Dhwani Soni

I am GVP of Product Management and Design at 8x8, and I work at the intersection of product, technology and human behavior.

Personally, I thrive in transformation, so it's exciting. I'm enjoying the bit where human behavior is coming to the forefront, with technology tapping into it in so many different ways that it's becoming ever more important to understand behavior and how we work through that. How am I feeling? As a leader in an organization, there's even more responsibility. There's greatness with AI: now you have sufficient evidence around things that you can see and measure, which before AI wasn't possible as much. Making sure we have enough evidence, and how much is enough or sufficient, is the judgment call that we as leaders have to make. But it comes with a responsibility. Just because AI has made quantity possible, the judgment still relies on a human being. What should be built? Why should that be built? How is it impacting a customer's outcome? And why would a customer trust you as a company versus company B? That's where we as product leaders have a sincere responsibility. And there are so many other things, not just this. Talent is getting compressed. Junior talent is being replaced by agents. Why? Because managers just can't keep up. They're asking, should I train a junior person or should I just give it to an agent to build it? So there are so many different things happening, and how we take responsibility to deliver AI responsibly, as a company and as individuals, starts to show up in our leadership.

Personally, I think we owe it to the next generation to make sure we have comprehensive programs, by companies or academically, where we sustain and support them growing up. We all know that what you learn academically is not the same as when you get into a company and apply those skills. You still need to understand the industry. You still need to understand how that company behaves. That's where we owe it to them. Programs like internships, apprenticeships and rotational programs are still extremely valuable to provide to the next generation. Those opportunities cannot be wiped away with AI, even if we tell them to build those agents. Having on-the-job experience should not be taken away by AI.

Agreed. But I think mentorship is different from teaching. Both are different skills.

When you're teaching, you are providing your principles and mindsets. As a teacher, you have different skills to manage a crowd, and you bring that storytelling ability to transfer a skill. Mentorship is slightly different. Mentorship is where you bring your own stories to a problem that a person is having, and have them live through a moment of, yes, this is possible if I change my perspective. So I think those are different skills.

You're absolutely right. Not everybody is cut out to be a teacher. Not everybody is cut out to be a mentor. I'm not cut out to be a teacher. I don't have that level of patience. But one-on-one mentorship, yes, I can, and I do. I believe in paying that forward.

I look for signal through the noise and tie it, like you said, tactically, to what we are trying to solve, and then try to connect that.

AI may not be new as a technology, but it's new in the hands of so many people, and the things you can produce with it are so compelling at times. Think about compelling demos, vibe-coded stuff, the content you just mentioned. I could have come up with that statement if I had more time. That lets people jump very quickly: let's do it, let's produce it, let's take it in front of the customer. Then you start to build a relationship with the thing you just produced, and then there are commitments from customers. So I think there's a moment to slow it down. Think about the gates that existed pre-AI: when a product is ready for a prototype, when you take it to a customer for research, how long it takes to productize. Execution has definitely compressed. But the time it takes to nurture that relationship, and the time it takes to work through the bureaucracy of audits and compliance to really productize something in the hands of a customer in a specific industry, those haven't accelerated yet. So we need to slow down and say, okay, we have something. Before we take it to the customer, let's make sure we set the expectation correctly.

I can answer it from the product side or from the operating model side.

