Every job in the economy can be thought of as a bundle of tasks, shared between humans and machines. Over the years, software has picked off more and more of these tasks, but even today, humans still own the vast majority of business processes. In every function, headcount costs dwarf software expenditures by orders of magnitude.
Beyond Bots: How AI Agents Are Driving the Next Wave of Enterprise Automation, Menlo Ventures
A few months ago I wrote my first primer on VC Cafe about AI Agents, defining what they are and why they represent an attractive investment opportunity. Since then, we’ve seen approximately $1.5 billion invested in coding agents and co-pilots alone, with Google, Salesforce, Meta, and other incumbents heavily investing in Agentic AI. In this post, I aim to take stock of the AI Agent ecosystem, compare incumbents versus startups, explore areas of opportunity for founders building in this space, and look ahead at the future of AI Agents.
The AI Agent Ecosystem: A Layered Approach
Industry analyst Jeremiah Owyang published an interesting mapping of the 2024 AI Agent Ecosystem
The 2024 AI Agent Ecosystem framework provides a useful categorisation of the emerging layers in AI development:
- Ecosystem Layer: Foundational models (like OpenAI’s GPT) and enterprise tools form the bedrock.
- Application Layer: Thousands of companies will build agent apps, leveraging multimodal capabilities and no-code tools.
- Management Layer: Critical functions include managing AI agents, securing permissions, and enabling payments.
- Data Layer: AI agents will rely on both exclusive private data and public datasets.
Startups vs. Incumbents: The AI Battleground
The rapid development of AI agents is creating a fierce competition between agile startups and established tech giants. While incumbents like Amazon, Google, and Microsoft have vast resources and existing integrations, startups have the advantage of speed, focus, and the ability to attract top AI talent interested in building from the ground up.
When it comes to Generative AI, the incumbents are not resting on their laurels. For example, take Salesforce. Agentforce, a customisable AI agent builder by Salesforce is the world’s first and only platform deeply integrated with your data and apps. Their belief is that to build truly powerful agents, you need more than static PDFs and basic drag-and-drop tools.
Google is also leaning in on AI Agents. They’ve recently shared 185 real world use cases for generative AI in the enterprise
In a matter of months, organizations have gone from AI helping answer questions, to AI making predictions, to generative AI agents. What makes AI agents unique is that they can take actions to achieve specific goals, whether that’s guiding a shopper to the perfect pair of shoes, helping an employee looking for the right health benefits, or supporting nursing staff with smoother patient hand-offs during shifts changes.
In addition, there’s Sierra AI, a startup working in the AI Agent space founded by Bret Taylor, who’s impressive resume includes being the Chairperson of OpenAI, former co-ceo of Salesforce, CTO at Facebook, ex PM at Google, etc. Sierra’s mission is to ‘be the conversational AI platform for businesses’. The company raised an $85M “seed” round.
With this in mind, can startups even compete? Aileen Lee recently published an interesting piece on this topic. The short answer is yes, startups have a chance, but it might take time. While it seems like the big will get bigger, new companies will emerge with an ‘AI-first’ approach, but similar to the early days of the Internet, many of the early players in the generative AI space might not make it.
There are plenty of opportunities for startups in the AI Agent Ecosystem
Despite the challenges, the Agentic AI space offers significant opportunities for startups. A few areas of opportunity (not exhaustive)
- Agent-specific developer tools – agentic adoption could start with developers. There is a need for infrastructure that can seamlessly manage, orchestrate, and scale the hosting of agents.
- Building AI agents that can serve as decision engines – RPA, or rule-based automations, typically break, because they’re rigid and are limited in their ability to adapt to new situations, handle multi-step actions, demonstrate complex reasoning, or account for uncertainty.
- Addressing trust and privacy concerns – As with all AI development, it is important to consider the ethical implications of AI agents. The ability to offer customers safe, transparent tools can impact adoption.
