A Voice-First Internal Knowledge Agent for Everyday Workplace Tasks
Written by
Subject
Published on
MegChem employees need answers from policies, safety records, HR systems, and booking tools. Finding them means searching across documents and apps, often while a task is already waiting. App Developer Studio built Meg AI as a voice-first internal agent so staff could ask a question or start a task the way they would with a colleague.
Critical information lived in separate systems: safety and incident reports, HR records, meeting-room and vehicle booking, email. Manual search was slow. Typed chatbots still felt like another tool to learn. MegChem needed one conversational entry point (grounded in company knowledge) for the quick office tasks that keep a busy day moving.
Meg AI combines live voice conversation with retrieval-augmented generation over a secure vector store. Employees speak naturally; the agent retrieves relevant internal documents before responding and supports practical tasks in the same session: incident summaries, leave balances, room and vehicle booking, email drafts. The result is an internal AI experience that is easier to reach and harder to misuse than a generic chat window.
The demo below walks through a typical Meg AI session: starting a voice conversation, asking about policies and operational records, checking leave balances, booking a meeting room, and drafting a professional email from a spoken brief. Watch how the assistant stays conversational while pulling answers from company knowledge and acting on workplace systems.
What the demo shows in practice: safety teams can surface lessons from incident reports without opening file shares; HR queries like leave balances resolve in conversation; and coordination tasks (rooms, vehicles, outbound email) stay in one place instead of three separate tools.
Employees speak to Meg AI as they would to a colleague. No form-filling, no menu navigation: just a natural conversation to get an answer or start a task.
Responses draw on internal documents through retrieval-augmented generation and a secure vector store. Answers stay tied to company knowledge, not generic model output.
Meg AI extracts and summarises incident and safety reports (contributing factors, root causes, corrective actions, and related RCA or ICAM guidance) when teams need them.
Employees retrieve leave balances and related information, with Meg AI guiding them toward the appropriate next step rather than leaving them at a dead end.
A spoken request or a brief set of notes becomes a clear, structured professional email that the employee can review and send.
Meeting-room availability, bookings, and company vehicle lookup: common coordination tasks handled inside the same conversational session.
Meg AI is built on retrieval-augmented generation with a secure vector store indexed from MegChem’s internal documents: policies, operational records, safety documentation, and other business resources. Voice interaction sits on top of that retrieval layer, so every response can be grounded before it is spoken. Task support (bookings, balances, email drafts) runs through structured integrations behind the conversation, keeping natural speech on the surface and precise data underneath.
RAG has to retrieve the right policy or report, not invent one. We scoped retrieval to indexed company documents and kept every answer traceable to source material in the vector store.
Workplace tasks need accurate data: dates, balances, bookings. We paired natural speech with structured tool calls behind the conversation so Meg AI could act on systems, not just talk about them.
Meg AI gives MegChem employees a more accessible internal AI experience: voice interaction, trusted document retrieval, and task support in one place. Staff spend less time navigating separate tools and searching for information, and more time on the work the answers were meant to enable.
Meg AI can extend to additional document sources and workplace systems as MegChem’s needs grow: more operational records, deeper HR workflows, and broader safety and compliance coverage, all through the same voice-first entry point.
MegChem shows what a bespoke internal AI agent looks like when voice, RAG, and task support are built together from the start, not bolted on as afterthoughts. Meg AI helps employees work more efficiently, make better-informed decisions, and spend less time on routine administration. Explore our AI Solutions for Clients service to see how we apply the same approach for other organisations.
From voice-first RAG assistants to full enterprise integrations, we build bespoke AI products for clients, the same engineering we applied to Meg AI for MegChem.
Our own engineering intelligence platform, unifying Cursor, Jira, GitHub, Figma, Slack, and Google Workspace into one dashboard.
Our own multi-agent AI platform, built to fix our own sales bottleneck. Cut our sales lead-to-conversion time from two months to three weeks.
AI-powered car damage detection, built with Tcheck on a model trained on 100+ cars.
A compassionate AI chatbot combatting gender-based violence. GRIT / Zuzi AI Award Winner, 2024.
