AI’s future: What is broken, and what’s next?

Keith Shaw: It’s time for a reality check on the state of generative AI. Companies have had a few years to create projects and look toward the future. So what’s been working, what hasn’t, and what’s coming next?

We’re going to talk about this with a leading CEO on this episode of Today in Tech. Hi, everybody. Welcome to Today in Tech. I’m Keith Shaw. Joining me on the show today is Ketan Karkhanis, CEO of ThoughtSpot.

Welcome to the show, Ketan — it’s great to have you on.

Ketan Karkhanis: Great to be on the show, Keith. Thank you so much for having me. I appreciate it.

Keith: Before we begin, I wanted to start with a quick icebreaker question. You’ve worked at massive companies like Salesforce and also led small, scrappy startups. What’s the one thing you’ve learned about leadership that you might not have expected 10 years ago?

Ketan: That’s a very good question. I’ll tell you this — what I’ve learned the most, and I’m sure I’ll learn more along the way, is that authenticity matters. Authenticity in leadership matters more than anything else.

What I’ve experienced is that I may not always be popular, but I will be trusted.

Keith: That’s nice. That’s really cool.

Ketan: I think so. Some of these lessons I’ve learned by being inspired by other authentic leaders, and some I’ve learned through my own personal failures. But I keep coming back to authenticity and trust. That’s the one thing we all have to strive for.

Keith: Let’s jump into the big question about AI. There’s a narrative out there that AI is replacing entry-level roles at companies — we’ve seen this with coders, developers, and customer service reps. Are we going to keep seeing this in the short term?

And how do we get companies to move away from that mindset and start seeing AI as a tool for creating jobs, not just cutting them?

Ketan: Some of this is clickbait, some is for attention. And some of it is just companies — myself included — trying to figure out what the future workforce will look like. What skills will we need? How can we amplify them?

But I’ll tell you one thing: everybody needs to be laser-focused on their customers. If you see everything you do through the lens of your customer, you’ll make the right decisions. Take customer success and service, for example. AI can amplify the value tremendously.

With AI, I can provide higher levels of support, reduce response times by 25%, and free up my team from mundane tasks so they can focus on higher-quality customer conversations. That’s how AI can enhance customer experience. Keith: Right.

The reason I ask is because some companies seem to be bragging about their headcount reductions as proof they’re “winning” with AI — especially those selling AI products. It almost feels like they’re laying off staff to prove their tools work.

It’s a little cynical — maybe even a conspiracy theory on my part — but have you heard the same thing?

Ketan: You don’t need AI to cut jobs — let’s be clear about that. If a company wants to reduce headcount, they’ll find a reason, AI or not. I firmly believe companies can do more with what they have and enhance value.

At ThoughtSpot, we’ve used AI to free up capital and reinvest in the business. We’re growing — we have about 100 open roles right now. We’ve increased pipeline generation by 25%, boosted customer service output, and scaled without reducing headcount.

AI allows us to serve customers more efficiently and grow faster.

Keith: Do you get mad when you see headlines saying, “We’re going to replace half our staff with AI”? Or do you just dismiss it as clickbait?

Ketan: They might mean it, they might not — but a company’s narrative shouldn’t be about job cuts. It should be about how AI is adding value for customers. For example, many of my customers have hundreds of dashboards that no one uses but still must be maintained.

With AI, all of that can be replaced with a single agent. That’s a better experience for everyone. For us, AI is about making work better, not just cheaper.

Keith: You mentioned agents — what have you seen working in the real world, and where are companies still figuring things out?

Ketan: In our domain, AI is the new BI. We built our AI analyst, Spotter — think of it like ChatGPT for BI. It can answer any question about your structured business data. Instead of building dashboards, you just ask Spotter questions and get instant answers.

That’s game-changing because you get the insights you need in real time, without waiting weeks for reports. In the future, people won’t be creating dashboards — they’ll be creating agents.

Instead of asking for a “customer success dashboard,” you’ll ask for an “NPS agent.” Everyone will have multiple AI agents working for them.

Keith: Before the show, you gave me some great examples of how you’re using AI agents at ThoughtSpot — not just in IT, but in onboarding, testing, and even in your legal department. Can you share those so our viewers can better understand how agents might work in the future?

Ketan: Sure.

We’re not just an AI company building AI products — we’re an AI-first company. One of the first things we did was require everyone in the organization to complete generative AI training — engineering, product, HR, finance, facilities, even executive assistants.

Once everyone was trained, there was an explosion of ideas. We didn’t need a corporate mandate to “adopt AI” — it was bottom-up. For example, our legal team now has an AI agent that automatically redlines commercial contracts. That saves huge amounts of time and gives customers quicker responses.

