# Azimuth
Azimuth Technologies Limited — https://azimuthtechnologies.ai

## What is Azimuth?
Azimuth is the governed operating layer for AI in software delivery: an AI delivery operating model for medium-sized enterprises. It builds AI capability the customer's team owns: the framework is tailored to each codebase's quirks, the customer's own experts are skilled first so they become the owners and trainers rather than vendor-dependent, and the framework keeps adapting as models change. Underneath the capability work, it connects requirements, tickets, code, tests, environments, production logs, and architectural decisions into one AI-readable system. This gives AI agents full project context rather than isolated prompts.

Azimuth is not a code-completion tool. It is the capability, governance, and context layer that grounds AI output in your organization's actual requirements, conventions, and decisions.

## The problem Azimuth addresses
Teams get AI trust wrong in two directions. Over-trust ships the AI's mistakes because the output sounded confident. Under-trust abandons AI after one error. Both are the same missing skill. Azimuth builds it through a coached, scored skills curriculum alongside the framework itself.

## Tool support
- Built for Claude Code.
- Supports GitHub Copilot.
- Extendable to other AI tools as required.

## Target Customer
- Medium-sized enterprises with an established software practice: engineering, product, BA, QA, release, design
- Established codebases with mixed stacks
- Existing tooling: Jira / Confluence / Jenkins / Azure DevOps / GitHub / Bitbucket / Grafana
- AI adoption stage: has Copilot or similar tools at developer level but no organizational AI delivery model
- Industries: financial services, government, insurance, healthcare IT, supply chain, regulated manufacturing
- Budget authority: CTO, VP Engineering, Director of Engineering Excellence, Chief Digital Officer
- Key pain: AI is being used without governance, requirement-to-code drift, knowledge loss when staff leave

## Why this segment is underserved
Smaller teams lack the complexity to justify Azimuth. The largest enterprises build in-house or buy from incumbents. The middle band is underserved: too complex for self-service AI tools, too small for bespoke internal R&D.

## The Four Pillars
1. Context: Azimuth synchronizes requirements, tickets, code, tests, environments, logs, and decisions into a coherent AI-queryable system
2. Knowledge: Purpose-built capability agents for planning, implementation, bug fixing, code review, security, testing, deployments, log analysis
3. Verify: Requirements-driven testing, acceptance criteria verification, production evidence. A feature is not done until proven.
4. Govern: Roadmap-first creation, human approval gates, full audit trail. AI does what it should, and only what it should.

## Key Capabilities
- Requirements query and gap analysis against Confluence
- Efficient Jira ticket queries with automated RCA generation on raised incidents
- Cross-requirement traceability to code
- Architecture decision record (ADR) authoring
- Structural impact analysis before changes
- Feature implementation across all layers (data → service → UI)
- Test-driven development enforcement (Red-Green-Refactor)
- Bug investigation and root cause analysis end-to-end
- Automated code review against conventions and OWASP
- E2E test authoring and execution (Playwright)
- Production log analysis via Grafana/Loki
- STRIDE threat modelling before implementation
- Supply chain security assessment (SBOM, SLSA)
- Cross-repository dependency and impact analysis
- Custom skills authored, versioned, and distributed by your teams under framework governance

