Tel Aviv, Israel-based Doti AI has raised $7 million in seed funding led by F2 Venture Capital and angel investors including Jared Kasner and Guy Flechter.
Doti is described as a ‘Work AI platform’; that is, a platform that uses artificial intelligence to improve, automate, and streamline standard office and business processes across distributed and hybrid environments.
The firm claims its platform can be deployed in less than an hour. It integrates with existing tools and consolidates data – both structured and unstructured – from multiple existing business platforms, providing a single interface to company data. Company data is kept within the company and under the company’s control – the system provides information from the data, not access to the source.
Enterprises, says Matan Cohen, co-founder and CEO of Doti AI, “need a secure, intuitive platform that not only understands their data but delivers the right context when and where it’s needed.” Since Doti is an AI product, it can handle tasks like analyzing codebases, drafting communications, retrieving historical data and answering policy questions.
An example of Doti in use could be the real time retrieval of client history and product updates to facilitate a personalized pitch during a phone conversation.
However, effective data retrieval is historically dependent on asking the right question – the more complex the answer required, the more difficult it is to phrase the right question. Doti assists this problem through the power of AI – it integrates with platforms like Slack to monitor ongoing conversations, can recognize a potential question and automatically deliver an answer before the question is formally asked.
Security is provided through both direct and indirect prompt injection guardrails that prevent both insiders and malicious outsiders manipulating the system to gain unauthorized information. If the network is compromised by an intruder, the intruder will only gain limited access to a limited part of Doti’s knowledge. This is through Doti’s use of authenticated access and its concept of ‘spaces’.
Firstly, explains Opher Hofshi, CPO, “Every interaction with Doti must be authenticated, either via OAUTH/SSO or using a generated API token in the system.” Secondly, he continues, “We always and only require READ ONLY permissions in systems we connect to. Doti cannot be used to manipulate the systems themselves and so cannot be used to deploy some sort of ransomware etc.”
The ’spaces’ part of the solution is Doti’s ability to segment user access to limited areas of the company information. In this way, an intruder with stolen credentials would only gain access to limited data (with no ability to change it), while a legitimate company user would be unable to access any data he or she shouldn’t. Malicious activity, assuming an intruder can by-pass the company’s other security controls and get into the network, is limited, while the more common problem of internal oversharing by legitimate users is prevented.
It’s a version of role-based access control (to information in real time) updated to the AI age. Its purpose is to make the right company information (data made meaningful through AI) available to the right company people, on demand and in real time. And to eliminate the problem of employees being forced to say, “I’m not sure of the answer – give me a few hours; I’ll look into it and get back to you.”
Doti AI was founded in Tel Aviv in January 2024 by Matan Cohen (formerly Software Group Manager at WIX and previously senior software developer at the IDF’s J6 & Cyber Defense Directorate), and Opher Hofshi (formerly security architects team leader at WIX and previously a security consultant with Rockwell Automation).
Related: Cisco Unveils New AI Application Security Solution
Related: AI Won’t Take This Job: Microsoft Says Human Ingenuity Crucial to Red-Teaming
Related: Emerging FunkSec Ransomware Developed Using AI
Related: Trolley Problem, Safety Versus Security of Generative AI