- Agents as a service / no code agents – Offering pre-built, specialised agents for common tasks like UI automation, tool selection, and prompt engineering can significantly reduce development time and effort for businesses looking to integrate AI agent capabilities
- Browser Infrastructure – Providing robust tools and services that allow agents to interact seamlessly with websites and web applications, enabling them to access, navigate, and parse information just like APIs
- Personalised Memory – improving accuracy and access historical data. Developing solutions for managing agent memory and context, ensuring agents have access to relevant historical data and user-specific information for more personalised and accurate responses
- Auth for Agents – managing security and permissions. Creating secure and user-friendly mechanisms for managing agent identities and permissions as they interact with external systems on behalf of users
- “Vercel for Agents” – version control and agentic orchestration. Creating secure and user-friendly mechanisms for managing agent identities and permissions as they interact with external systems on behalf of users.
- Vertical AI agents – automate workflows in specific industries (say health or industrial). Can be niche markets where the large companies or foundational LLMs are not likely to compete with. This is particularly relevant for industries and departments that rely on manual, procedure-driven processes, such as customer support, recruiting, and security operations.
The current Agentic AI landscapes
By the time of writing this post, the AI Agent Directory lists 285 AI Agent companies, across 11 categories.
The latest curated mapping of the AI Agent landscape was published by Menlo Ventures on September 26 2024. Their analysis concludes that AI Agents are the evolution of RPA (Robotic Process Automation) by offering more adaptive, multi-step decision-making capabilities compared to traditional bots and rule-based automation systems.
Madrona Ventures published their own mapping of the AI Agents Infrastructure in June 2024. They note that agents still face challenges like scalability, memory, and reasoning limitations. As the Agentic AI field progresses, companies are developing tools and platforms to support the next wave of agent-driven innovation.
Dawn Capital’s AI Agent landscape refers to agents as productivity boosters (and in some cases replacement) for employees. I.e. Copilots for coding, marketing, customer support, etc.
- AI agents are expected to join workplaces, automating repetitive tasks and enhancing decision-making processes
- Marketing, accounting, coding, are already starting to get disrupted
- These AI tools will significantly boost productivity and innovation within teams.
Finally, the Insight Partners landscape, published in May 2024 (and featured in my previous post) offers a clear disctinction between fully autonomous AI agents and human in the loop agents.
These agents can interact with external tools, manage data, and use machine learning models to solve complex enterprise challenges, making automation more efficient and adaptable to changing environments
Final Thoughts: The Road Ahead for AI Agents (and a shameless plug)
As AI agents evolve from theoretical constructs to everyday tools, they represent a new paradigm in the digital economy. The race is on between incumbents and startups to dominate this emerging field, with a unique window of opportunity for innovation, particularly in agent application development, data management, and security.
Success in this space will require a deep understanding of the AI ecosystem and a focus on specific, actionable use cases. As we move forward, the next wave of billion-dollar companies may well emerge from today’s startups – especially those that can navigate the competition and create solutions for tomorrow’s autonomous AI-driven economy.
For VCs and investors, keeping a close eye on Israeli AI startups could yield significant returns. The country’s unique blend of technical expertise, entrepreneurial spirit, and cross-industry innovation makes it a hotbed for potentially game-changing AI agent technologies.
Shameless plug: At Remagine Ventures, we’re first cheque investors actively backing founders with deep passion and expertise in this growing space. Many companies in the Agentic AI space choose to remain in stealth, but just know that it’s never too early to speak with us. We’d be happy to be a sounding board, provide feedback and help where we can, regardless of investing.
I'm a former general partner at google ventures, head of Google for Entrepreneurs in Europe and founding head of Campus London, Google's first physical hub for startups.
I'm also the founder of Techbikers, a non-profit bringing together the startup ecosystem on cycling challenges in support of Room to Read. Since inception in 2012 we've built 11 schools and 50 libraries in the developing world.
- Finding Product Market Fit is More Than Just Checking Boxes - November 5, 2024
- The Emerging Israeli AI Landscape in 2024 - November 4, 2024
- Weekly #FIRGUN Newsletter – November 1 2024 - November 1, 2024
