AI Features Deep Dive — 7 Agents, Voice & Smart Automation

Nyura's AI layer is not a single model — it is a team of seven specialized agents, each trained for a specific domain. Amara handles task prioritization and deadline reasoning, analyzing urgency and impact scores before surfacing what matters today. Kai owns the calendar and meeting intelligence layer, connecting events to contacts and projects. Kenji powers travel — parsing booking emails, enriching flight data, and tracking live itineraries. Lina manages the contacts CRM, suggesting follow-ups and enriching profiles via multi-source lookups. Mila drives the music and focus modules, generating personalized playlists and ambient soundscapes. Ravi handles company intelligence, monitoring news and financial signals for your tracked organizations. Zara is the orchestrator — the conversational agent you talk to directly, routing your requests to the right specialist and synthesizing their answers. Each agent exposes a set of tools (search, create, update, summarize) that Zara can call in sequence to answer multi-step queries.

[SCREENSHOT: AI Assistant tab showing Zara chat interface with 'thinking' indicator and specialist routing label]

The voice layer in Nyura runs on ElevenLabs Conversational AI, giving Zara a natural, low-latency voice that responds in real time over a WebRTC audio channel. Open the AI Assistant tab and switch to Voice mode — you will see a pulsing orb that visualizes the audio signal. Speak naturally: 'Add a task to call Marcus tomorrow morning, high priority' and Zara will confirm the task was created, speaking back the details so you can verify without looking at the screen. Beyond task creation, voice supports fifteen tools across eight categories — searching your travel itinerary, querying contacts, checking your calendar, starting a meeting standup, and navigating between app sections. A live transcript appears below the orb so you always have a text record of the exchange. The microphone button mutes your input without ending the session. Voice capture also works in Quick Add — tap the mic icon next to the title field and dictate directly, with device-level speech-to-text transcribing in real time even when the full Conversational AI is offline.

[SCREENSHOT: Voice mode showing the animated orb, live transcript panel, and mute button]

Every task in Nyura receives an AI priority score computed by Amara. The score combines four signals: days until deadline, your historical completion patterns for similar tasks, the project's overall urgency weight, and any explicit priority you set. The result is a 0–100 score that adjusts throughout the day as deadlines approach or as you complete related tasks. You will see this score displayed as a small colored dot on each task card — red above 75, amber 40–75, green below 40. Amara also parses natural language when you type a task title. Write 'Call the accountant re Q1 taxes before Friday' and Amara will suggest extracting a due date (this Friday), linking the task to any contact named 'accountant' in your CRM, and tagging it with 'finance'. You can accept or dismiss each suggestion individually with a single tap. The AI breakdown feature goes further: for complex tasks with a vague title like 'Prepare investor deck', Amara can decompose it into three to seven concrete subtasks using Gemini, each with its own suggested due date and priority.

[SCREENSHOT: Task list with priority dots visible, and an AI suggestion banner proposing subtask breakdown]

The Meeting Brief feature is powered by Kai and the generate-meeting-brief edge function. When you open a calendar event in Nyura, a 'Generate Brief' button appears at the bottom of the event detail sheet. Tap it and Gemini compiles a structured dossier in under five seconds. The brief pulls context from three sources: open tasks assigned to or mentioning any attendee of the meeting, CRM profile data for each attendee (job title, company, last interaction date, notes), and any past meeting summaries linked to the same contacts. The output is a markdown-formatted card with four sections: Attendees (photo, role, last contact), Open Actions (tasks due or overdue involving them), Context (recent email or calendar history), and Suggested Agenda (three to five bullet points Gemini infers from your task and note data). The brief is not saved automatically — it lives in the event detail sheet and you can copy or share it before the meeting. Combine this with the AI Meeting Copilot (live transcription) to cover both pre-meeting prep and post-meeting task extraction.

[SCREENSHOT: Calendar event detail with AI brief expanded, showing attendee cards, open actions list, and suggested agenda]

The Decision Journal is a dedicated module inside Nyura for logging important choices you face — job offers, vendor contracts, product pivots, personal trade-offs. Each decision entry has a title, context field, a list of options, and a deadline. Once you save the entry, a 'Run AI Analysis' button triggers Gemini to generate a structured pros and cons breakdown for each option. The analysis goes beyond a simple list: Gemini assigns a confidence score (0–100) to each option based on the context you provided and flags which factors it considers most decisive. Confidence scores are displayed as colored badges directly on the decision card in the Decision Journal list view — so you can see at a glance which options the AI considers strongest. The analysis also surfaces questions you may not have considered, labeled as 'Blind Spots'. These are not prescriptive — Gemini will not tell you what to decide — but they help you notice gaps in your reasoning before committing. All analysis data is stored locally in your decision entry and can be exported as plain text or PDF.

[SCREENSHOT: Decision Journal entry showing AI analysis panel with confidence badges, pros/cons columns, and blind spots section]

If you are new to Nyura's AI features, the recommended order is: first, read [Getting Started](/blog/getting-started-complete-guide) to understand the core task and project model — the AI layer builds on top of these primitives, so knowing them well will make every AI feature more useful. Second, try Smart Task Suggestions with a handful of real tasks from your current week; observe how Amara adjusts scores as you mark things done. Third, generate a Meeting Brief for your next scheduled meeting before it happens — compare the suggested agenda against what actually gets discussed. Fourth, enable Voice mode and practice a few voice commands: create a task, ask for your schedule, query a contact. Once those four areas feel natural, explore Decision Analysis for a real trade-off you are currently weighing. Explore [Analytics](/blog/analytics-insights-guide) to see AI insights in action across your productivity data and discover patterns in how you work. The AI team learns from your usage patterns over time — the more consistently you use Nyura, the more personalized the suggestions become. There is no configuration required: the learning is automatic.

[SCREENSHOT: Blog article list showing Getting Started and Analytics articles linked as next steps]