How we work

No big separate functions. No layers of communication. No complex approval chains. A single project team goes from idea to deploy independently. Read our Operational Principles and Work Principles for the philosophy behind this structure.

No sprints

We don't run sprints, and there's no Scrum Team or sprint planning ceremony. Work runs on daily commits, rolled up into a weekly CROSP report - see Daily commitment and Weekly commits below for how that works.

Show & tell

End of each week. We celebrate the week's work. People talk about what shipped, problems they worked on, solutions they found. The team votes on a winner who takes home the Golden Mic for the week.

Development flow

Every feature moves through three roles and a fixed set of steps, documented in the brain repo: Product Owner (the what and why), Engineering Owner (the how and the test code), and Test Owner (manual end-to-end validation on staging). See our Engineering Principles for the technical standards behind each step.

1. 1Requirements & scenarios (Product Owner) - write requirements and UAT scenarios in the brain repo, in plain behavioral language. No endpoints, no selectors, no code.
2. 2Mockup gap analysis (Product Owner) - map every requirement to UI coverage and flag what's missing before engineering starts.
3. 3Service design & API contract (Engineering Owner) - domain model, architecture decisions, exact endpoints and schemas.
4. 4Code & unit tests (Engineering Owner) - written with Claude Code and Cursor, AI-assisted review, nothing merges without it. TDD, 80% minimum coverage.
5. 5Automated + manual E2E (Engineering Owner + Test Owner) - automated frontend and backend tests run on CI, manual E2E runs on staging against real infrastructure.
6. 6Sign-off (Product Owner) - review pass/fail results, sign off or raise defects for Engineering to fix.
7. 7Customer docs (Product Owner scopes, Engineering Owner declares the public surface) - written after sign-off, against the shipped and tested feature.

🤖 AI is how we work

Every team member at Intempt works with AI agents daily. Not as a side experiment - as the primary way work gets done. We use Claude Code with harness and loop engineering for backend services, Lovable for frontend prototyping and full builds, Claude on claude.ai for strategy, writing, analysis, and coworking sessions. Every person has AI agents running alongside them throughout the day.

• Engineers run Claude Code with harness and loop engineering - multi-agent workflows, automated code review, test generation
• Product coworks with Claude on specs and prototypes in Lovable. Marketing uses AI for content, campaigns, and creative
• Everyone uses claude.ai as a thinking partner. AI agents handle support, code review, docs, and deployment checks

🕐 Working hours and commitment

Core hours

Full-time employment means a full working day. Core hours are 9:00 AM to 6:00 PM, structured as an 8-hour workday plus a 1-hour lunch break. During this window, you are expected to be online, responsive, and actively working - using Claude and the other AI tools the company provides.

Daily commitment

At the start of each day, state your commitment: the specific tasks you intend to deliver. Your work is reviewed against that commitment.

• If you undercommitted (finished early with capacity to spare): this is a signal to calibrate - take on more next time. It is not a basis for ending your day early.
• If you overcommitted (couldn't finish): say so, and explain what happened. Adjust and carry the work forward. Honesty about the miss matters more than hitting an artificial number.

Weekly commits: the CROSP report

Daily commitments roll up into a weekly CROSP report, reviewed at the start of each week: what you delivered against what you committed to, and why - if there's a gap in either direction.

Level and growth

How well you calibrate and deliver against your commitments - not hours logged - is the primary signal used to assess your level and growth.

Pushing beyond core hours

When the business asks you to push - a 12-hour day, a 60-hour week, a critical deadline - that's a distinct category, outside the standard full day, and it is specifically rewarded. Pushing is never the baseline expectation; it's the exception, recognized when it happens.

🏗️ Everyone builds

Every role builds

Marketing sources and closes deals - there is no sales team, no SDR, no AE. Product writes specs engineers can build from without back-and-forth, prototypes in Lovable before engineering picks it up, and evaluates AI output quality across the product. Engineers own services end-to-end - you built it, you monitor it, you fix it. Everyone uses AI tooling (Lovable, Claude Code, Cursor, claude.ai) to multiply their output. Nobody is just a coordinator.

Customer support is AI-first

First-line customer support is fully AI-driven via Plain. Blu Chat (our AI agent) resolves the majority of tickets autonomously from our docs across Slack, Discord, and email. Sales questions route to marketing. Product questions route to product. Confirmed bugs land in the engineering backlog. Nobody is stuck in a queue - AI handles the queue.

On-call availability

On-call is opt-in and compensated. Team members who participate receive 15% of their daily rate per on-call day, plus incident-based multipliers for production issues. Rotation is weekly among those who opt in. No penalty for not participating - teams coordinate coverage among themselves.

Customer support rotation

Engineering teams coordinate a weekly customer support rotation. The person on rotation triages customer-reported issues, works with the customer directly, and closes the loop. This keeps engineers close to real user problems instead of building in a vacuum.