The startup, which is coming out of stealth, is building an AI assistant around the idea of ‘proactive AI’. This identifies commitments users make in emails and starts the work needed to fulfil them, instead of waiting for users to write a prompt or assign a task.
Sol currently has more than 100 users on its pre-launch platform and is initially offering selective access through a waitlist. The company plans to open the product more broadly in the coming weeks.
“People whose work runs on email often have to repeat themselves: first on mail, then again when they ask an AI tool to act. There’s no good reason for that,” Karan told ET in a conversation.
“The workflow today is surprisingly redundant. If I have expressed, with reasonable intent, that I will do something, there should be an AI product that can spot that commitment, do the work, and bring it back to me for approval,” he added.
The startup said the product scans Gmail for commitments, gathers the relevant context, and can carry out tasks including research, creating documents and presentations, and coordinating meetings. The user gives the final approval before anything is sent, scheduled, or shared.
The company is initially targeting people whose work involves a high volume of commitments and where the cost of missing one is relatively high. This includes founders, executives, and company leadership, as well as people in sales, recruiting, consulting, and agencies.
Karan said the startup’s primary competition is not necessarily other AI assistants, but the workflows users have already built around existing AI tools.
“Users can configure AI tools to create routines, reminders, and workflows, but maintaining these workflows requires them to continuously configure and update them,” he said.
The company said it has built Sol to operate with limited user intervention. Once connected to Google Workspace, including Gmail, it builds a contextual understanding of the user — including their role, the people they work with, the companies they deal with, and how they communicate with different people — before it begins to act.
Karan said the challenge in building such a product is not simply connecting an AI model to email, but making the system reliable across multiple contexts and types of work.
Under the hood, Sol has a library of more than 100 specialist skills, according to the company. These skills can improve based on how users interact with the work produced by the system.
The startup is model-agnostic and uses different AI models depending on what works best for a particular task, Karan said.
Data security
Karan emphasised that Sol cannot send an email, schedule a meeting, or share an output without the user’s explicit approval. Permissions can also be revoked at any point.
The company said it has completed CASA, a cloud security assessment, and is SOC 2 Type 1 certified — this is an independent audit framework used to assess a company’s controls in areas such as data security and privacy. It plans to obtain SOC 2 Type 2 certification, which assesses whether the controls continue to operate effectively over a period of time, in the coming months.
The company expects most of its near-term hiring to be in the US as it looks to stay close to customers there.
The fresh capital will primarily be used for R&D, including AI processing and computing costs, hiring, and working with customers across geographies. Some of it will also be used for go-to-market and scaling the product, Karan said.