Skip to content
POET
[POET] Work

The work we take on, and how engagements actually run.

POET runs across four capabilities — Build, Ship, Scale, Speak — and a small set of recurring engagement shapes. Below is what we typically build, how we engage, and the cadence a typical project runs at.

[01] What we build

Ten recurring kinds of work.

Most engagements start with one of these. Many compose into multi-capability programmes once the first surface lands.

  • AI agents

    Production agents that operate inside your stack — your CRM, ticketing system, data warehouse — with explicit policies, scoped permissions, and full audit logs.

  • Internal AI platforms

    Custom RAG, retrieval, and knowledge platforms built on your private data. Replaces 8-tab juggling with one operating surface for your team.

  • Native iOS and Android

    Real native builds in Swift and Kotlin (or Expo when cross-platform earns its tradeoff). End-to-end ownership through App Store and Play Store submission.

  • Web product platforms

    Modern web on Next.js + Vercel. Type-safe end to end, server components, and the rigor you'd want for a customer-facing surface or an internal admin.

  • Shopify Plus storefronts

    Theme architecture, checkout extensions, and customisation that holds up at peak. Internationalisation, multi-currency, and the integrations the finance team needs to close the books.

  • Headless commerce

    Hydrogen and Oxygen builds when the shape of the business genuinely needs it. We pick the level of decoupling that earns its complexity, no more.

  • Subscription and lifecycle

    Recharge, Skio, or custom subscription engines. Lifecycle marketing flows in Klaviyo. Attribution that maps cleanly to spend.

  • Editorial systems

    Sanity or Contentful, modelled on how your team actually creates. Multi-channel publish targets so a single asset earns its full reach.

  • Social operations

    Channel architecture across LinkedIn, Instagram, TikTok, X, YouTube. Creator and partnership workflows. Reporting your CFO can read.

  • Custom internal tools

    Admin platforms, ops dashboards, workflow tools. Fast, opinionated, designed around your team's real workflow — not a generic SaaS persona.

[02] Engagement models

Four engagement shapes.

We pick the engagement shape from the problem. Most operators land on one of these four, with the option to evolve toward an operating-partner retainer over time.

  • Greenfield build

    8 – 16 weeks

    Start to ship. We own the spec, the architecture, the build, and the launch — including App Store submission, domain setup, and the first weeks of telemetry.

  • Replatform

    10 – 20 weeks

    Migrate off legacy with a zero-downtime cutover plan. We've moved brands and operators from Magento, WooCommerce, bespoke CMS stacks, and homegrown internal tools.

  • Operating partner retainer

    Ongoing

    Ongoing engineering and operational ownership across the systems we build (or systems your team already runs). Not hours-billed. Outcomes-billed.

  • Audit

    2 – 4 weeks

    Security, performance, or architecture deep-dive on a system you operate today. We deliver a punch list ranked by impact, with recommended remediation paths.

[03] How it runs

The cadence of a typical engagement.

Four phases. Each ends with something tangible. No "we'll get back to you in 90 days." Discovery ends with a signed scope. Build ends with a production surface.

  1. [01]

    Discover

    Stakeholder interviews, system access, baseline metrics. We end this phase with a one-page scope your operators can sign off on without ambiguity.

    1 – 2 weeks
  2. [02]

    Design and scope

    Architecture, key flows, and the prototypes that resolve the biggest questions. We lock the v1 surface area and write down what's deliberately out of scope.

    1 – 2 weeks
  3. [03]

    Build and iterate

    Weekly working sessions, demo-driven progress, and shippable increments. You see the system in production-shaped form long before launch.

    4 – 12 weeks typical
  4. [04]

    Operate

    We either keep operating on retainer or hand off a documented repo, infrastructure-as-code, and runbooks your team can inherit cleanly.

    Ongoing or handed off