Reviewing Menlo Ventures’ Take on AI Agents and Why a New Architecture is Essential for 80%+ Automation
In their recent article, Menlo Ventures beautifully outlines the evolving landscape of enterprise automation, focusing on the critical role that AI agents are now playing. Their piece captures several important insights that deserve acknowledgment, especially as we consider the future trajectory of automation technology.
1.Background of Automation Companies
The article traces the development of Robotic Process Automation (RPA) systems, which have been pivotal in improving productivity by automating simple, rule-based tasks. Companies like UiPath and Zapier were early pioneers, enabling enterprises to automate sequential workflows like extracting data from documents, moving files, or filling forms. However, despite these advances, RPA faces significant limitations that have constrained its scalability:
Inability to Handle Exceptions in Unstructured Data: While RPAs excel in structured, predictable environments, they struggle when exceptions arise, especially when working with unstructured or semi-structured data. This inability to intelligently process diverse data formats has been a persistent limitation.
Underdeveloped User Interfaces (UI): Another issue lies in the UI experience. Many RPA systems are not designed for interactions that require user input or decision-making. In fact, every SaaS platform can be viewed as a form of automation that depends on gathering information from users. RPAs tend to fall short here, particularly when automations require more advanced, user-friendly interfaces.
Lack of Integrations and API Support: Finally, the article points out that RPA systems often suffer from limited integrations. While API-based platforms like Zapier provide more stability, they still lack the necessary depth and breadth of integration to handle more complex software environments, often leaving humans to fill the gaps and perform the necessary manual interventions.
2. Evidence of Moving Beyond Rule-Based Automations
Menlo Ventures further highlights a crucial evolution: we’re now seeing automation systems that go beyond just rule-based frameworks. This is a significant advancement as more dynamic, AI-powered systems are emerging, capable of handling exceptions, adapting to new scenarios, and even making intelligent decisions. The ability to process unstructured data and deal with real-world complexity marks a new frontier in automation technology. It’s no longer about simple, sequential steps, but rather about automating end-to-end processes with greater autonomy.
3.Approaches to AI Adoption
Another strength of the Menlo Ventures article is the market map they present, which classifies automation solutions along two axes: horizontal versus vertical applications, and rule-based versus AI-driven systems. Horizontal platforms aim to provide general-purpose automation across industries, while vertical solutions focus on industry-specific workflows (e.g., sales, healthcare). The vertical shift is particularly noteworthy as more companies recognize that specialized, tailored solutions tend to perform better in specific domains, as they offer deeper, more reliable automation. At the same time, the shift towards full AI autonomy promises to push automation far beyond the limits of rule-based systems.
4. The Horizontal Platforms’ Struggle with Monetization
Menlo Ventures astutely observes that horizontal platforms, while ambitious, often struggle with monetization. Their lack of depth and user experience in specific verticals makes it difficult for them to compete with industry-specific solutions that can offer more value to customers. Although horizontal platforms aim to cover broad use cases, they often fail to deliver the level of detail and reliability needed for more complex business processes. As a result, many horizontal platforms are now pivoting towards vertical solutions, where they can provide more tailored experiences and, in turn, capture greater market share.
The Bold Truth: Current Architectures Aren’t Ready for the Future of Automation
While Menlo Ventures provides a strong analysis of the current landscape, we believe the fundamental issue preventing enterprise automation from reaching its full potential is the technical architecture. Shifting between vertical and horizontal approaches or adjusting go-to-market strategies will not address this underlying problem. The reality is, none of the current solutions—including those mentioned by Menlo—are designed to achieve the level of automation that the market truly expects: 80%+ automation.
Based on our conversations with hundreds of customers, traditional RPA systems, no-code platforms, and even code-based agentic solutions hit a ceiling at around 20% to 40% automation. Even with the advances in AI agents and novel LLM technologies, businesses still find themselves limited in the degree of automation they can achieve.
The problem is not in the ambition or approach but in the architecture. To get past this ceiling and reach the 80%+ automation level, we need a different architectural model—one that supports both autonomy and explainability, modularity, and citizen developer empowerment. This is where generative graph-based reasoning comes in.
What is Generative Graph-Based Reasoning?
