AI can be a tool for that solution. Instead of saying, I will build an AI model to predict customer churn, reframe it as I will reduce the churn by x percentage over the next couple of years by predicting at risk customers and proactively engaging with them. This type of approach grounds the AI solution in a very measurable outcome. An example here that I have put here is Netflix. Netflix focused on improving user engagement through personalized recommendation. Most of you who've used Netflix would see that. So that not just developed an AI model, it resulted in longer viewing times of their audience, and it reduced churn and aligned the AI product with the core business metrics.

And last but not the least, when you build goal driven AI products is set smart goals. Specific, measurable, achievable, relevant, and time bound because that will provide clarity and a way to track progress. An example here is in Amazon. They have an AI powered inventory management system. What they did is they aim to reduce the overstock and stockouts by 20% within six months. SMART goals by x percentage within time bound, six months. This goal was very specific, measurable, and tied directly to operational efficiency, driving cost savings and improving customer satisfaction. The next key piece in, business value realization is assessing your risks. Very important, right? Building AI products comes with inherent risks. Identifying and mitigating these risks is crucial because that's how you get long term success.

When deploying AI products, it's crucial to identify and manage the risks that could derail your success, so few things are data quality and availability. AI relies heavily on data, so poor data quality can lead to inaccurate models, while insufficient data can stall the product completely. So an example here is Tesla. When Tesla, which is an autonomous, driving system, it faced challenges around data diversity. They faced a lot of challenges, but what they started doing is they started leveraging massive real time data collection network from Tesla owners who were their beta testers. They were able to improve the model performance and safety, and then that way, they were able to ensure they had high quality representative data. Next is bias and fairness. AI models can inadvertently perpetuate bias, right, which can harm customer trust and it can lead to regulatory issues.

So let's say you want to build an AI based hiring tool, something that Amazon did. They built an AI recruitment tool. As a a product leader, you would need to assess for potential biases. You need to conduct audits and stress tests with different demographic groups. You need to have a proactive approach that can help identify and mitigate any biases that are there in your model. That way you can ensure fairness and compliance. Amazon did the same thing. Initially, their AI recruitment tool favored male candidates due to biased training data, but they recognized this issue, and they revised the system to ensure fairness and reduce gender bias. Third is model risk, the technology risks that comes with the model. Right? Models can degrade over time, and that's something we need to keep in mind.

This is because of changes in the environment known as model drift. So regularly monitoring and model training is very crucial. An example here is Facebook's content moderation system. In the past, we've seen that Facebook has encountered challenges with emerging forms of hate speech, And as user behavior evolved, cultural behaviors evolved, the model struggled to identify harmful content across different languages and contexts. What did Facebook do? They responded to this by retraining the model on a regular basis to adapt to any new trends, any new environments, and reduce the errors. Last but not the least is regulatory and ethical risks. As AI adoption increases, so do regulations around its use, particularly important in sectors like finance, health care, and education. Especially in regulatory environment where you face hurdle in the medical field, you need to have stringent medical standards and regulatory approvals, and that highlights the importance of aligning AI with legal and ethical standards. How do you define success now that you've identified the risks? The third piece of the puzzle is how do I define success for metrics, success metrics for AI products.

Right? Tracking the right metrics ensures that AI product delivers tangible value. It needs to deliver tangible value to both users as well as the business. So defining the success for your AI product goes beyond just saying, hey, this is the technical performance of the model. It's about delivering value to the business and to the customer. So the first one here is business impact metrics. Right? So business impact metrics focus on how AI product drives business outcomes. For instance, revenue growth, cost savings, customer satisfaction. These are all key indicators. I'll give an example, a real world example. In one of my previous companies, we had AI driven customer service chatbot, and what we did was we defined success in terms of reduction in average response time.

So we said my target is 50% reduction in average response time. Then we said customer satisfaction score. We want the target to be 4.5 out of five. Then we said we want x percentage. In our case, we said we want a 30% decrease in customer service cost. So we defined business impact metrics to drive the outcomes. In this case, it was average response time, customer satisfaction scores, and the customer service cost. Right? So similarly, you could do the same thing. Google did this in their AI driven search system, RankBrain. RankBrain was designed to improve result relevance and reduce the bounce rates, And this directly contributed to very high user engagement and ad revenue growth. Right? So making it as a key business driver. So, something to keep in mind. Right? Next is, okay, I have the business impact metrics.

