The shapes of intelligent software.
Six patterns that describe what our work looks like, when you look at it closely.
Intelligent software tends to take a handful of recognisable forms. The patterns below are the ones we build most often. Each is a shape, not a client. A kind of problem, and the kind of software that solves it.
Patterns, not portfolios.
Every engagement is different in the details and similar in the shape. The same six patterns keep appearing in the work, across industries and team sizes. Recognising the shape is how a conversation becomes a plan.
The Quiet Automator.
The work that runs itself while your team sleeps.
An agent that handles the repetitive work between systems. Invoice reconciliation, support triage, data entry, routing. It reads, decides, and acts inside your existing tools, with human review where it matters.
A finance team that closes its books in a morning, not a week.
We wired an agent into Xero, Stripe, and a shared inbox. It reads statements, matches charges to invoices, posts the reconciliation, and only surfaces what does not match cleanly. A process that used to take three days a month is now a short morning review.
The Second Brain.
Everything your team knows, in one place, ready to answer.
A retrieval system that turns scattered documents, wikis, tickets, and conversations into a single source of answers. Ask it anything about your business, and it replies with the relevant passage and the source it came from.
One place to ask the company anything.
A SaaS team's knowledge lived across Notion, Google Drive, Intercom, and years of Slack threads. We indexed it nightly, wired a retrieval layer, and gave support, sales, and product a shared ask-anything interface. Every answer comes back with the source document and the date it was last touched.
The Ambient Advisor.
Intelligence that lives inside the tool you already use.
A reasoning layer embedded in your existing product. It understands the state of the work in front of you and offers the right help at the right moment. Not a chatbot in the corner. A quieter, smarter part of the experience.
An operations dashboard that explains its own next move.
Operators were reading tickets all day and guessing what to do next. We added a reasoning panel beside each case that reads the history, the policy, and the current context, then proposes an action with the why spelled out. Accept, edit, or override — one click. The product stays in their hands.
The Signal Finder.
Noise in. Clarity out.
A system that watches your data and surfaces what matters. The pattern that was not there yesterday. The outlier that deserves a look. The change that should reach a leader before it reaches a customer.
The five accounts most likely to churn, in Slack at 8am.
A B2B product team wanted early warning, not a weekly report. A nightly job reads usage, billing, and support signals, scores the accounts that look like they are drifting, and drops the shortlist into a Slack channel with the reasons attached. Customer success works the list over morning coffee, before the call gets awkward.
The Orchestra.
Many specialists, one performance.
A multi-agent system where each agent does one thing well, coordinated by a lead that decides who does what next. Research, analysis, verification, synthesis. The outcome is something no single model could produce alone.
A diligence memo written by four agents and one partner.
For a small investment team, a lead agent plans the research and routes subtasks to specialists — market sizing, competitor map, team background, financial sanity check. Claims come back with sources attached. A partner edits the memo in a morning, where the old version took the better part of a week.
The Conversational Surface.
The product people talk to, and the one that talks back.
A customer-facing product where the interface itself is intelligent. It understands intent, remembers context, and answers in the language of the user. Support, sales, onboarding, self-service. Conversation as the application.
A policy-aware chat replacing a labyrinth of help pages.
An insurer's help centre was six hundred pages nobody read. We replaced the front door with a conversation that knows the policy the customer holds, pulls the relevant clause, and hands off to a human the moment the question turns commercial. Most queries now close without a ticket.
Every shape starts as a conversation.
A discovery call surfaces the problem. A Compass engagement turns it into a proof of value. The Weave, the Build, and the Partnership are how the shape becomes permanent.
The path is always the same. Start small, prove fast, deepen as the value is earned.
One of these your shape?
If a pattern on this page looks like a problem you are holding, a discovery call is the next step.
Book a discovery call