1Mby1M Virtual Accelerator AI Investor Forum: Alex Benik, Encoded Ventures (Part 4)

Sramana Mitra: Let me ask you some of the broader trend questions and how you’re processing all this. As we discussed earlier, the market is currently in a human-in-the-loop, highly domain-specific and domain-expertise-oriented business space. The logical next step is going to be more automation, especially as agents become more active inside applications.

So how do you see that playing out on many fronts, including enterprise adoption, price structures, and so on?

Alex Benik: I think for each domain — I’m an infrastructure guy — if I think about infrastructure, there are many companies doing human-in-the-loop augmentation and potentially agentic replacement of data analysts, SREs, DevOps practitioners, and security analysts.

In business-critical systems, automation is not new, and there has historically been a reluctance to completely turn machines loose. That’s led to human-in-the-loop setups, where agents describe a possible security remediation but still require approval from a human operator.

It’s the same thing on the DevOps side. I think that phase will continue for a reasonable amount of time in infrastructure and business-critical operations.

For situations where there’s more margin for error, I think people will be more comfortable letting machines make autonomous decisions. It’ll be interesting to see how it all plays out.

Sramana Mitra: What is your assessment? You must be making some assumptions on the timeline.

Alex Benik: Things are accelerating so rapidly now, it’s hard to pin it down with any specificity. But in the phase of a new investment I make today at the seed stage — which, if a company is successful, could be a 4 to 7-year journey — I think it’s possible that we’ll make this transition.

People can build around the appropriate guardrails and build confidence in turning over low-stakes decisions to agentic operators, and then build that trust over time, just as you would with a junior engineer as they gain more expertise and experience.

That transition will probably happen within the scope of some investments I’m making now.

Sramana Mitra: In dinner parties, at least here in Silicon Valley, I’m having lots of conversations that point toward AGI or superintelligence in the next 5 to 10 years. That could throw a monkey wrench into this whole process if it happens.

Alex Benik: I’m not a big believer in the philosophical debate about what is consciousness or superintelligence or AGI. At the end of the day, these systems multiply a bunch of small numbers by each other — with incredible benefit.

It’s fascinating and it’s a great opportunity for founders and investors, but I don’t dwell on the semantics of consciousness and AGI.

Sramana Mitra: I’m not interested in the definition of consciousness either, at least in the context of startups. But there is a tremendous and fast-moving automation opportunity as the intelligence level of these systems improves dramatically. That brings a lot of questions.

Alex Benik: No doubt. Modern machine learning and AI systems have already surpassed human capability in many domains, and that will continue. The question is how fast that change happens and whether buyers feel confident turning over business-critical activities of varying severities and criticalities to agent operators.

Sramana Mitra: I’ll tell you how I’m thinking about it, and you can tell me if it resonates.

The first change will happen on the customer front. Right now, customers are much more comfortable with human-in-the-loop systems. I’m seeing a lot of case studies that are automating parts of a process — enabling humans to do more — without disrupting workflows.

But as customers get more comfortable and AI gets more powerful, that starts to become full automation and human replacement, not just augmentation. That change is inevitable and likely in the next 2 to 5 years.

Another big change is in software development. We’re seeing all these case studies of huge amount of software being built using AI — at the architecture, design, and coding levels. In the hands of really good engineers with a lot of experience, vibe coding is very powerful.

In the hands of amateurs, it’s not as powerful as they can produce a lot of fragile, low-quality code. But experienced engineers are getting major productivity gains. And that trend is going to continue.

Alex Benik: Absolutely. I see this across my own portfolio. I have a company in the security domain — today is actually their eighth birthday — and they’ve aggressively adopted modern code development tools. They built the entire front end with a halftime designer and Cursor. They reached their first customer purchase order at one of the fastest rates I’ve ever seen.

So yes, that accelerant is real. All of my companies are making use of it. And then there are questions, as you said, about code quality, debuggability, maintenance — all kinds of downstream issues.

But these issues have always existed. We have tools, and we’ll develop better ones to address them.

This segment is part 4 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: Alex Benik, Encoded Ventures
1 2 3 4 5