AI copilots are increasingly the rage as the generative AI boom reaches its first-year anniversary (since the launch of OpenAI’s ChatGPT in November 2022). But even with companies from Microsoft to recruiting software iCIMS launching copilots — the question remains: how useful, reliable, and helpful will they be to end-users?
Layer, a new startup based in Columbus, Ohio, seeks to help other enterprises, particularly small-and-medium sized enterprises (SMESs) build copilots into their own software that are so good, the end-users can’t imagine life without them. Layer plans to do this by focusing on ensuring reliability and minimizing AI hallucinations (wherein the program generates inaccurate, harmful, or unwanted information).
And the company has some wind in its sails, announcing its $3 million seed round today led by Drive Capital, with additional participation from Resolute Ventures, Detroit Venture Partners, Alumni Ventures, Expansion Venture Capital, and several other funds and angel investors.
“What a copilot is, is it’s really like a personal assistant to the end user,” explained Layer co-founder and CEO Jonah Katz in an exclusive video call interview with VentureBeat. “It allows them to interact with a platform in plain English, they can just tell it what to do, and the copilot will do it for them, without a user having to like click and drag a bunch of buttons on the screen.”
In this way, Layer seeks to build upon the AI copilot trend already emerging in software more generally, but to broaden it to businesses of all sizes and their end-users, be they customers or employees, wholly transforming how people interact with software.
In a blog post announcing the raise, Katz wrote that “the copilot should be able to do anything a user can do on the platform (and often more) ten times faster.”
“The core challenge today when you try to build a copilot is around reliability,” Katz told VentureBeat. “It’s the fact that large language models (LLMs) hallucinate. And when you try to build a copilot, you’re basically plugging an LLM directly into your [software] product, so you can get into some pretty nasty situations.”
Whether it’s LLMs producing biased responses that exhibit sexism or racism or undesirable traits picked up from a corpus of human-created data by us flawed people over the many years of changing social mores, or delivering answers or performing tasks that are not what the user desired, enterprises looking to harness the power of generative AI face some surmountable and potentially legally dicey obstacles.
In order to minimize the effects of LLM hallucinations and undesirable behaviors, Layer has created a system wherein its AI copilot building platform ingests and analyzes the documentation of a client’s software and uses this as its grounding framework for deciding what actions it should, and should not take on behalf of the user.
“We have a sequence validator that basically checks ‘is this path valid?’” against the client’s software documentation, Katz explained. “If it’s not, then we either tell the LLM to generate the path again or feed it to a developer-chosen standard fallback procedure where the copilot tells the user, ‘sorry, we can’t figure that out.’”
Rather than try and do something that it thinks the end-user wants, Layer’s copilots are governed by whatever ruleset the client software developer has chosen, and won’t go beyond those boundaries.
Katz told VentureBeat that Layer was initially targeting “companies in the financial services” sector as its customer-base, but envisioned a world in which nearly every SME reliant on software could benefit from its LLM copilot infrastructure.
“The actual technology we’re building, from a purely technical standpoint, is pretty applicable across platforms,” Katz maintained. “Any company with a software platform, three to four years down the road will be able to use our technology to build their own copilot.”
For a specific example and inspiration to him personally, Katz pointed out that he previously used the business leads organization platform ZoomInfo for a job and found himself doing many repetitious and tedious, low mental-effort tasks that he figured could be automated, with a smart enough tool — like the kind of AI copilots Layer is seeking to help companies build.
“I would spend hours just working the product to get it to do what I wanted,” Katz reminisced. “I would, filter through a bunch of lists of leads, I would export them to a sales flow and put them into Engage and then send out a campaign. It was this super, like tedious process. The way I got there was just clicking and dragging a bunch of different buttons on my screen.”
But, this grind led Katz to the major source of inspiration behind Layer.
“If I had a personal assistant or a copilot, I could say, ‘hey, copilot, create me a sales campaign like the one I did yesterday and send it out to 20 people tomorrow morning,’ or some natural language query like that, and the copilot would just automatically execute all of those workflows for me with a complete understanding of the platform.”
Now armed with its new seed round, this rare Midwest-located AI startup plans to expand its hiring and continue building out its platform.
