Skip to content
Aurum AI LabsAurum AI Labs

Five steps. Your approval last.

The same sequence every run, and nothing acts until you say so.

See it run
From context to checked work
  1. Your context

    The brand, the brief and the target

  2. A defined workflow

    Specialist work, with your approval

  3. A checked result

    Ready to review and put to work

How it works

What a running workflow does

Five steps, in this order, every time.

  1. Detect

    It watches the data you already have, every day.

  2. Diagnose

    It works out what changed, and what caused it.

  3. Recommend

    It says what to do about it, and why.

    Your approval — on every send

  4. Act

    It does the work — once you have said yes.

  5. Verify

    It checks what the change actually did.

The pipeline

Who does the work

Five roles, each with one job — the team you would build if budget were no object.

  1. Detect
  2. Diagnose
  3. Recommend
  4. Act
  5. Verify

  1. Planner

    Scopes the job

  2. Researcher

    Digs for the evidence

  3. Executor

    Does the work

  4. Reviewer

    Challenges the result

  5. Formatter

    Puts it in the shape you asked for

The build pipeline runs inside every step of the ops loop.

Why it holds up

What the AI knows, and how it works

Two pillars. Five parts. Built into every workflow that runs.

Intelligence

  1. Skills

    Methodology, not prompts.

    1,000+ skills

  2. Capabilities

    Real-time data in, action out. We pull and push. We don't replace.

    100+ capabilities

Tradecraft

  1. People

    Agents that specialise, each with one job.

  2. Process

    Method fixed, output adapts.

  3. Practice

    A second opinion, built in.

    80+ models

    Cross-checked, not guessed.

The combined effect

All five work together for cost, quality and reliability.

  • Cost

    The right model for the task, every time.

  • Quality

    Different models see different things.

  • Reliability

    Cross-provider checks catch contradictions.

Your stack

What comes in, and how

Select a logo to explore that part of your stack.

Tools it reads from and writes to

Your tools · Ads

  • Google Ads
  • Meta
  • LinkedIn
  • TikTok

Your workflow

Read. Analyse. Check.

Ready for your review

You approve the next action.

Illustrative workflow · your existing tools stay in place

Data capabilities read live from the tools you already run — ads, analytics, social, search, your own first-party data.

Action capabilities write the result back where your team already works.

Your stack stays — we pull and push, we don't replace.

What it remembers

What it remembers, and how it gets smarter

Four things worth keeping, and one door in.

ProposedYour approvalRemembered

Everything enters as a proposal and lands on your approval. Nothing is remembered that you did not agree to.

The Brand Brain
Brand, products, audience, assets — the things that stay true.
A run history
What ran, when, and what it produced.
The signals worth keeping
What the data showed that is worth remembering.
Learnings
What one run taught the next. Learnings age, and stale ones are surfaced.

Nothing locks you in

What runs where

On our side
  • Assess it
  • Try one workflow
Either side — you choose
  • Keep a workflow running
  • Run all workflows in one place
In your cloud
  • Have workflows built around you
Your keysYour audit boundary

Assess it and Try it run on our side. Keep one running and Run all of it in one place run on our side or in your cloud — you choose. Have it built around you runs in your cloud, on your keys, inside your audit boundary.

All of it is published.

The architecture and the practices are open to read, and to check.

Read the Open Framework

Knowledge Hub

Explore the thinking behind the work.

Browse the Knowledge Hub

Marketing Workflows, Automated.

The method is built in. You close in on the target.

Stop prompting. Start marketing.

Automate your marketing workflows

Assess it — see where you stand, free.