Skip to content
Tell us what you needTalk to us

Results you can check. The rest waits for permission.

This page shows what we have built and run ourselves, and how we report results on a client engagement. Client results are published only with the client’s permission and the record behind them.

Tell us what you need Ask Sage on WhatsApp
The Lotus is our AI worker, made visible. What it runs is below, and most of it can be opened.

We don’t just talk about AI. We ship it.

Live products, an AI agent answering on WhatsApp right now, and the systems our own agency runs on every day.

11 products and systems, built in-house.

Sage, MapLift, Portrana and more, live and running today. All of it stands on 200+ branding & marketing projects since 2018.

Live

FrameRally

A client workspace for wedding and event photo studios: families pick favourites, albums get approved, the delivery is handed over.

Live

LeadFlow

A patient CRM for clinics: one WhatsApp inbox, follow-up and refill reminders, consent on record.

Live

AgentsWisdom

Ratings and reviews for AI agent tools, so a team can choose one on evidence.

Live

AdsAI

Mines Google Ads search terms to find the waste and tighten campaigns.

We run on what we build

  1. An enquiry arrivesThis website
  2. Sage answers and booksSage, on WhatsApp
  3. Agents research and draftOur agent team
  4. Content, tasks and images, preparedEveryday assistants
  5. A person approvesOur team
  6. Delivered and loggedOur CRM

Tools we build with

  • Claude (AI model)
  • OpenAI (AI model)
  • Gemini (AI model)
  • WhatsApp (Messaging)
  • Google Workspace (Workspace)
  • n8n (Automation)
  • Figma (Design)
  • Meta (Ads and social)
  • Claude Code (Coding agent)
  • Next.js (Web apps)
  • Vercel (Hosting)
  • Supabase (Database)
  • Perplexity (Research AI)
  • Apify (Web data)
  • ElevenLabs (Voice AI)
  • Codex (Coding agent)
  • Grok (AI model)
  • Hermes (AI agent)
  • Pi (Coding agent)
  • Antigravity (Coding agent)
  • Cursor (Coding agent)
  • DeepSeek (AI model)
  • Mistral (AI model)
  • Llama (AI model)
  • Nano Banana (Image AI)
  • Midjourney (Image AI)
  • Runway (Video AI)
  • Make (Automation)
  • Zapier (Automation)
  • Airtable (Database)
  • Bright Data (Web data)
  • Firecrawl (Web data)
  • React (Interfaces)
  • TypeScript (Language)
  • Python (Language)
  • GitHub (Code)
  • Convex (Database)
  • Capacitor (Mobile apps)
  • Docker (Servers)
  • Cloudflare (Network)
  • Hostinger (Hosting)
  • DigitalOcean (Cloud)
  • Canva (Design)
  • Adobe (Design)
  • Google Ads (Ads)
  • LinkedIn (Social)
  • Google Maps (Local search)
  • Telegram (Messaging)
  • Slack (Team chat)
  • Notion (Docs)
  • NocoDB (Database)
  • PostHog (Analytics)

Work we hand to AI

Handled by our agents every day

  • Answer WhatsApp enquiries after hours
  • Book appointments into Google Calendar
  • Log every conversation to a sheet
  • Alert the owner when something fails
  • Send WhatsApp campaigns on schedule
  • Answer questions from your own material
  • Hand complaints to a person
  • Draft follow-ups with the history attached
  • Stop chasing when someone replies
  • Keep one view of who is waiting
  • Prepare briefs before meetings
  • Cite a source for every claim
  • Draft content from your brief
  • Adapt drafts for each channel
  • Propose schedules within your rules
  • Collect documents for each job
  • Keep an audit trail of changes
  • Coordinate appointment requests after hours
  • Notify whoever is on duty
  • Answer order and delivery questions
  • Chase client documents
  • Summarise meetings into action lists
  • Track production status across accounts
  • Answer admissions questions from your prospectus
  • Coordinate property viewings
  • Request reviews after a stay
  • Triage requests against the service agreement
  • Write review templates for Google Maps

Every engagement reports against one question. We write it down before the work runs.

The measure is agreed in scoping, so a result is read against what you asked for, not against whatever happened to move.

  1. Agree the question

    Before anything runs, we write down what the work is meant to change, what we would measure, how soon it could be read, and what would make us stop.

  2. Note the starting point

    We record where things stand before we begin, in your own numbers where you have them, so the change can be read against something.

  3. Report in your language

    Each review reports what moved against that question, in your language rather than the platform’s.

  4. Keep or stop

    What tested well is kept and the next round starts from it. What did not is retired, including the expensive ones.

No promised numbers.We will not promise a result before we have seen your business.

Client results are published with permission. Until then, they stay with the client.

A results page is the easiest page on a website to fake. We would rather publish less and have all of it hold.

  • A client is named only with their permission.

    Who we work with is the client’s information before it is ours. We ask before we name anyone on this site, and we publish a case study only when the client has agreed to it.

  • A client result ships only with its record.

    A result from client work is published with the measured record behind it and its scope attached. If we cannot show where it came from, it is not here.

  • Examples are illustrations.

    Examples on this site are illustrations, not client results. The businesses, dates and messages in them are invented to show how the pieces fit together.

Bring the job you want done. We will tell you how we would measure it.

Ask about scope, measurement or what we would refuse, or start a brief.

Brand Wisdom A person reads every enquiry and replies within one working day.

Tell us what’s going on and we’ll help you work out the next step. A person reads every one of these, and you’ll hear back within one working day.

What should we work on?

Or ask first

We store your answers so we can reply. Read what we keep and why in our privacy notice.