If you think about the operating model pre-AI and post-AI, at the end of the day, what is an organization trying to do? It's trying to make money and stay relevant. How do you do that? By delivering meaningful customer outcomes. Because AI has compressed the time to code, you now spend more time connecting with the customer on their outcomes and defining those outcomes. You can only get ROI from AI if you can measure it, and AI has made that possible today. AI also makes it possible to use mathematical models at scale to figure out how to measure that. Now you can apply that to the operating model overall. If your end goal is supporting a customer's outcome, or the top-line P&L for your company, you work backwards. Things that were previously execution-focused were very manual. Think about how a product progresses from research to MVP to prototype, through the different gates: beta program, alpha program, content, tech writing, taking it to the customer, demoing internally, building that environment, and eventually solving for the customer's outcome. All of that can be generated with AI today. You can apply that lens from a factory-model perspective and automate all of it. Then you know very quickly where your outcome is stuck for the customer. Your context layer is built, your data layer is comprehensive, and you're running AI on top of it. I literally think of it as a light bulb moving around the London Tube map. That's you as a company, and where you are in delivering that outcome for that customer. And you can measure that today. We've gotten to KPIs that existed before but you could never measure. Think about organization-wide promise tracking for a customer: how many people in my organization committed so many different things to so many customers, and how have I scaled as a company to deliver them? AI can measure that today, if your data is connected, your layers are clean and there's not too much garbage in. And when you can measure it, you can put an ROI on top of it. Operating models used to be more factory-oriented: I am releasing this, and the next stage is adoption. Great, but today that's just a leading indicator. How are you fixing the customer's problem and the outcome they're chasing? That's where you spend the majority of your time now, with the quick early prototyping and vibe coding. And that's where the whole industry is shifting toward a more service-focused model with AI. That's where it has real value.

I won't dismiss that, because it's very new. But I also believe data will never be enough. Human judgment is what decides what is sufficient. We have so much data: financial and commercial data, CRM data, interaction data from recordings of how you work with customers, telemetry on how they use your product, and what people outside your company are saying about it. You can pull in all of that voice of the customer to understand the gaps and what they're trying to do. You will use AI to apply that judgment, but that judgment is definitely biased by whoever wrote the code. With AI, the asks are very easy. Help me optimize my cost. Help me get ROI on this thing. I got budget approval from my organization; can you help me scale my AI transformation journey? And you sit down with them and understand what that journey means for them. So yes, it's still early. Sufficient data, maybe, but not enough data. I think we're getting very philosophical.

I won't be able to describe it better than you did. There's a part of society that is trying to understand and just keep up, and it's exhausting. There's a part that is against it because it's not environmentally friendly. And there's a part that says you cannot ignore that this is going to happen. I think there's truth in all of them. Technology and science will happen before society reaches consensus. That is inevitable. We saw it with electricity. And at that point, as companies and as leaders, our responsibility is to do our best to make sure it happens rightly for society. Whatever you can do: the carbon footprint, transparency around what you're using, how you're using it and how it's implemented. We cannot just say, that's somebody else's problem, I'm just going to use it. That's ignorant, and it's not sustainable for society. Rules came in with electricity. The right expectations of companies started coming in. That will happen with AI as well, and all of us as companies have that responsibility.

I hear you, and I think the concerns I'm hearing are absolutely valid. Even the most hopeful and optimistic people, all of us, should have a backup plan, with AI or without AI, in life in general. I'm sure some people disagree with me. They'd say, no, you need to go all in, otherwise you won't achieve your goals. But we are at a point where not everybody understands what AI is capable of, and it's capable of a lot of things. How are jobs going to evolve? Should I be sending my kids to college? Should I get a backup job? How do I guide my kids? Those are fundamental questions, and they're still prevalent in the Valley too. We are at an inflection point, and we cannot disregard that this is happening. It is happening, and it is going to happen. To each their own in how they accept it. There's nothing wrong with having a backup plan. There's nothing wrong if a bunker is your backup plan, or a second job, or a second, third or fourth source of income. The point is that this is happening, and how you plan to accept it, and how quickly, is what matters.

Yeah, that's a very interesting way to put it.

Oh, that's a loaded question. I don't know where to start. Where we started was, have we gone through a change like this in our living history? As you mentioned, the pandemic. There were pandemics before, but we didn't live through them.

Correct. But the closest, I feel, was the arrival of electricity. Even that wasn't the same inflection point we're at with AI. Electricity was inevitable once you had it. We had the Industrial Revolution and everything that came after it. But it wasn't at the accelerated pace at which AI has brought us to this point.

