The Future of AI Agents: Key Insights from Our CEO AMA

Last week, our CEO, Sean Blanchfield, hosted an extensive Ask Me Anything session on the Jentic Community Discord, covering everything from the future of AI agents to why SaaS might be dead. The discussion generated incredible insights about where the AI agent ecosystem is heading and how developers can position themselves for success. Here are the key takeaways from our community conversation.

Defining AI Agents: Beyond the Hype

One of the most persistent questions we encounter is: "What actually is an AI agent?" The term has become frustratingly overloaded, with everyone from automation platforms to simple scripts claiming the "agent" label. At Jentic, we believe an agent must have three core characteristics:

1. LLM-driven: The language model is in charge, making decisions about what to do next
2. Asynchronous: It runs independently in the background, not requiring constant human interaction like a chatbot
3. Tool-enabled: It has access to at least one external capability

The fundamental distinction is this: traditional software calls LLMs as a service, but with true agents, the LLM calls software as needed. This represents a shift from deterministic to non-deterministic computing and it is revolutionary. As Sean noted during the AMA:

"Deep research is an agent because it has search, runs away in the background and calls you when it's done. Amazing what you can do with just one tool!"

Why SaaS Is Facing an Existential Crisis

Perhaps the most provocative insight from the AMA was Sean's prediction about the death of traditional SaaS. The reasoning is compelling:

Most SaaS exists because humans need UIs to interact with databases. AI doesn't have this limitation. We just need to give it a database connection, and it can work directly with the data.

Agents replace entire workflows, not just individual tasks. As they do so, they eliminate the need for the specialized software that humans previously required to complete those workflows.

Competitive moats are eroding rapidly. What used to take years of engineering can now be built in weeks with AI assistance. Network effects get undermined when agents can interact with multiple platforms simultaneously. The "email work" that employs millions of people such as moving information between Slack, spreadsheets, and dozens of SaaS applications becomes low-hanging fruit for automation.

The venture capital implications are stark:

"VCs invest on 10-year timelines and can't see any growth opportunity for SaaS in 2035. And if it's in decline, there won't be any exits. So they won't invest in any new SaaS companies."

The AI Agent Infrastructure Market: Fragmentation vs. Consolidation

The community asked whether we are heading toward a few dominant platforms or continued fragmentation in the AI agent space. Sean's perspective was refreshingly contrarian:

Agents will unbundle existing market power. Consider how your agent could simultaneously shop across Amazon, Temu, AliExpress, and thousands of Shopify stores. The agent is tireless while humans are lazy. This unbundles the marketplace advantage.

Social networks face similar disruption when agents can parse LinkedIn spam for you, meaning you never need to visit the platform directly while remaining reachable everywhere.

Search remains a critical infrastructure layer, but even here, competition is intensifying. As Sean noted: "Maybe that just allows DuckDuckGo to step up, and API pricing."

Building in the Age of AI: Opportunities and Risks

The Opportunities

• Everything. Seriously. Every boring job and chore is fair game for automation
• Focus on outsourced work, especially Business Process Outsourcing (BPO) functions. These are the first targets for enterprise automation
• Data moats remain valuable, particularly tacit knowledge that can be transferred from humans to machines

The Risks

• AI development stalls before reaching true planning and reasoning capability
• Security becomes a nightmare with credentials leaking and prompt injection attacks evolving faster than our defenses
• The bubble bursts, taking good companies down with bad ones

Staying Adaptable in an AI-Native World

When asked how Jentic ensures adaptability as traditional software becomes obsolete, Sean emphasized culture over technology:

"I think adaptability is cultural. We're trying to build an AI native company. I think that means that the team is small and dedicated to building and maintaining a multi-agent system that does the work."

The key insight: AI doesn't need most of the software we have been building for the last 15 years. It can solve problems internally, generate single-use code, and doesn't need React frontends to use databases.

However, managing the communications layer, including discovery, authentication, and security, runs deep and enables AI rather than being replaced by it. This is precisely what attracted Sean to the agent infrastructure space as an entrepreneur.

The Challenge of Intellectual Property in an AI World

One of the more nuanced discussions centered on protecting intellectual property when AI can reverse-engineer and replicate so quickly. Sean's hot takes were particularly thought-provoking:

• Copyright doesn't fit in an age of generative IP
• Patents face uncertainty around AI-generated inventions
• Trade secrets become less valuable when software is nearly free to create

The practical advice: don't build your business assuming prompts can stay secret or remain inherently valuable. Instead, focus on data moats and capturing tacit knowledge that hasn't been systematically mapped yet.

Open Source Strategy: Building on Solid Ground

Jentic's commitment to open source is very pragmatic and not idealistic. As Sean explained:

"Software is on a path to becoming free. There's no point trying to build a company on a foundation of sand, pretending that the special software we have is magic in some way."

Our rule is simple: if it has public provenance, we open source it. If it has private provenance, like novel workflows discovered by customer agents using Jentic, it remains proprietary. This extends to our recently released Standard Agent library, which Sean described as an "anti-agent" or "anti-LangChain." The library is as simple as possible to demonstrate that building agents isn't the hard part. The entire codebase can be read and understood in 30 minutes.

Looking Forward: The Jentic Vision

Our long-term vision is to build a bridge between AI and APIs. First, by connecting agents to the real digital infrastructure of the world. This represents decades of existing code. Second, by mapping all the business logic needed on top of those APIs to get real work done reliably and securely.

This vision has crystallized into something larger. As Sean emphasized in follow-up discussions, the missing ingredient for reliable agent planning isn't more compute power or larger context windows: It's knowledge. Declarative, structured, machine-readable specifications of what actions are available and how to execute them.

Jentic aims at creating a shared public infrastructure layer for agentic knowledge. This includes not just web APIs, but CLI tools, configuration files, and agent-to-agent protocols, essentially any machine interface an agent might need to interact with.

This is how agents evolve from "bumbling around" to becoming fast, reliable, cheap, and secure: through access to a rich, openly accessible knowledge layer built on open standards like OpenAPI, Arazzo, and Model Context Protocol.

What's Next

The AMA reinforced our conviction that we're at a pivotal moment in computing history. The shift from human-driven software to AI-native systems isn't just changing how we build applications, it's fundamentally altering what needs to be built at all.

For developers and entrepreneurs, the message is clear: the old rules are breaking down, but the opportunities are immense. Success will go to those who can move fast, think differently about problem-solving, and build for an AI-native future.

Interested in diving deeper? Check out our Standard Agent library or our Arazzo Engine and join our waitlist for early access to Jentic-native agents that require no coding.

Have questions about AI agents or infrastructure? Join our Discord community where we continue these conversations daily.