DESIGN SPRINTS

Software designed around what agents can do.

A browsable set of design prototypes from our sprints. Each one takes a real, recognizable problem and explores what AI-native software — software that acts, not a chatbot bolted on — could look like. Every one is an interactive prototype you can open.

THE IDEA

Not chatbots bolted onto old software.

The easy way to add AI to software is to staple a chat box onto a tool that already exists. The more interesting question is what the software would look like if it were designed around what agents can now do — take actions, hold a plan, and carry out work, not just answer questions. That is what these prototypes explore.

Each one starts from a real, recognizable problem and imagines the AI-native version: the workflow and the interface rethought around agents-in-the-loop, from the ground up. They share one design language, which is how a handful of sprints can produce work that all feels like it came from one hand.

SELECTED WORK

13 prototypes, one thesis.

Open any prototype to try it, or read the case study for the problem it explores and the design thinking behind it.

Hiro — Agentic ops triage

Hiro

Agentic ops triage

An operations triage console that reads the inbox, Slack, Stripe, Snowflake and Linear overnight, auto-clears what's routine, and escalates only what needs a human — taking safe autonomous actions and logging the rest.

Business-process & workflow automation
Drift — Declarative planning for software teams

Drift

Declarative planning for software teams

Treats the spec as the desired state and code as the actual state, so the work is the measured gap. Coding agents pull the codebase toward the spec, and scheduled checks surface drift as reviewable spec PRs.

Software that builds software
Saga — Agent observability

Saga

Agent observability

A control room for a workforce of production agents — web chat, voice, dev copilot, fraud triage — tracking pass rate, spend, latency, compliance regressions and behavior drift against a live test suite.

Monitoring, observability & QA of AI systems
Fidé — Compliance console

Fidé

Compliance console

Continuous compliance monitoring across frameworks, checklists and audits — one live row per target (repositories, entities, new hires) with failing, attention and due states kept current automatically.

Continuous monitoring & compliance automation
Data Rooms — Records, organized

Data Rooms

Records, organized

Turns a pile of personal records into a structured, askable room — reading lab PDFs into data, building a timeline, flagging what needs attention, and drafting the appeal or the follow-up for you.

Document processing, data collection & analysis
Author — Drafting console

Author

Drafting console

Send a paragraph of raw notes and get a clean, finished draft back — refined by talking to it, clicking a line to edit, or selecting a phrase to change, always in a chosen voice.

AI-native document generation
Flow — Process operations

Flow

Process operations

Maps how work actually gets done across roles as living swimlanes — each step manual, assisted or automated — with sub-processes, bindings, and the exceptions and escalations spelled out.

Workflow & process automation
Autonomous Humans — Event program management

Autonomous Humans

Event program management

Runs an entire events program end to end — series and cohorts from planning to live, registration progress, speaker and organizer approvals — synced to the tools the team already uses.

Business-process automation
North — Strategy operations

North

Strategy operations

Makes a company's strategy operational — mission, objectives, metrics and a live strategy map with on-track and at-risk signals, so the plan and the numbers that prove it stay in one place.

Business strategy & operations
Cozy — Relationship manager

Cozy

Relationship manager

A personal relationship manager built on the AI Hero design system — keeping the people who matter, the context around them, and the next nudge close at hand.

Consumer agentic web service
Lemur — Learning management system

Lemur

Learning management system

A learning management system where every lesson ships with a tutor that answers only from the lesson itself — teaching a concept, then checking that the learner can apply it.

AI-native learning software
Auth — Identity & access

Auth

Identity & access

A prototype exploring identity for the agentic era — one login for every app a person is entitled to, no second password, no second prompt, with entitlement computed live from team membership rather than stored as stale grants.

Authentication, authorization & identity
Primer — Agent skill library

Primer

Agent skill library

A library for authoring and observing a team's agent skills — measuring how often each skill actually fires, so a description that doesn't match how people ask shows up as a fixable zero.

AI agent tooling & observability

THE SYSTEM BEHIND THE SOFTWARE

What ties the prototypes together.

None of this coheres without a shared visual language. One monochrome ink ramp, one accent, one composition rule — the design system is why every prototype above looks like the same hand made it.

Have a workflow that doesn't fit off-the-shelf software?

Send us a paragraph. We'll send back a one-page design within a week.

Talk to us →