What else do I need to focus on? Focus also is on the technical performance of the AI model itself. So accuracy, precision, recall rates, f one score, these are all the model performance metrics that you can look at. Like I promised, I wanna give a real world example here. Microsoft had a predictive maintenance AI tool, which achieved 95% accuracy, directly reducing their downtime and maintenance cost for their clients. So in this way, success is measured through precision recall, system uptime, and ensuring the product's technical effectiveness align with the business goals. Third is user engagement. You have business impact metrics, you have model performance, but how do your users interact with the AI product? That is very important, isn't it?

You need to know the engagement metrics that could include adoption rates, usage frequency, and retention. I give real world example because that really sticks with us. Right? So, a lot of you might have used Snapchat or versions of Snapchat. So Snapchat has a AI powered filter. It drove really high engagement through their daily active user, increase. Their, engagement started increasing with the filter usage. The frequency of engagement increased. The AI innovation directly impacted the user attention, and that way, they were able to drive the platform growth. So this illustrates as an example as to how engagement ref, metrics also reflect product success. The more the customers are engaged with your product, that also shows the success metrics for your AI product and how it can grow over time.

The last but not the least is operational metrics. This includes measurement of efficiency of your AI product. For example, it's deployment, uptime, processing speed, and scalability. So we set a for example, in one of the previous, firms that I worked for, we set a target for our real time AI recommendation system. We said it will process 1,000,000 users per minute without compromising on speed, aiming for 99% uptime. That type of operational metrics like this helps to improve the user experience and drive the satisfaction. Let's come to the last piece of the puzzle, priorities and business commitment for AI. You might have done all of this, but if you don't secure business commitment and drive alignment across teams, you cannot ensure that you'll have an AI product success.

So for AI products to succeed, they need to have clear alignment with business priorities, and it needs full commitment with the leadership. So first example is build cross functional alignment. So AI products require collaboration across teams, from engineering to data science to marketing. You need to ensure that there is alignment between these groups and the overall business strategy. I'll give an example of Apple's Siri. The development team of Apple's Siri included engineers, designers, data scientists, product managers, all of them, to ensure that the voice assistant system seamlessly integrated with the Apple's ecosystem. That's how the collaboration across teams brings about the success of the overall business strategy. Right? So this cross functional collaboration was crucial, and in this case, it made Siri a standout product in a competitive market. Next is organizational change.

AI often requires a cultural shift, a shift in mindset. You need to educate the teams about the potential of AI, and you need to encourage experimentation and learning. This also means addressing any potential fears about automation. I'll give an example here of Capital One. Capital One embraced AI beautifully with a strategic pivot to AI driven chatbots. So by educating both employees and customers, they ensured widespread adoption of their chatbot, and their chatbot, I think it was named Eno, which now handles millions of customer interaction, and it illustrates how organizational commitment fuels AI success. And I wanna repeat that statement. Organizational commitment fuels AI success. So last but not the least is long term investment. AI is not a one off project. Right? It's a long term investment. You need to make sure your leadership is committed to the long term nature of AI product development. Google's DeepMind is a classic example. It's an example of long term commitment to AI innovation.

Google has invested in DeepMind for over a decade now, resulting in breakthroughs like AlphaGo and AI powered health care solutions. The sustained investment in AI research demonstrates the importance of a long term vision in achieving, meaningful results. Right? I would like to leave you guys with a thought. Right? The key to success of any AI product is aligning everything. You need to align from strategy to execution towards a unified vision. So AI product development requires a very clear strategy, careful risk management. You need to look at the human angle of things. You should have measurable success metrics, and you need a strong business commitment. Some of the real world examples that I gave you from Spotify, Tesla, Amazon, and others, I hope I was able to share how these principles come to life in practice.

By aligning your AI products with business' objectives, by assessing the risks, tracking the progress with metrics, and ensuring a buy in, you all can drive innovation and business growth. I really thank you guys a lot for your time today, and I look forward to seeing you how you will lead, you know, in your organization with AI powered tools. Thanks a lot. I hope this was an engaging discussion. Thank you, Deepshika. Yes, please, Nirmen. I would love to connect and talk more as well. So, yeah, please feel free to, engage with me. I would love to hear from all of you what you are doing in terms of the business value realization in your companies through AI product strategy. Happy to have further discussions offline. Thank you to Women in Tech and everybody who has been part of this discussion. Appreciate it.