Anthropic and OpenAI
Applied AI
Production builds on the Anthropic and OpenAI stacks.
- Engagement window
- 2–8 weeks
- Pod composition
- 1–3 senior experts + orchestrated agents
- Pricing
- Stated in proposal
What we do here
Custom agents, Claude Code skills and workflows, Custom GPTs and ChatGPT Apps, Realtime voice surfaces, retrieval pipelines.
Our internal substrate is Anthropic, so we know it cold — Agent SDK builds, Claude Code skills, Managed Agents deployments, Cowork integrations. When a client has standardised on OpenAI — GPT models on the Agents SDK and Responses API, Custom GPTs in ChatGPT, Realtime voice — the deliverable runs on that stack. The pod composition stays consistent; the platform meets the engagement where it lives.
What we ship in this discipline
- Custom discipline agents on the Agent SDK or Agents SDK
- Claude Code skills for internal teams
- Custom GPTs and ChatGPT Apps
- Realtime voice surfaces
- Retrieval and grounding pipelines
Capabilities
Anthropic stack
OpenAI stack
Models
Build tools
Retrieval
The list is what we build with — not what we promise to install. The right tool for the work is named in scope.
Other capabilities
The same operating model, applied across the practice.
Greenfield
Bespoke Apps
TypeScript-first product engineering for web apps, mobile-shaped surfaces, integrations, automations, and migrations. The fastest path from request to running system, on a stack your team will recognise.
Discipline
ServiceNow
ITSM, CSM, HRSD, and ITOM extension. Flow Designer, Integration Hub, Service Portal, and scoped applications — built and accepted past sign-off.
Discipline
Microsoft
Significant government and enterprise estate runs on Microsoft. We extend it where the platform alone needs custom logic, integration, or a delivery cadence the partner network cannot match.
Newest discipline
Salesforce
Flows, Apex, LWC, and integration work for organisations running Salesforce as the customer-record-of-truth. Ramping in line with confirmed client demand.
Cross-cutting and standalone
Testing
Automated regression suites, test infrastructure standup, coverage audits, and post-incident hardening. Engaged as its own pod when the work is test-shaped, and folded into every other discipline as a Capability by default.
Begin
If the work lives here, take the assessment.
Free, self-driven, no sales call. A proposal — outcome pod or day-rate pod — within 24 hours.