Pablo PalafoxCo-founder & CEO

We help enterprises deploy AI agents in complex environments like telecommunications, utilities, financial services, and supply chains. Our agents handle the calls, emails, and messy coordination tasks that keep businesses running. We focus on executing work and gathering real-world insights rather than just answering basic questions.

We went through Y Combinator in the summer of 2023 and realized we needed to change our direction. We pivoted and started the current business in early 2024. Our first sale occurred in April of 2024, followed by rapid funding rounds that turned the company into a unicorn.

Intelligence itself is not the limiting factor. The key is how you wield that intelligence. Large language models are widely available, but enterprises struggle with coordination issues and transferring information between departments, such as customer service, finance, and external trucking companies. That coordination is what we automate.

We started in freight logistics because the industry is highly dependent on coordination. We realized that these problems are not specific to supply chains but represent a general enterprise coordination challenge. Serving early customers like DHL helped us connect with telecommunications firms to deploy our solutions horizontally.

We built one of the first voice agents in January of 2024 because freight and supply chain processes are traditionally run over calls and emails. Voice agents were the key unlock to work with nine of the top ten US freight forwarders, though we now support email, WhatsApp, and chat.

We work with a large parcel company that handles up to forty thousand outbound calls and follow-up texts daily. Our agents call customers who have not paid import duties on packages. This helps clear warehouse space, prevents return shipments to origin, and reduces repetitive work for human customer service agents.

Instead of shallow support, our agents handle coordination tasks like dispatching technicians to fix routers or boilers. When a customer calls with an issue, the agent coordinates the technician schedule and manages the escalation to human teams, resolving the issue without manual coordination.

A software platform alone is not enough; clients need experts to customize the implementation. We assign full-time forward deployed engineers to sit onsite with customers for weeks. They understand both the technology and the business, which allows us to deploy complex integrations in less than four weeks.

Even in a two hundred person company, you need a catalyst to make things happen. In a massive enterprise with thousands of employees, a forward deployed engineer acts as a bridge between the operators in the warehouse or energy plant and the executive team, helping them make fast, data-driven decisions.

Our model relies on three pillars: the platform itself, credits for consumption, and our deployment services. The platform is the vehicle, the consumption credits represent the fuel, and our engineers teach the client's team how to drive. Eventually, we transfer full control to the customer's data scientists.

The execution layer runs the agentic workflows, such as monitoring email inboxes or fielding inbound calls. The context layer, which we call Twin, integrates directly with CRM and data systems. Finally, the interface layer provides code-based applications and user interfaces for human teams to monitor and guide the agents.

We wanted to put exceptional care into voice quality. We have a research team of seven machine learning engineers who combine transformer models and open-source systems to build custom text-to-speech models. This proprietary technology ensures that our voice agents sound natural and professional.

Small language models allow us to distill the knowledge of massive LLMs into specific, day-to-day tasks. You do not need a PhD-level language model to call a driver and ask for their location coordinates. Small language models reduce token usage costs, lower compute requirements, and increase customer ownership.

While we offer a multi-tenant cloud solution, we also deploy single-tenant, on-premise installations for customers with strict data sovereignty requirements. Ensuring that large enterprises have complete ownership and security over their communication data is a key priority for our platform.

Intelligence will not automatically deploy itself in a business. Large language model capabilities will fluctuate, so enterprises should not tie themselves to a single model provider. Instead, they need an orchestration platform like ours that is interoperable and can utilize multiple models to solve specific operational workflows.

Intelligence compounds in the enterprise. Building your first agent takes time, but building subsequent agents is much faster because you have already completed the data integrations and captured the company's context. Earning the right to expand across different departments creates deep stickiness for our platform.

We are focusing our resources on product development and deployment teams. We are growing our research team to improve our small language models and text-to-speech capabilities, while expanding our team of forward deployed engineers to unlock value and deploy new agents for our enterprise customers.