We’re joined by Vijay Raina, an expert in enterprise SaaS and software architecture, to discuss a major shift in the help desk industry. With AI automation becoming a central theme, we’ll explore how startups are leveraging this technology to challenge established players by re-imagining workflows. The conversation will touch on the specifics of building reliable AI infrastructure on top of foundational models, the practical impact of this technology on reducing administrative overhead for IT teams, and a forward-looking vision where specialized AI tools operate within a larger, centralized AI ecosystem, fundamentally changing how we interact with enterprise software.
Congratulations on the recent $10 million seed round. How will this capital specifically accelerate your product roadmap, and what are the key milestones you aim to achieve with this runway? Please share some details on your immediate hiring or R&D priorities.
This $10 million is a game-changer for us; it’s the fuel we need to truly scale our vision. Our immediate priority is to double down on the infrastructure that makes our AI so effective. This means aggressively hiring engineers who specialize in AI systems and data pipelines. We’re not just building a product; we’re building an intricate system of prompt libraries and evaluation suites that requires immense R&D. The goal for the next 18 months is to expand our library of real-world examples tenfold, which is crucial for training the AI to handle more complex and nuanced support tickets. This funding allows us to move from theory to widespread, robust implementation with our enterprise clients.
Your platform is built on a third-party model, but the core is your infrastructure. What constitutes this ‘special sauce,’ and how do your prompt libraries and real-world examples keep the model’s non-deterministic nature in check?
That’s the critical question, isn’t it? The ‘special sauce’ is everything that sits between the raw power of a foundational model and the customer’s messy reality. A general LLM is like a brilliant but wildly unpredictable new hire. Our infrastructure is the rigorous training, the detailed playbooks, and the constant performance reviews that turn that raw talent into a reliable employee. The prompt libraries are meticulously crafted to guide the AI, and our evaluation suites act as a quality control system, constantly testing its outputs. For example, a simple password reset request isn’t just one prompt; it’s a chain of them that verifies user identity, checks account status, and logs the action, with checks at every stage to ensure the AI doesn’t go off-script. It’s this disciplined framework that tames the model’s unpredictability and delivers consistent results.
With a client like Gusto, your system automated 60% of support tickets. Could you walk us through the implementation process for a new enterprise customer and describe the specific types of tickets that are most effectively resolved autonomously?
Achieving that 60% automation with Gusto was a landmark for us, and it starts with a deep integration process. When we onboard a new enterprise client, we first connect our platform to their existing ticketing system, like Jira, and their internal tools. Then, the most important phase begins: learning. We analyze thousands of their historical tickets to identify high-volume, repetitive issues—think password resets, access requests for specific software, or basic payroll queries. These are the low-hanging fruit and the best candidates for automation. Our system is then trained on these specific examples, learning the exact steps a human agent would take. We track metrics like resolution time, first-contact resolution rate, and error rates to continuously refine the AI’s performance.
Given that some companies employ entire teams just to manage platforms like Jira, how does Risotto’s AI layer specifically reduce this administrative burden? Can you detail the step-by-step process for how your tool tames this complexity for a typical IT team?
It’s astonishing, isn’t it? We have a customer with four people whose full-time job is just wrangling Jira. Our AI layer essentially makes the ticketing system invisible for common tasks. Here’s how it works: An employee submits a ticket. Instead of an IT person seeing it, triaging it, and manually navigating a labyrinth of internal tools, our AI intercepts it. It understands the request, authenticates the user, executes the necessary action in the correct backend system, and closes the ticket with a resolution note, often in seconds. The human IT team is freed from that mundane, repetitive work. They only get involved when the issue is novel or requires human judgment, allowing them to focus on high-impact projects instead of just keeping the platform running.
You are preparing for a future where the primary interface for work is a central LLM. How does Risotto function in this paradigm, and what makes your focused tool more valuable than a general-purpose system like ChatGPT for Enterprise operating alone?
We absolutely see that future coming, and we’re building for it right now. In a world where a central LLM like ChatGPT for Enterprise is the main coordinator, our role becomes even more critical. Think of the central LLM as an orchestra conductor. It can direct the whole performance, but it doesn’t play every instrument. Our platform is the first-chair violin—a highly specialized, reliable, and context-aware instrument that the conductor calls upon for a specific, complex task. A general-purpose system might understand a request to “reset a password,” but it doesn’t have the deep, secure integrations and thousands of trained examples to actually do it reliably and safely within a company’s specific infrastructure. Our value is in that focused, dependable execution, which is something a general model will always struggle with on its own.
What is your forecast for the help desk automation industry?
I believe we’re on the cusp of a fundamental shift away from the human-in-the-loop ticketing system as we know it. The future isn’t just about AI helping humans close tickets faster; it’s about creating an autonomous layer that handles the vast majority of issues before a human ever sees them. The role of help desk software will change from being a human-centric interface to a hub for specialized AIs that manage, resolve, and learn. The companies that thrive will be those that master reliability and deep, contextual integration, becoming essential components called upon by a central intelligence rather than just another screen for an employee to look at. Human-friendly interfaces will become less important than machine-to-machine reliability.
