How Initialized founders view the future of human-in-the-loop AI


Less than a year ago, ‘agentic AI’ was an industry-specific term that, even for journalists covering the space, was hard to define. Flash forward to today: AI agents are not only part of the cultural zeitgeist, they have started infiltrating everything we do. 

As LinkedIn gets flooded with posts from workers being laid off and “replaced by AI,” we decided to vibe check founders in the Initialized portfolio to see how they are leveraging agents, and what they think the role of humans will be as the outcome.

As part of an ongoing Initialized dinner series, we invited a select group of founders across a variety of industries who have put AI agents at the core of their business to provide their perspectives. The question we posed was:

When is it appropriate for AI agents to augment humans versus replace them, and how do you see these agents transforming your industry?

Chrys Bader, cofounder and CEO of Rosebud, on agents in mental health:

The number one use case for GPTs today is therapy. While we firmly believe that AI shouldn’t replace therapists, there is clearly a need for self-reflection and insights that aren’t being met. The power of AI is that it can recognize patterns in large amounts of data, making it a perfect tool to help us learn about ourselves, and recognize our blind spots.

At Rosebud, we use agents not as therapists, but as tools to enhance the journaling process. Rather than replacing therapists, we work with them to leverage common practices and therapeutic approaches to optimize the experience and output of journaling. Research shows that journaling is an impactful way to enhance our mental healthcare, but the blank page can be a stopping point. Our agents help provide prompts to get users going, they help you set goals, and they can show you insights and patterns that help users grow.

There will continue to be attempts to use agents as a replacement for therapists. While we should provide better access and affordable ways to democratize mental health resources, we believe the best way to help people is to leverage this technology to enhance the work therapists are already doing. This is the reason so many therapists recommend Rosebud to their clients – the real work happens at home, every day.

Dmitry Alexin, founder and CEO of Handoff, on agents in blue-collar work:

When people think of AI replacing jobs, they are often referring to knowledge workers. AI isn’t generally replacing the headcount at SMBs, where the average company has one worker: the owner. Growing an SMB is difficult, with low margins and a lack of resources to integrate and manage software.

This is the power of AI agents: they are both the software and the operator in one package. AI agents are particularly beneficial for small businesses in the construction industry, where businesses typically struggle with operational challenges due to low margins and the inability to hire sufficient staff.

Our AI agents empower small businesses in construction, enabling them to “create headcount” without the costs associated with hiring. Consumers can instantly get quotes on construction, and on the backend, we facilitate the process end-to-end. We now have 10,000 monthly active users trusting us to handle over $6 billion in annualized construction projects, proving that agents are creating a renaissance moment for SMBs.

Rahul Raina, cofounder and CTO of TRM Labs, on agents in national security:

AI is shifting the bottleneck of what it takes to conduct criminal activity from humans to computers. Criminals can use AI to instantly spin up millions of personas for hyper personalized phishing, custom malware, or synthetic CSAM which fundamentally changes how we need to approach cybersecurity: more humans simply cannot fight this level of crime.

At TRM, we are deploying AI “defense agents” at scale to map illicit networks, triage threats, and surface early warning signs to help global law enforcement agencies move faster, trace complex laundering schemes, and focus on the highest-risk activity. Through a combination of blockchain tracing, AI-driven pattern recognition, and human expertise, we’ve assisted in cases ranging from deepfake investment scams and romance fraud to illicit cryptocurrency exchanges laundering ransomware proceeds.

Two weeks ago, we spoke to the United States Congress about the shifting dynamics of criminal activity due to AI. As threat actors scale up their schemes with AI, our collective defenses — across society, but particularly government and law enforcement — must scale up as well. AI will undoubtedly shape the next generation of financial crime, but with collaboration and innovation, it can also become the key to preventing and defeating those crimes.

AJ Asver, founder and CEO of Parcha, on agents in compliance:

$200,000,000,000 is spent annually on compliance, and specifically, on fighting financial crimes like identifying terrorist financing and money laundering. Between rising costs, regulatory complexity and the rate of change in this space, compliance teams are frequently the bottleneck for business growth as they are burdened with an immense amount of manual labor.

Parcha is building AI agents to accelerate those manual workflows, conducting tasks in minutes instead of hours, and providing recommendations to be audited by a human worker. Having agents research evidence not only dramatically reduces false positives, but it decreases the risk from BPOs (business process outsourcing).

Regulators will always require humans to manage compliance, and we don’t disagree. But if we want FinTechs to succeed, and institutions to sustain, AI agents can eliminate the bottlenecks that keep companies from scaling.

Ruben Harris, cofounder and CEO of OutRival, on agents in education:

We build AI agents for admissions teams at universities. We call them digital workers—because they act as digital labor, working alongside human teams to reimagine administrative operations in higher education. Once we prove we can drive enrollment, we quickly show ROI and expand from there.

Most schools don’t have a real contact center or outbound motion. They rely on email, non-personalized text blasts, BPOs, and outsourced marketing firms just to keep up. Our digital workers handle millions of personalized outbound and inbound calls, texts, emails, and voicemails every month — from recruitment to retention to alumni giving.

Zooming out: while some schools have record enrollment, many are facing the enrollment cliff. At the same time, policy shifts are forcing schools to reduce headcount— they’re expected to do more with less. If we want universities to focus on academics and the student experience, the infrastructure behind them needs to be modernized. This isn’t just about surviving. It’s an opportunity to build something better.

And if America wants to win in AI, we can’t just invest in infrastructure and chips — we need to lead in talent and education too. That means empowering the universities that shape our future workforce to lead by example.

Higher ed is our beachhead. But if this works here—if we can transform one of the most complex and relationship-driven industries. We believe digital workers can be deployed anywhere there’s a contact center: education, healthcare, financial services, and beyond. 

Siqi Chen, cofounder and CEO of Runway, on agents in finance:

Within 30 minutes of leaving Sandbox VR as CEO, I was thinking about why we used basic spreadsheets for our financial modeling. It’s one thing to see the numbers, but it’s another to critically process how the numbers relate to each other and the broader business in order to make strategic decisions. While the pandemic gave me impetus to want a tool like that, I knew it would be helpful for CFOs of any organization.

Runway’s AI agents are built to help finance teams not only manage their spreadsheets, but to understand them. If you hold your mouse over any figure, our agents will provide a summary of the meaning behind that figure in order to see the bigger picture. These agents aren’t replacing workers, they are filling a gap that existing tools have never been able to breach.

We joke that Runway is the ‘finance tool you don’t hate.’ Agents can help make finance easier to understand, perhaps even slightly more fun. More importantly, when the next big event happens, our businesses will have a way to quickly predict impact and plan accordingly. For many businesses, AI agents could become the tool that keeps them alive.