AI that finishes the job

Agentic Research

01 Practice 01 of 04

Context-aware RAG and agentic systems with the tools to finish complex, multi-step work, not just answer questions. Our public showcase, StreetBriefs, researches, writes, fact-checks and publishes financial analysis autonomously, every day.

Start a conversation What we build
Abstract research system with translucent data panes, server blocks and connected nodes tracing teal light paths

Not chatbots. Working systems that plan, use tools, check their own work and deliver something you can act on.

Most AI projects stall at the demo. A model answers a question well in a meeting, then nobody can say what it should do on a Tuesday afternoon when the data is messy and the stakes are real. We build for the Tuesday afternoon.

What we build

Inside Agentic Research

The work that sits under Agentic Research, and what you can expect us to take on.

  1. Retrieval-augmented generation (RAG)
  2. Multi-agent pipelines and tool use
  3. Evaluation and fact-checking harnesses
  4. Autonomous reporting and agent operations

A working sketch

Try a small version

Not a screenshot. A small, working model of the kind of system Agentic Research builds, running on invented data. Change the inputs and watch what it does.

Research agent

Pick a question and watch how it works it out.

Scripted sample

Summarise this week's filings

  1. Plan Find the filings, compare with last week, summarise what changed
  2. Tool search_filings 6 filings found for the week
  3. Tool compare 3 material changes
  4. Check Every change linked to the filing it came from

Answered

Three material changes this week: a new debt facility, a revised guidance range and a change of director. Each line links to its filing.

A recording, not a live model. It plays the same way every time.

Public showcase

StreetBriefs runs this in the open

An agentic pipeline that scores around 210 stocks daily, then writes, fact-checks and publishes its own analysis. Nobody presses publish.

Visit StreetBriefs
Abstract research system with translucent data panes, server blocks and connected nodes tracing teal light paths

In motion

Agentic Research, in motion

A short, silent look at the kind of system this practice builds. It plays on its own and loops, and costs nothing until you scroll to it.

  • Retrieval-augmented generation (RAG)
  • Multi-agent pipelines and tool use
  • Evaluation and fact-checking harnesses

Proof, not promises

A case study

  • Abstract autonomous publishing pipeline with charts, blank publishing sheets, fact-check symbols and teal data paths

    Agentic Research

    StreetBriefs

    A fully autonomous financial publication: an agentic pipeline that scores ~210 stocks daily, writes and fact-checks its own articles, and publishes every day without human hands.

Four practices, one team

Keep exploring

Let's work together

Ready to start, or just want to talk it through?

Whether you're ready to start a project, want to explore an idea, or just want to talk data, we'd love to hear from you.

Start a conversation [email protected]
Response time
Within two business days
Where we are
Hamilton, New Zealand
What you get
Source code, documentation and training, handed over