-
Orby, a generative AI startup, has raised $35 milion to automate business workflows with a large action model (LAM) instead of the typical large language model (LLM)
-
Orby’s platform automates tasks by observing user actions, eliminating the need for users to write code and fitting within the “no code” trend
-
The platform also uses generative process automation (GPA), which is different from traditional robotics process automation (RPA)
With $35 million of funding in its pocket, generative artificial intelligence (AI) startup Orby AI is entering the scene with a vision of business automation with little to no effort. Notably, the company’s AI approach uses a large action model (LAM), rather than the large language model (LLM) used by most AI platforms today.
Here’s how it works.
The Orby platform can build automation as it watches behavior across all apps, workflows and essentially, anything a user is doing. Orby looks at HTML, where a mouse or trackpad is positioned, what images are involved and what kind of repetitive task a user is performing. Then the platform generates each workflow with code, automatically.
Orby CEO and founder Bella Liu said the platform aligns with the emergence of a “no-code” enterprise technology category. Basically, making tools easily accessible for business users. But the problem is that most no-code tools, “they’re not really no code,” Liu told Fierce Network.
(Bella Liu, Orby CEO via Orby)
“Many tools in this generation, they say ‘no code’ but then when the business user really starts to use it, it still requires you to understand how to write software. You need to understand all the logic, the variables, arguments,” she added.
With Orby’s platform, the idea is that each user doesn’t need to write code at all, or really have to do anything differently than he or she normally does. The system simply watches and learns, then it builds an automation code itself.
The LAM approach
A key difference here is that Orby uses an LAM. Many AI models thus far are language-based and require a user to understand how to craft queries to extract information and insights. Popular generative AI (GenAI) applications that use language-based models include OpenAI’s GPT-3, Google’s Smart Compose and IBM Watson Assistant.
Newer to the scene, an LAM is action-based, and rather than relying on prompts, it relies on observation of actions and learning as it “works” alongside users. It delivers a cognitive assist to users in a way that LLMs and queries don’t do.
According to one analyst, theCUBE Research’s Shelly Kramer, the beauty in Orby’s platform is that the LAM can “watch, learn, automate and adapt.” Kramer noted Orby’s solution works alongside a user much like the app Grammarly. It’s there watching everything they do, learning their motions, behaviors, preferences, workflows, etc., just as Grammarly watches a user create written content and applies AI to improve that writing. Essentially, “it’s self-running.”
“It’s a larger, and more robust action model than a typical LLM and one with use cases that I believe will make it infinitely more attractive to organizations looking to adopt and infuse gen AI throughout their organizations,” she told Fierce.
Kramer said Orby isn’t the first AI platform to leverage a LAM, but the approach isn’t necessarily common, yet. LAMs are “easily” going to be the next big thing on the AI front, she added. “As you know, this is a crowded sector and I believe the many use cases of an LAM will be attractive, both from a development standpoint and a user standpoint.”
GPA is juice ‘worth the squeeze’
Orby’s work is also part of a market category called generative process automation (GPA), which is different from traditional robotics process automation (RPA).
RPA uses rule-based software bots to automate simple, repetitive tasks like data entry. GPA uses AI to handle more complex tasks that require understanding and creating content, like writing emails or making decisions. That’s why Orby is branding its platform a “no rules” AI.
Liu said implementing RPA can be time-consuming, often taking months or years for a single process. Additionally, RPA projects are costly due to the need for specialized developers, data scientists and IT support, leading some customers to feel that the “juice isn’t worth the squeeze.”
Keeping humans in the loop
Realistically, it should be noted that many of today’s AI use cases still need some sort of human oversight — and Orby’s is no exception.
For instance, once an automation is launched, if changes need to be made, it can be modified by a human at any point.
Because the nature of AI is learning through experience, the platform also might run into situations it has not seen before, Liu said. When its “confidence level is low,” Orby will send a notification to the user for confirmation, learning from that feedback.
“This is actually one of those things we really believe: keep humans in the loop,” said Liu, “I will say that the current AI technology is actually really good. We can deliver very good accuracy in workflow from beginning just by watching the user once, but it’s usually very hard to say that AI is perfect, because even humans are not perfect in that case.”
Competition to come
There doesn’t seem to be “a great deal of competition” for Orby at the moment, Kramer said. Although, some players might still be flying under the radar.
“Much like Orby has been in stealth mode over the course of the last year or so, I’m certain they are not alone,” Kramer added. “I don’t think it will be long before we see more conversation about LAMs and more activity in and entrants to the space.”
Both Microsoft and Salesforce are exploring LAM use, for example, and she’s “certain they are not alone.”
To date, Orby has $35 million funding in total. In June the company announced its $30 million Series A funding that was raised last year. Why the delay? “We were just really busy with our products and customers,” Liu said. “Now we’re really ready to announce to the world our Series A and all the traction that we have achieved so far.”