You're right. The big difference is that thinking was not overtaken, if I may say that. But what it did overtake, which is similar with AI, is the workflows. In a moment of inflection, what changes is the workflows and how you adapt to the new world. If you apply AI to the workflows that exist today, you'll probably get a faster horse cart. But honestly, with AI, you won't have horse carts anymore. So what does the workflow look like in an AI world? Does it help us go faster? Sure. Is it more efficient? Do we even need to be involved? That's where AI is doing the thinking for us.

Yeah.

This goes back to how we started: what is our responsibility in all of this? Electricity came. Calculators came, and calculators were banned in schools. I think development, vibe coding and building products with AI will become part of the curriculum, but how much AI you use in your learning needs to be controlled. I go back to anthropology. We bring our biases to technology, and that's essentially what AI is doing too. All of us are feeding our biases into AI. That's where we preserve our judgment as humans. If a product is built by a human and tested by a human, they carry the same bias they pushed in when building it. We have to separate that out and use our judgment about what is right and what is wrong, even with the biases we picked up in childhood. That's what we owe the next generation: helping them lean into learning and reading. Yes, they can read through AI, but also actually reading books. We owe them the internships and apprenticeships we talked about, giving them the ability to experience things without AI, or with AI, but still learning from the ground up. That needs to be preserved, and that's our responsibility. Not just somebody sitting in Silicon Valley or at a company, but all of us today. We have a responsibility to the next generation.

I agree. And I also think we can do the stewardship, but it has to start with acceptance that AI is here and it's going to happen. Then we have to figure out the responsibility around it.

I don't want to make a blanket statement, but where I feel I need to preserve judgment, I would still use AI to do data crunching, look at evidence and provide that to me. But at the end of the day, the decision should stay with me. It can give me proposals, for sure, but I want to retain the judgment: is this the right thing to build? Is this what the customer is looking for? Is this what I would invest in as an organization? That judgment still needs to be preserved. Tomorrow, with the pace of AI, will that go away? I don't know. Probably. Hard to say. But as of today, I want to preserve that judgment. Give the right tools and the right mediums to everybody in my org, tell them what the outcome should be, and give them the autonomy to run with it. That's what I think leadership should do in the world of AI.

Here's an example. If you're doing a repetitive task and you have a way of doing it, you'd want to verify what sources AI brings and how it did the thinking. That's how you build trust. Even as a leader in design: how do you build trust with the end user of your product? You show them the thinking behind the AI. Then they can verify: okay, I trust this, I would have done it this way. Once they've done that, and it's a repetitive task, you can convert it into a skill that runs automatically. At that point you don't need to be babysitting it. It's done, and you can trust it. So I can't give you a blanket yes, but for repetitive tasks, once I've conquered one area where I applied AI and it worked, it fits my mental model. I trusted it, I verified it, and I can move on to the next one.

Yes, things change. You keep doing your checkpoints. But what you just described was a judgment call. A model or a skill that was working before isn't working anymore, so you go back and check. That's exactly what you did.

I think our responsibility, right or wrong, is that they need to hear all sides of the changes happening. We can instill our ethics and our thinking, but they should hear both sides, and at the end of the day they are their own people. They have to make their own decisions, and that's their life. I know it's very philosophical, but how long can you keep a mobile phone away from them? How long can you say no iPad? They're going to see things and talk, and their friends will tell them about Roblox or whatever. You can't shield them forever. The best way forward, at least in my philosophy, is to give them as much exposure as possible. Read, write, watch, whatever it might be. Try it with your hands, and learn.

I'm glad to be at an age where I can embrace this. I have enough years ahead of me to actually learn it, embrace it, use it and apply it. If I were much older, I don't know how much capacity I would have to embrace a change this huge. There will always be naysayers, but I think the answer is to keep an open mind, listen to everything, and do what you want to do.