Workflow automation · Data · AI

We build software that does what your people already expected it to do.

Workflow automation, data pipelines, and AI applications for solo and small teams — fit to your real work, and shaped around the way your people already think about the job. Less interface for users to learn, better outcomes delivered.

What we believe

Three things we hold to.

We're a research-led studio. Our work tends to be quiet, infrastructural, and built to last — which means the relationships we like best are the long ones.

01 / Commitment

Long-term, on purpose.

We pick a small number of clients and stay with them. Most of what we build becomes infrastructure other things depend on — that's not a one-quarter relationship.

02 / Investigation

Their mental model, first.

Before scoping anything, we learn how your people already think about the work — what they expect to happen when they act, where they look for an answer. The build follows the mental model; the quote follows the build.

03 / Plain experience

Meets people where they are.

The best software disappears. Agents, natural-language flows, automations — designed so the path to the outcome matches what the person using it already imagined it would be.

Selected work

A few things we've shipped.

Client names stay private; the work is real. A longer list — including the messier procurement and research projects — lives on the work page.

Public data
Procurement signal from public noise.

Public procurement data is voluminous and almost unreadable in raw form. We built a pipeline that cleans, enriches, and turns it into monthly signals feeding a newsletter and a small dashboard product.

Data pipeline Newsletter Dashboard
Type
Data product
Status
In production
Research
Not surveys, customer conversations.

Static survey forms slowed product research and missed the interesting answers. We designed an AI-moderated feedback experience that adapts per session and drops into the product where it matters — no separate tool to learn.

AI experience Embedded Adaptive
Type
AI application
Status
Shipped
Hospitality
A POS that finally reports itself.

Daily sales, cost, and inventory reports were stuck behind a legacy POS and pulled by hand. We built a cloud automation that logs in, extracts the reports, and loads a warehouse every morning — quietly, without anyone touching it.

Browser automation Warehouse Daily run
Type
Workflow automation
Status
Running daily
All work
How we work with you

We learn the work first.

Software fails when it asks people to think the way the software thinks. We start by learning how the people in your team already think — then build outward from there.

01 / Listen

Sit with the work.

We learn the language, the workarounds, and the expertise people carry in their heads. What gets built sits on top of what's already there — not around it.

02 / Map

What people expect to happen.

Before scoping, we map the moments where someone reaches for an answer or makes a decision — and what they already think should happen next. That map becomes the spec.

03 / Iterate

Build with the people using it.

We ship in small pieces, in front of the people whose work it touches. Where what we built doesn't match what they expected, we adjust until it does.

04 / Stay

Long after delivery.

We maintain what we built, evolve it as the work changes, and keep listening. Most engagements outlast the original scope by years.

Get started

Have something that should just run itself?

Send a short description. You'll hear back within one business day.