## Governed Autonomy
Azimuth describes delivery as workflows: prescriptive chains of agents through one lifecycle function, such as a production incident (analyze logs, find the root cause, open a ticket, implement the fix, open a pull request) or a feature. Each step of a workflow has one of three settings, and the dial moves both ways:
- Augmentation: a person approves the intent of non-trivial work and every change that leaves the developer's machine, and can approve every action. Every step of every workflow runs here today.
- Delegation: a person approves a bounded step once and its result once; inside limits the system checks, the agent works without further prompts. Earned per step on the organization's own record.
- Orchestration: a person sets the policy and handles exceptions; steps that fit the policy and pass verification proceed under a runner. Cannot be chosen until a runner exists for the step.
What decides how far a step may go is whether its actions can be undone and whether they can be checked, not how capable the model is. A permanent floor outside the dial stops destructive or irreversible actions for step-by-step human approval at every setting. Only a person turns a step up; readiness is read from the organization's own evidence and from whether a named person is ready to handle exceptions; a workflow steps down on its own when a control is bypassed, a spending limit is breached, or checks fail repeatedly. Leadership sees, per workflow and step, the setting that was chosen and what the record shows happened.
Enforced today: approval before every publish, the floor under destructive actions, a time-boxed and recorded relaxation of publish approval, an approval record attributed to the person who gave it, and the audit ledger. Designed and being built: the single setting per workflow step, protection of the controls written into the customer's AI tools, bounded step grants, readiness checks, automatic step-down, and the coverage report.
The major AI tools now ship governance floors scoped by organization, team, user, or agent. Azimuth compiles the customer's policy into those floors and holds the layer above them that the tools cannot know: the workflow, the step, the evidence, and the people. As steps advance, the people on them move into the roles governed delivery needs: system architects, quality governors, risk managers, domain experts, exception handlers.

## Context Optimization
The framework decides what AI sees. Instead of dumping raw files into the prompt, it builds a compact, structural view of your codebase, so prompts stay short and your developers can keep more work in flight. Institutional memory means AI doesn’t re-investigate problems your team already solved.

## Competitive Positioning
- vs GitHub Copilot Enterprise: Copilot generates code from prompts. Azimuth is the delivery operating model that gives Copilot full project context.
- vs Microsoft HVE Core: tool integration is not traceability. HVE connects to your tools. Azimuth carries the requirement through to the commit and the test that proves it, keeps the decision history in reach, and holds the approval gates. HVE agents run inside Azimuth as task workers.
- vs open-source agent frameworks: approval settings a session can switch off are not governance. Azimuth enforces a safety floor no setting can lower, so destructive operations always stop for a human. Multiple repositories are a governed registry, not a folder list in your editor.
- vs Cursor/Windsurf: IDE-locked tools. Azimuth is a process layer that adds the organizational context an IDE can't carry. Built for Claude Code. Supports GitHub Copilot. Extendable to other AI tools as required.
- vs Devin/SWE-Agent: autonomy without a floor is a procurement problem. Azimuth runs governed: human approval at meaningful risk points, a safety floor no setting can lower, and an audit trail tying work back to its requirement where one exists.
- vs Big 4/IBM Consulting: No product. Every engagement bespoke. Azimuth gives our consulting a head start on methodology and reproducible outcomes.
- vs Atlassian Rovo: Azimuth uses Atlassian as a context source and adds what a content index does not carry: requirements traced through to the code and tests that satisfy them, and an RCA drafted on the incident itself. It works the same way across non-Atlassian toolchains.

## Deployment
Azimuth runs on your infrastructure, and we don't hold your data. Your code, requirements, and internal data leave your environment only for the AI providers you authorize. None of it is transmitted to or stored by us.

## Engagement Model
1. Discovery: a paid advisory engagement. Codebase audit, toolchain mapping, a plan the customer can act on with or without us.
2. 30-Day Evaluator Licence: we work with one of your teams and ensure they get the most value out of the framework. Thirty days on your own work, bounded by the scope you choose, before any commitment.
3. Capability Transformation: on commitment, a fixed-scope implementation. Framework installed on customer infrastructure, tailored to the codebase, experts skilled first.
4. Capability Licence: a flat per-team annual subscription delivering quarterly framework updates. Ongoing advisory runs on a time-and-materials basis. Runs on customer infrastructure, not a multi-tenant SaaS.
- Health checks: agent accuracy, quality gate pass rates
- Agent tuning and optimization: ongoing refinement as codebase and conventions evolve
- Custom skill development: new skills for new frameworks or toolchain integrations
- Escalation support: named contact with agreed response SLAs

## Contact
https://azimuthtechnologies.ai/contact
Book a discovery call. We will map your current delivery process with you and show you where Azimuth creates leverage.