In engineering, 30–40% of our code is now written by AI agents. That’s allowed us to innovate faster because engineers focus on solving complex problems instead of repetitive tasks. Most of our UI testing framework is written by agents, too.

When an issue happens in production, instead of manually sifting through data, we have an agent — nicknamed QA 007 — that does the analysis. We also have agents with names like Lumos. Our teams have fun naming them, but they’re serious productivity boosters.

Keith: Given that you name them, do you think of these agents as “digital employees” or just lines of code? You’re not giving them an Employee of the Month award, right?

Ketan: (Laughs) No, we’re not there yet. But every team has a group of AI “team members” now. My legal team has legal agents; my engineering team has engineering agents. Naming them makes it fun, but they’re still tools, not people.

Keith: I like that you call them “minions” — that makes me think of the Despicable Me movies. Maybe giving them names changes how people interact with them.

Ketan: It does, and we also embrace a “beginner’s mind” when working with AI. Don’t just force an AI solution into an existing workflow — ask, “What would this process look like if I had two agents helping me?” That clean-slate thinking leads to better solutions.

Keith: We talked earlier about onboarding. How has AI sped up the process?

Ketan: In sales, onboarding a new enterprise rep usually takes six months. We’ve cut that to four months. Instead of cramming all the training into a few weeks, we give them an AI agent — SpotGPT — that’s with them 24/7.

If they’re on a customer call, they can ask SpotGPT a quick question and get an answer immediately, even referencing the most recent call notes or product definitions. In engineering, new hires can use agents to instantly understand code without reading it line by line.

Even I use our legal agent to explain contracts and check compliance benchmarks. AI has shifted onboarding from a “big training block” to “instant, in-context learning.”

Keith: You’ve said AI helps you compete with bigger companies. How so?

Ketan: We’re competing with the biggest players in BI — and winning. AI makes each salesperson and engineer more productive. My sales reps can sell more and faster; my engineering team can ship more code without adding hundreds of hires. For companies like ours, AI is the great equalizer.

Keith: When you talk to other executives, do you find they’re also doing good things with AI? Or do you see common traps companies fall into when integrating generative AI into their workforce?

Ketan: We’re all learning from each other.

I speak regularly with other CEOs, and the conversation is always, “How are you using AI?” or “What’s working for you?” A big pattern I’ve noticed: some companies try to “boil the ocean” with AI — launching overly broad initiatives that try to do everything at once. Those usually fail.

The AI projects that succeed are tied directly to ROI. And ROI can mean different things — better customer satisfaction, lower operating expenses, increased productivity — it’s up to the company to define it. Another lesson I’ve learned from my time at Salesforce is that tactics drive strategy.

Move fast, but not recklessly. Iterate, experiment, and adapt solutions to your company culture. AI adoption isn’t just about technology — it’s also about people. You can’t separate the culture conversation from the workforce conversation. For example, our engineering productivity is high, but our engineering satisfaction is also high.

Surveys show they’re spending less time on tedious tasks and more time on challenging problems. That’s a win-win.

Keith: That makes me think — if a company doesn’t focus on culture, employees might feel AI is being forced on them, leading to resistance or even sabotage. A culture that embraces tools will benefit more than one that says, “We’re using this — deal with it.”

Ketan: Honestly, I didn’t do anything special here — my team deserves the credit. I believe most people will make the smartest choice for their future if you give them the right tools. The only thing I mandated was that everyone had to take the AI training — no exceptions.

The training was practical, with real use cases, and it sparked excitement.

Now, in leadership meetings, my general counsel will share what legal is doing with AI, my CFO will share finance use cases, my CMO will share marketing examples — and they’re competing in a fun way to out-innovate each other. This is a bottoms-up movement.

Leaders now include “What could we do with AI?” as part of their offsite agendas.

Keith: How do you avoid problems like software sprawl or “rogue AI,” where teams use too many competing tools?

Ketan: Never limit ideas. Never limit aspirations. But give them a clear process for implementation. We quickly created security and governance policies, along with a centralized process for introducing new tools. Most of the time, teams don’t need entirely new tools — they can leverage our corporate standards.

The key is speed. The governance process must be fast so innovation isn’t stifled. Initially, we made the mistake of focusing too much on “tools” rather than ROI.

Now, the conversation isn’t, “I want Tool X,” but, “I want to improve developer productivity by five percentage points — here’s how.” Tying every AI initiative to ROI and culture makes it work.

Keith: Many companies are dabbling in AI right now. What’s the next big wall they might hit? Will it be hallucinations, guardrails, governance, or something else?

Ketan: First, we’re still in the very early days. What I think many companies will face is the AI adoption gap — you’ll have a group of employees who use AI extensively and another group that barely uses it at all. That gap will exist even inside a single company.

