1Mby1M Virtual Accelerator AI Investor Forum: With Gaurav Chaturvedi, Partner at Kae Capital (Part 2)

Sramana Mitra: Okay. The reason I’m asking this question Gaurav is, so as you know, AI is easier to control in a more constrained mode. The more degrees of freedom you give it, the more it is difficult to control all the issues like hallucination and precision. AI has a tendency to make things up and go a little bit haywire. I’m very curious about what is going to happen when this is actually operating in real world scenarios and how are they constraining?

How are they preventing? Based on the conversations we’ve had here, people are already in the market looking for ways to control that out of whack behavior. So is this a conversation this company is having?

Gaurav Chaturvedi: Absolutely. I’m so glad that you asked this. That’s where their approach is very different. Most companies that have developed similar solutions have taken a purely agentic approach, where you provide input to an AI agent or digital worker, and then receive an output. In many cases, that output goes directly into the workflow or to the customer, which makes some customers uncomfortable, for reasons you’ve already mentioned.

SuperAGI, on the other hand, has designed software that is still used by a human worker as part of the entire workflow, and that’s where they’re seeing much more interest. In this case, for example, if an SDR wants to run a campaign, the SDR creates the campaign plan. The AI conducts research and presents the results to the SDR, who can then approve and set the cadence. In some cases, the AI will also set the cadence for the next steps. Throughout this process, human oversight is maintained—there’s always a human in the loop. So, this is built to augment, not replace, salespeople.

On the UX side, this approach provides a lot of comfort to customers since it enhances, rather than replaces, the salesperson. It also helps address issues like hallucinations, which you mentioned. On the backend, because they’ve custom-built the model and done extensive research, they’ve been able to significantly reduce hallucinations.

Of course, it still happens occasionally, as it’s a known limitation in AI technology right now, but hopefully that will improve over time. Having a human in the loop mitigates many of these issues.

Sramana Mitra: Good. In the coding story, you’re saying you’re getting traction from the open source community. What are they enabling the open source programmers to do?

Gaurav Chaturvedi: There are lots of use cases here. SuperCoder currently supports two or three languages and frameworks, mainly focused on front-end development. The community has primarily been using it for creation. Entrepreneurs are using a lot of fantastic products that are more like co-pilot.

SuperCoder’s approach is to build an end-to-end front-end application. Again, the key thesis here is keeping a human in the loop and empowering the developer. It’s more than just a co-pilot. Many users apply it to projects where they specify the type of front-end they need, or they provide a Figma file, and SuperCoder generates the front-end based on that. But there’s control at every step, allowing the developer to step in and take charge whenever needed.

Sramana Mitra: Great. So both your examples have highlighted the importance of the human in the loop. So let me ask you this question, is human in the loop part of your investment thesis in the kinds of companies, AI companies you are investing in?

Gaurav Chaturvedi: In a lot of cases, yes. Part of that is because we believe that technology right now is not in shape for all the reasons that you also highlighted – hallucinations, data privacy, data security, etc. We still don’t understand LLMs and transformers that well. It’s always better, and the customer wants output and outcomes. The customer doesn’t want a technology. We believe that at this point of time, it is better to be on the spectrum of human plus technology. At some point of time, it will start moving towards fully AI, but we are still far away from that.

We have also made investments in AI-enabled services or AI-powered services.

Sramana Mitra: I was just going to get to that. Good that you got there. That is the natural next question. So let me actually kind of establish that question so that people who are following this conversation can understand the thought process.

So if you remember the trajectory of the software as a service industry, we kind of came along in that industry with the do it yourself (DIY) key. So people sold software and the enterprises use that software to do stuff themselves. So it’s a DIY mode.

Now in the AI, especially with the emphasis on human in the loop, comes the question of do it for me (DIFM). Instead of DIY, if you can not only get the AI technology, but if you get AI enabled services to actually do the entire function using AI with trained people who understand how to use that AI and do that job function, whatever that in this case we’re talking about, in some cases we’re talking about coding, in some cases we are talking about sales, some cases we’re talking about customer support, whichever function it is, it is a trained set of people who understand how to use that AI. So it’s in a way it is AI enabled business process outsourcing is where the trend seems to be going towards. So Gaurav, now I’ve set it up, take it away and maybe give us some examples if you have invested in this category.

Gaurav Chaturvedi: Absolutely. I’ll add another dimension to it. Right now, enterprises have a lot of excitement about AI, there’s a lot of budget allocated to it. Every CXO is mandated to identify the use cases where they can use AI and they have budgets to back it up. On the funny side, they don’t know what could be the use cases. They don’t know how the technology works. They don’t know how to get that outcome.

What we’re seeing with our portfolio companies—who are focused on pure B2B software sales—is that a lot of solutioning is involved. In a sense, it’s evolving into a model where companies are saying, “Can I just give this to you, and you deliver the outcome?” So, it starts with a product pitch, then moves into solutioning, and eventually to services. This is the trend we are observing in our AI and B2B AI software companies. That’s the future.

This segment is part 2 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Gaurav Chaturvedi, Partner at Kae Capital
1 2 3 4