Generative graph-based reasoning allows AI to autonomously solve novel problems, execute actions, and interface with external tools and APIs, while also generating a transparent, structured format called a Thought Graph. This format enables the AI to articulate its reasoning in a way that users can easily understand, modify, or enhance.
The three key characteristics of generative graph-based reasoning are:
- Autonomy with Structured Reasoning: The AI doesn’t just respond to queries—it generates a Thought Graph that outlines its reasoning in a clear, structured way before executing it. This ensures that even though the AI can autonomously act, its decision-making process is fully transparent and understandable.
- Human-Understandable and Modifiable: The system is designed so that humans—whether they are developers or not—can understand and modify the AI’s reasoning. This allows users to tweak processes, add specific workflows, or build out parts manually if needed, all while ensuring control over what the AI is doing.
- Empowering Citizen Developers: Perhaps most importantly, generative graph-based reasoning empowers citizen developers—people without coding expertise—to create complex workflows and applications using natural language and no-code tools. This modularity and flexibility mean even non-technical users can build sophisticated applications like custom CRMs or e-commerce storefronts, using the same underlying platform.
Our Place in the Menlo Ventures Taxonomy
When considering Menlo’s taxonomy of automation systems, which focuses on two axes—AI autonomy versus rule-based systems, and horizontal versus vertical solutions—we consider our platform to be a hybrid solution.
In terms of autonomy, we fall into a hybrid category with our generative graph-based reasoning. Our system can indeed handle novel, autonomous automation tasks while maintaining explainability. Citizen developers can understand, modify, and improve the reasoning for more robust, real-world use cases that demand sophisticated reasoning.
On the horizontal versus vertical axis, we don’t position ourselves strictly as either. Instead, we consider ourselves “verisontal.” We have a strong horizontal platform, supported by our Thought Graph architecture, that scales across various industries. At the same time, we’ve developed specific vertical applications like the Lexie Sales Assistant, which focuses on go-to-market personas, especially business development and sales teams. Within that vertical, we target specific sectors such as healthcare and crypto, creating tailored solutions.
The key to our approach is that the more we learn from vertical applications, the better our horizontal platform becomes. This creates a flywheel effect, where vertical solutions generate revenue and resources, which are reinvested to improve the horizontal platform and make it easier to enter new verticals. But our flywheel goes beyond just this dynamic.
The Flywheel: Empowering Citizen Developers for Faster Growth
What truly sets our approach apart is our emphasis on enabling citizen developers. While many platforms are trying to create a flywheel between their horizontal and vertical components, the real accelerant comes from empowering a much larger group of users. By enabling citizen developers, we multiply our platform’s capabilities by a factor of ten, a hundred, or even a thousand times.
This creates a much faster feedback loop between vertical applications and the horizontal platform because citizen developers can build, modify, and enhance the system without relying solely on internal developers. This broad base of contributors means we can enter new functions and verticals far more easily than a platform that relies exclusively on its own development team.
Furthermore, our hybrid generative graph-based reasoning supports full-stack applications, with rich UIs and autonomy. This enables us to build sophisticated applications, while also taking advantage of the generative and novel capabilities that Menlo Ventures highlights as critical for the future of automation.
Call to Action: Join Us in Shaping the Future of Automation
As we’ve outlined, the future of automation depends on having the right architecture, one that embraces generative graph-based reasoning and empowers users at all levels. While our Thought Graph development environment is still in its alpha stage and limited to our early customers, we are already making waves with our vertical solutions, such as the Lexie Sales Assistant, which is designed to augment business development and sales activities.
If you’re interested in exploring how our vertical solutions can boost your sales and business development efforts, we’d love to hear from you. Whether you’re looking for tools to streamline your processes or are intrigued by the power of Thought Graph, we’re ready to help.
Additionally, if you are a thought leader in the automation space, have feedback about this article, or believe there are interesting ideas worth discussing, we’d love to connect with you. Feel free to reach out to us—we are always excited to engage in thoughtful conversations and collaborations.
Contact us at:
- Maysam Lavasnai: [email protected]
- Nitin Naik: [email protected]
- Leily Moazami: [email protected]
We look forward to speaking with you and shaping the future of automation together.