It reminds me of 25–30 years ago when people had to be taught how to use the internet. Are we doing the same for AI? In many cases, no. The other big factor is what Gartner calls the trough of disillusionment.

Many AI tools are overpromising, and ROI will become the main measure of success. You’re going to hear the word “ROI” more often than “AI.”

Keith: Especially since companies have already been spending a lot of money on this.

Ketan: Absolutely — and I think we’re just getting started. The AI economy will be bigger than we imagine. Another wall we’ll hit involves enabling the workforce of the future while also using AI to drive societal impact.

We haven’t scratched the surface of what AI means for society — our engagement with each other, our communities, and beyond. At ThoughtSpot, our mission is to make the world more fact-driven.

Yes, we want to grow and succeed, but we also ask: how do we use AI to give everyone access to facts? The societal change that could bring is profound.

Keith: You just zoomed way out to a much broader discussion — like, who should “own” AI? Or should AI belong to everyone? Ketan: Exactly.

This is like the early internet — when we still said “HTTP” or “WWW.” We didn’t yet understand its full impact. We haven’t fully grasped AI’s potential to make society better — multiply the internet’s impact by ten and that’s what AI can do.

Keith: Of course, the internet brought both good and bad. Are we talking equally about both sides with AI?

Ketan: Humans are flawed creatures, so there will always be negatives. But we can’t stop progress because of fear. To me, AI’s potential to improve lives — through better healthcare, education, jobs, or even entertainment — is far more important than the downsides.

Yes, bad things will happen, but that shouldn’t be the headline.

Keith: Do you think other CEOs are thinking this way, or is this just a “Ketan thing”?

Ketan: It depends on the values of the company. Mission-driven organizations like ours see AI as a vehicle to fulfill that mission. For us, trust is our number one value.

I believe that in five years, people won’t just buy ThoughtSpot AI for its features — they’ll buy it because they trust us.

Keith: That brings up the trust question. Are we going to reach a point where business leaders let AI make major decisions? We’re already seeing agents that not only provide answers but also take actions. With multi-agent platforms emerging, how close are we to really trusting these systems?

Does AI need to explain its decisions before we can rely on it?

Ketan: AI has to earn your trust — period. Let me give you a simple example. You probably have a circuit breaker in your home, right?

Keith: Yes — and it shuts off automatically when too much power is being used. Ketan: Exactly.

You trust it to do that without asking questions. It’s purpose-built to make a decision and take an action. Now, scale that up 10,000 times. We already have agents at ThoughtSpot that take action.

For example, our Spotter AI agent can tell you what marketing campaign to run — and with one click, it can execute that campaign.

But here’s the key: I’ll often follow up with, “Tell me why you made that recommendation.” Spotter can explain the question it was asked, the data it used, and the reasoning process — right down to the algorithmic approach.

We’re using techniques like ReAct (Reasoning and Action) where the AI inspects its own thinking. Transparency and explainability are essential. Over time, as the system earns trust, you ask “why” less often — just like you don’t check your circuit breaker every day. Trust is earned.

It’s not given, and you can’t just label something “trustworthy.” It has to prove itself through consistent, explainable results.

Keith: Final question — do you think AI will create a big divide where the largest, richest companies reap most of the benefits? Or will it democratize access so everyone can benefit?

Ketan: I’m going to do everything I can to make sure it’s not just the biggest companies. Yes, large companies have resources and GPUs, but innovation has never been stopped by that. We’ve seen small teams — sometimes just a few people — completely disrupt industries.

What’s important is equalizing access to technology. Governments can help here, but I also believe that 10–20 years from now, the companies making the biggest AI impact will be ones we haven’t even heard of today. The cycle time is shrinking — it’s no longer 25 years between major shifts.

It’s closer to 10.

Keith: One more “final” question. How can business leaders separate genuine AI transformation from “AI theater,” where it’s all hype?

Ketan: Leaders are under pressure — boards, investors, media. It’s easy to get caught up in the hype. The key is to have a beginner’s mind.

Don’t ask, “What can I do with AI?” Ask, “How can AI enhance my customer’s life?” Your “customer” could be a business client, a community, or society at large. Start there, and the right AI projects will emerge.

I made the mistake early on of focusing too much on the tech itself — deploying cool tools and showing them off. But the real question is: is this helping my customers? Is it improving resolution times? Is it letting us innovate faster? That’s how you find the real value.

Keith: Ketan, thank you for sharing your insights — it’s been a fantastic discussion.

Ketan: Thank you, Keith, and thanks to everyone for tuning in. I appreciate it.

Keith: That’s all the time we have for this week. If you’re watching on YouTube, be sure to like, subscribe, and share your thoughts in the comments. Join us every week for new episodes of Today in Tech. Thanks for watching.