Adept raised $350 million in a Series B round to develop AI that can act inside existing software. The startup’s goal is to translate natural-language instructions into digital actions, positioning its system as an “AI teammate” for people who work with computers.
From generating content to taking action
Many prominent AI systems focus on producing text or images. Adept is studying how people browse the web and navigate software, with the aim of training a model to carry out tasks across a range of tools and APIs.
The distinction is practical: rather than asking a person to use an AI-generated answer as a starting point, Adept wants its system to perform steps in software on the person’s behalf. The idea depends on interpreting an instruction and mapping it to a sequence of actions in the relevant tool.
Adept describes its approach as a new kind of foundation model. In plain terms, the company is trying to build a general system that can work with software interfaces, rather than one designed only to generate content.
ACT-1 puts the idea over existing software
The company’s minimum viable product, ACT-1, can perform tasks such as importing LinkedIn URLs into recruiting software, according to Forbes. The prototype appears in an overlay window on top of programs such as Google Chrome or Salesforce.
That overlay approach suggests users could ask for help while working in familiar applications. Adept has a desktop prototype and plans to bring it to mobile in the near future, according to the source article.
Turning the concept into a useful product will depend on how well the system handles different tools and the steps people ask it to take. The source describes ACT-1’s capabilities and ambitions, but does not establish how broadly or reliably it performs beyond the examples given.
Investors are backing the opportunity
The round was co-led by General Catalyst and Spark Capital, with participation from Addition, Greylock, Atlassian Ventures, Microsoft, Nvidia, Workday Ventures, Caterina Fake, Frontiers Capital, PSP Growth, SV Angel and A.Capital. Forbes reported that Adept’s valuation was “at least” $1 billion.
The new funding brings Adept’s total raised to $415 million. CEO and co-founder David Luan said the money would support productization, model training and headcount growth.
Several investors also sell software that could potentially benefit from an AI assistant. Microsoft, Nvidia, Atlassian and Workday participated in the round, giving the company backing from firms with products that may intersect with its approach.
Adept enters a competitive field
Adept is not alone in exploring AI that learns to use computers. A February 2022 paper from scientists at Alphabet-backed DeepMind described an AI observing people’s keyboard and mouse actions as they completed instruction-following tasks, including booking a flight.
DeepMind co-founder Mustafa Suleyman also teamed up with LinkedIn co-founder Reid Hoffman to launch Inflection AI, which aims to use AI to help people work more efficiently with computers. These efforts point to a broader interest in making AI useful within everyday digital work.
Investors appear willing to fund the space despite competition. Adept had 25 employees at the time described in the source. The article also reported that co-founders Ashish Vaswani and Niki Parmar had left for another startup, while product development continued.
The central test for Adept is whether its model can turn a broad ambition into dependable help with real work. ACT-1 offers an early example of the direction: an assistant layered over existing software, responding to instructions by taking actions across tools.