Skip to content
Platform Brief
brand → customers, suppliers, candidates and others
Nurix.ai is building a platform called AgentX that allows companies to build a conversation AI agents. These agents speak with the brand’s customers (to handle queries, clarify product doubts, and perform after sales support in general), suppliers (to negotiate terms, do followups etc), prospects (either job candidates or sales prospects to do lead qualification) and others.
Conversational AI would need to include multiple channels to be able to automate the current role of humans and capture market effectively. These communication channels include email, chat (whatsapp, instagram, telegram etc) and voice (telephonic) with the ability to maintain context across channels or switch between them seamlessly for things like uploading documents while on a phone call.
This takes a load off of human agents performing customer support (56% of B2B AI use cases include customer support and is ranked #1) and sales operations (AI cannot effectively do later stages in the sales funnel but early stages such as lead qualification can be done effectively in large volumes).
Creating agents is done on Nurix’s online platform. This is done by employees of Nurix itself currently in the foreseeable future of 3 to 6 months but the platform capabilities will be intuitive enough for companies to build their own agents in the future.
An agent on Nurix is conversational, either by text or voice (to be selected even before creating an agent because the backend structure is different for both). Agents can perform tasks, via integrations (bundle of 200-300 integrations currently being provided by a startup called unified.to). Agents can have custom behavior, tasks, goals etc in the form of a prompt. Agents have access to a live knowledgebase as well via integrations of google drive and other online sources.
A follow-up prompt can be given to define the behaviour of upcoming calls. Call summaries can be generated (optional with a custom prompt), calls can be categorised immediately after they end (optional with a custom prompt again). Speech settings can be customised (including text to speech variables, speech to text variables and voice activity variables). Language model can be configured. SIP trunk can be configured.
Agent creation is a slightly involved process that is expected to be done only once and then maintained over time to keep it optimised for the usecase and the brand. Nurix offers a service to have it done for the company by employees of Nurix (creation + maintenance). For a SaaS company, the spread between software and service revenues is usually 80-20 percent. But in our case we expect it more to be something like 50-50 as Nurix is based out of India and labour cost is cheaper here + Nurix employees will be better at doing it than in-house people that a company might hire.
The tasks that this agent performs are not exactly deterministic because it’s based on prompts. To make the agent behaviour more readable, predictable and scalable, Nurix offers a workflow studio. Workflow studio includes a canvas where blocks can be placed and tied together to create a flowchart of how it should behave. These blocks can include agents that were created before, but also tools (actions from integrations), API endpoint integrations, conditional logic (that can be gotten from javascript variables or have natural language descriptions), and more.
This workflow studio allows process builders at companies to create workflows that adhere to regulations (these people can include customer success managers for example), while people that are more well versed with technology can build specific agents that can be used by these workflows. This separation allows these different stakeholders to do what they are best at.
Voice agents have been 2 types from what was observed in the market so far. Conversational and Logical. Conversational is when you give a prompt regarding its personality and behaviour and goals while logical is more similar to IVR systems that are more strict in nature and can benefit from having a visual flowchart.
Nurix GTM will be focusing on the MM+ (mid market plus, roughly $50M+ ARR) markets of retail and insurance in the US. India can also serve as a secondary market due to the founders’ extensive connections in India that can garner clients quickly and serve a reputation. Catering to customer support usecases serves the retail market, while catering to sales usecases serves the insurance market. Things like issue resolution, product inquiries, after sales support can come under retail, while lead qualification, FNOL/claim registration, claim procedures can come under insurance.
Bare agent capabilities will be the highest priority to execute right now, while things like feeding context, having monitoring abilities will come next. The GTM is selectively narrow as we have extensive connections in retail and insurance is something we have seen evidence of traction in. These verticals (support, sales, hr, etc) and horizontals (retail, insurance, finance, automobile, supply chain etc) will have to be experimented with to find PMF.
🧭 Dashboard Design Implications: What the UI Must Support
To design a dashboard that supports this architecture well, your Information Architecture (IA) needs to consider these structural implications:
1. Modular Agent Overview Page
2. Workflow Designer Interface
3. Scenario Simulator
4. Live Performance Monitoring Panel
5. Conversation Explorer
6. Prompt + Guardrail Library
7. Human Handoff Monitoring
🏁 TL;DR
Z-Studio is a full-stack platform for designing, deploying, and managing conversational AI agents across channels and tasks. It has an agent builder, a process orchestrator, testing sandbox, and performance reporting — and each part has implications for how the dashboard should surface configuration, usage, and results.
The dashboard must support modular editing, scenario-based testing, real-time insight, and process-driven views to truly empower product and ops teams.
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.