Another generative AI startup has raised a significant amount of funding, and like the others before it, this one has high hopes.
Emergence, whose co-founders include Satya Nitta, the former head of global AI solutions at IBM's research division, emerged from stealth on Monday with $97.2 million in funding from Learn Capital and a credit line totaling more than $100 million.Emergence claims to be building “agent-based” systems that can perform many of the tasks typically handled by knowledge workers, in part by routing those tasks to first- and third-party generative AI models like OpenAI's GPT-4o.
“At Emergence, we are working on many aspects of the evolving field of generative AI agents,” Emergence CEO Nitta told TechCrunch. “In our R&D labs, we advance the science of agent systems and approach this from a 'first principles' perspective. This includes critical AI tasks like planning and reasoning, as well as agent self-improvement.”
Nitta says the idea for Emergence came to him shortly after he co-founded Merlyn Mind, a company that developed virtual assistants for education, and realized that some of the techniques developed at Merlyn could be applied to automating workstation software and web apps.
So Nitta recruited ex-IBMers Ravi Koch and Sharad Sundararajan and founded Emergence with the goal, in Nitta's words, “to advance the science and development of AI agents.”
“Current generative AI models, while good at language understanding, still lag behind in the advanced planning and reasoning capabilities required for the more complex automated tasks that are the source of agents,” Nitta said. “This is what Emergence specializes in.”
Emergence has a very ambitious roadmap, which includes a project called Agent E, which aims to automate tasks like filling out forms, finding products on online marketplaces, and navigating streaming services like Netflix. An early version of Agent E is already available, trained on a combination of synthetic and human-annotated data. But Emergence's first finished product is what Nitta describes as an “orchestrator” agent.
The orchestrator, which was open-sourced on Monday, doesn't perform tasks itself. Rather, it acts as a kind of automatic model switcher for workflow automation. The orchestrator considers the task to be performed (such as creating an email), taking into account things like the model's capabilities and the cost of using it (if it's third-party), and then chooses a model from a list curated by developers to complete that task.
An early version of Emergence's Agent E project. Image courtesy of Emergence
“Developers can add appropriate guardrails, use multiple models for their workflows and applications, and seamlessly switch to the latest open source or generalist models on demand without worrying about issues like cost, rapid migration and availability,” Nitta said.
Emergence's orchestrator appears to be conceptually very similar to AI startup Martian's model router, which takes in prompts for AI models and automatically routes them to different models depending on uptime, capabilities, etc. Another startup, Credal, offers a more basic model routing solution that's driven by hard-coded rules.
Nitta doesn't deny the similarities, but he's not too shy about suggesting that Emergence's model routing technology is more reliable than the others, noting that it also offers additional configuration features like manual model selectors, API management and a cost overview dashboard.
“Our orchestrator agent is built with a deep understanding of the scalability, robustness and availability required for enterprise systems and is backed by our team's decades of experience building some of the largest AI deployments in the world,” he said.
Emergence plans to monetize Orchestrator with a premium version that will be hosted and available via an API in the coming weeks, but this is just one part of the company's grander plans to build a platform that handles claims and documents, manages IT systems and triages customer inquiries by integrating with customer relationship management systems like Salesforce and Zendesk.
To this end, Emergence says it has formed strategic partnerships with Samsung and touch display company New Line Interactive (both of which are existing MerlinMind customers, which seems like a coincidence) to integrate Emergence's technology into their future products.
Another screenshot of Emergence's Agent E in action. Image credit: Emergence
What specific products will be available and when? Nitta said Samsung's WAD interactive displays and New Line's Q and Q Pro series displays, but didn't answer the second question, suggesting it's still early days.
There's no denying that AI agents are a hot topic right now: generative AI giants OpenAI and Anthropic are developing agent products to perform tasks, as are other big tech companies like Google and Amazon.
But beyond a big cash injection at the start, it's not clear what Emergence's differentiator is.
TechCrunch recently profiled Orby, another AI agent startup with a similar pitch: AI agents trained to work with a variety of desktop software. Adept was also developing technology along these lines, but is reportedly on the verge of bailout from Microsoft or Meta, despite having raised over $415 million.
Emergence is more research and development intensive than most companies — it's like the “OpenAI of agents,” with a lab studying how agents plan, reason and self-improve — and it draws from an impressive talent pool: Many of the company's researchers and software engineers hail from Google, Meta, Microsoft, Amazon and the Allen Institute for AI.
Nitta says Emergence's guiding principle is to prioritize openly available research while building a paid service around it, a strategy borrowed from the software-as-a-service sector, and he claims that early versions of Emergence's services are already being used by tens of thousands of people.
“We are confident that our work will lay the foundation for automating multiple corporate workflows in the future,” said Nitta.
Skeptics may be skeptical, but they don't believe Emergence's 50-person team can beat other players in the generative AI field, nor can they solve the overarching technical challenges that plague generative AI, such as hallucinations and the enormous cost of developing models. One of the best-performing agents for building and deploying software, Cognition Labs' Devin, only has a success rate of around 14% on GitHub's benchmark test that measures problem-solving ability. Clearly, there is a lot of work to be done before agents can navigate complex processes unsupervised.
For now, Emergence has the money to try, but in the future, as VCs and companies grow increasingly skeptical of the path to ROI for generative AI technologies, it may not have the money to try.
Exuding the confidence of a startup CEO who just raised $100 million, Nitta argued that Emergence is well positioned for success.
“Emergence is resilient because of its focus on solving fundamental AI infrastructure problems that deliver clear and immediate ROI for enterprises,” he said. “Our open core business model combined with premium services ensures a stable revenue stream while nurturing a growing community of developers and early adopters.”
You'll see it soon.