VentureBeat presents: AI Unleashed – An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More
AI copilots are increasingly the rage as the generative AI boom reaches its first-year anniversary (since the launch of OpenAI’s ChatGPT in November 2022). But even with companies from Microsoft to recruiting software iCIMS launching copilots — the question remains: how useful, reliable, and helpful will they be to end-users?
Layer, a new startup based in Columbus, Ohio, seeks to help other enterprises, particularly small-and-medium sized enterprises (SMESs) build copilots into their own software that are so good, the end-users can’t imagine life without them. Layer plans to do this by focusing on ensuring reliability and minimizing AI hallucinations (wherein the program generates inaccurate, harmful, or unwanted information).
And the company has some wind in its sails, announcing its $3 million seed round today led by Drive Capital, with additional participation from Resolute Ventures, Detroit Venture Partners, Alumni Ventures, Expansion Venture Capital, and several other funds and angel investors.
“What a copilot is, is it’s really like a personal assistant to the end user,” explained Layer co-founder and CEO Jonah Katz in an exclusive video call interview with VentureBeat. “It allows them to interact with a platform in plain English, they can just tell it what to do, and the copilot will do it for them, without a user having to like click and drag a bunch of buttons on the screen.”
Event
AI Unleashed
An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies.
In this way, Layer seeks to build upon the AI copilot trend already emerging in software more generally, but to broaden it to businesses of all sizes and their end-users, be they customers or employees, wholly transforming how people interact with software.
In a blog post announcing the raise, Katz wrote that “the copilot should be able to do anything a user can do on the platform (and often more) ten times faster.”
Addressing reliability as a ‘core challenge’
“The core challenge today when you try to build a copilot is around reliability,” Katz told VentureBeat. “It’s the fact that large language models (LLMs) hallucinate. And when you try to build a copilot, you’re basically plugging an LLM directly into your [software] product, so you can get into some pretty nasty situations.”
Whether it’s LLMs producing biased responses that exhibit sexism or racism or undesirable traits picked up from a corpus of human-created data by us flawed people over the many years of changing social mores, or delivering answers or performing tasks that are not what the user desired, enterprises looking to harness the power of generative AI face some surmountable and potentially legally dicey obstacles.
In order to minimize the effects of LLM hallucinations and undesirable behaviors, Layer has created a system wherein its AI copilot building platform ingests and analyzes the documentation of a client’s software and uses this as its grounding framework for deciding what actions it should, and should not take on behalf of the user.
“We have a sequence validator that basically checks ‘is this path valid?’” against the client’s software documentation, Katz explained. “If it’s not, then we either tell the LLM to generate the path again or feed it to a developer-chosen standard fallback procedure where the copilot tells the user, ‘sorry, we can’t figure that out.’”
Rather than try and do something that it thinks the end-user wants, Layer’s copilots are governed by whatever ruleset the client software developer has chosen, and won’t go beyond those boundaries.
Real world applications
Katz told VentureBeat that Layer was initially targeting “companies in the financial services” sector as its customer-base, but envisioned a world in which nearly every SME reliant on software could benefit from its LLM copilot infrastructure.
“The actual technology we’re building, from a purely technical standpoint, is pretty applicable across platforms,” Katz maintained. “Any company with a software platform, three to four years down the road will be able to use our technology to build their own copilot.”
For a specific example and inspiration to him personally, Katz pointed out that he previously used the business leads organization platform ZoomInfo for a job and found himself doing many repetitious and tedious, low mental-effort tasks that he figured could be automated, with a smart enough tool — like the kind of AI copilots Layer is seeking to help companies build.
“I would spend hours just working the product to get it to do what I wanted,” Katz reminisced. “I would, filter through a bunch of lists of leads, I would export them to a sales flow and put them into Engage and then send out a campaign. It was this super, like tedious process. The way I got there was just clicking and dragging a bunch of different buttons on my screen.”
But, this grind led Katz to the major source of inspiration behind Layer.
“If I had a personal assistant or a copilot, I could say, ‘hey, copilot, create me a sales campaign like the one I did yesterday and send it out to 20 people tomorrow morning,’ or some natural language query like that, and the copilot would just automatically execute all of those workflows for me with a complete understanding of the platform.”
Now armed with its new seed round, this rare Midwest-located AI startup plans to expand its hiring and continue building out its platform.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.
Author: Carl Franzen
Source: Venturebeat
Reviewed By: Editorial Team
