Do you want smarter insights in your inbox? Sign up for our weekly newsletters to get what is vital for AI leaders, data and security. Subscribe now
DustThe two-year platform of artificial intelligence, which helps enterprises build AI agents capable of making entire flows of work, reached $ 6 million annual revenues-a thinning boost from $ 1 million only a year ago. The rapid growth of the company signals the change in the admission of entrepreneurship from basic chatbots to sophisticated systems that can take specific actions in various business applications.
The startup from San Francisco announced on Thursday that he was chosen as part of the ecosystem “powered by Claude” Anthropica, emphasizing the novel category of AI companies building specialized company tools on the boundary language models instead of developing their own AI systems from scratch.
“Users want more than just conversation interfaces,” said Gabriel Hubert, CEO and co-founder of Dust, in an interview with Venturebeat. “Instead of generating a project, they want to automatically create a real document. Instead of getting a summary of the meeting, they need to update CRM records without manual intervention.”
The Dusta platform goes far beyond AI in the style of chatbot, which dominated the early reception of the company. Instead of simply answering questions, AI Dust agents can automatically create github problems, plan calendar meetings, update customer records, and even send codes reviews based on internal coding standards-all, while maintaining corporate class safety protocols.
How AI agents turn sales connections into automated GitHUB tickets and CRM updates
The company’s approach becomes clear through the described specific example of Hubert: a company sales company dealing with business with many dust agents for processing transcripts of sales connections. One agent analyzes which sales arguments resonated with potential customers and automatically update the combat cards in Salesforce. At the same time, another agent identifies the demands of the customer’s function, maps them to the road map of the product, and in some cases automatically generates GitHub tickets to diminutive functions considered ready for development.
“Each connection transcript will be analyzed by many agents,” explained Hubert. “You will have an agent of the Sales Battle Optimalizer, which will look at the arguments created by the seller, which were powerful and seem to resonate with the perspective, and which will go into the process on the side of Salesforce.”
This level of automation is turned on by Context Protocol (MCP)The novel standard developed by Anthropic, which allows AI systems to be safely connected with external data sources and applications. Guillaume Prinen, head of EMEA in Anthropic, described MCP as “like a USB-C connector between AI models and applications”, enabling agents to access the company’s data while maintaining security limits.
Why Claude and MCP supply the next wave of automation AI Enterprise AI
Dust’s success reflects wider changes in how companies are approaching AI implementation. Instead of building non -standard models, companies such as Dust operate more and more talented models of foundations – especially Anthropic’s Claude 4 Suite – and combine them with specialized orchestration software.
“We just want to provide our clients with access to the best models,” said Hubert. “And I think that at the moment Anthropic is at the beginning of the lead, especially in coding models.” The company charges customers 40-50 USD per month and serves thousands of working spaces, from diminutive startups to huge enterprises with thousands of employees.
Claude Anthropic models have recorded a particularly powerful acceptance of coding tasks, and the company has reported a 300% boost in Claude code consumption over the past four weeks after the release of the latest Claude 4 models. “Opus 4 is the most powerful coding model in the world,” noted Prinen. “We have already run a coding race. We strengthen it.”
Enterprises’ security becomes elaborate when AI agents can actually take action
The transition towards AI agents who can take real action in various business systems introduces novel security complexities that did not exist in the case of basic Chatbot implementation. Dust concerns this through what Hubert calls the “native layer of permissions”, which separates the rights of access to data from the agent’s rights.
“The creation of permissions, as well as managing data and tools are part of the implementation in order to limit confidential data exposure when AI agents operate in many business systems,” explains the company in technical documentation. This becomes crucial when agents have the ability to create GitHub problems, update CRM records or modify documents at the pile of organization technology.
The company implements the infrastructure of a corporate class with the principles of retaining zero anthropics, ensuring that confidential business information processed by AI agents is not stored by the model supplier. This applies to the key concern for enterprises considering the adoption of AI on a huge scale.
Growth of native AI startups on foundation models instead of creating your own
The growth of dust is part of what anthropic calls the ecosystem of “native AI startups” – companies that could not exist without advanced AI. These companies are building companies not by developing their own AI models, but by creating sophisticated applications on existing foundation models.
“These companies have a very, very strong idea about what their end customers need and want this particular case of use,” explained Prinen. “We provide them with tools to build and adapt their product to these specific customers and cases they are looking for.”
This approach is a significant change in the structure of the AI industry. Instead of every company that needs to develop your own artificial intelligence abilities, specialized platforms such as Dust, can provide a layer of orchestration, which makes powerful AI models useful for specific business applications.
What Dust in the amount of USD 6 million in boost in revenues about the future of the company’s software
The success of companies such as Dust suggests that the AI Enterprise market goes beyond the experimental phase in the direction of practical implementation. Instead of replacing wholesale employees, these systems are aimed at eliminating routine tasks and changing the context between applications, enabling employees to focus on higher value activities.
“By providing universal AI primitives, which make all the company’s flows more intelligent, as well as the appropriate permit system, we set the foundations of the agent’s operating system that is future,” said Hubert.
The company’s customer database includes organizations convinced that artificial integency will fundamentally change business activities. “The common thread between all clients is that they are quite created in the future and convinced that this technology will change many things,” noted Hubert.
Because AI models become more talented and protocols, such as mature MCP, a distinction between AI tools that simply provide information and those that take action can become a key distinguishing feature on the enterprise market. A rapid boost in revenues suggests that companies are willing to pay premium prices for AI systems that can complete the actual work, and not just aid with this.
Implications go beyond individual companies for a wider structure of corporate software. If AI agents can easily integrate and automatize work flows in disconnected business applications, it can transform how organizations think about software orders and work flow design – limiting the complexity that has long been harassed by the piles of company technology.
Perhaps the most articulate sign of this transformation is how naturally Hubert describes AI agents not as tools, but as digital employees who appear to work every day. In the business world, which spent decades, combining systems with API interfaces and integration platforms, companies such as Dust prove that the future may not require the combination of everything – simply teaching artificial intelligence to move around the chaos that we have already built.