From AI experimentation to controlled, evidence-driven software delivery
Agentic AI across the regulated software lifecycle
Discover how Agentic AI in the SDLC can help accelerate software engineering and modernize legacy components — while keeping verification, traceability, quality and human accountability at the centre.
Join a practical, two-hour workshop. See live demonstrations, the engineering controls behind reliable AI agents, and where to start in your environment.
An end-to-end feature creation workflow, with human approval built in
~80 min — end-to-end workflow in the demonstrated example
Under $10 — model-token cost per demonstrated cycle
Governance — a repeatable framework with 12 structural governance dimensions
AI can accelerate development.
Can your quality process keep up?
AI agents can work across requirements, code, tests and reviews. But in a regulated environment, producing code is only part of the job.
The opportunity is not simply to use more AI. It is to engineer a controlled way of working around it.
The real questions are harder:
- AI wrote the code. Will your evidence hold up? Every AI-assisted change needs traceable actions, recorded verification and objective evidence that stands up to IEC 62304 and FDA scrutiny.
- How do you standardise AI use across teams? Every engineer uses different tools, prompts and habits. Without a shared way of working, quality depends on who made the change, not on the process.
- How do you scale without overwhelming reviewers? When implementation gets faster, review and quality assurance can become the bottleneck.
- How do you manage aging software components? Replacing a dependency is not enough. You need a defensible way to assess risk and demonstrate that behaviour remains correct.
These are the practical challenges of AI-native software development for regulated industries: improving engineering workflows without losing control of quality, verification or accountability.
The missing piece: an engineered AI harness. A capable model alone does not make an AI workflow production-ready.
The surrounding system — the AI harness — provides the context, tools, rules, verification and evidence that make agentic work repeatable and reviewable. A governed workflow can follow seven steps:
This means agents work against established project information and explicit constraints. Automated checks and independent reviews help identify problems before work reaches a human approver. Each run produces a trace of what happened and the resulting artefacts.
This structured approach to agentic software development helps connect AI execution with the engineering controls needed for reviewable outcomes.
The principle is straightforward: automate the work where it is appropriate, engineer the controls around it, and keep consequential decisions with accountable people.
One practical workshop. Two real engineering challenges.
Designed for engineering leaders, software architects, quality and regulatory professionals, and cybersecurity teams working in medical device and other regulated software environments, this two-hour AI-native software development workshop connects the technology to real engineering decisions.
Rather than offering a generic introduction to AI, we explore two practical applications: Agentic AI across the software development lifecycle and safe SOUP modernisation.
You’ll see how the workflows operate, which controls matter, where human judgement remains essential, and how to identify a sensible starting point for your own team.
See what it takes to move from a chat-based assistant to a governed engineering workflow.
AI-assisted software development can help engineers generate code, explore solutions and work through technical tasks. But in regulated software development, producing an output is only part of the job.
Teams also need to understand the context behind the change, verify the result, review potential risks and retain appropriate evidence.
The workshop explores how a structured engineering workflow can connect AI-generated outputs with:
- Relevant project and codebase context
- Defined tool permissions and execution boundaries
- Automated verification and testing
- Human review and approval
- Traceable records of what happened and what was checked
The key question: How can you increase engineering throughput without sacrificing control, reviewability or accountability?
Replace legacy dependencies based on evidence, not guesswork.
Regulated software teams must understand the components they use, assess their suitability and manage the risks associated with maintaining or replacing them.
SOUP—Software of Unknown Provenance—is an important consideration in medical-device software engineering. Legacy components and dependencies can create challenges around security, reliability, sustainability and available evidence.
For teams managing SOUP software, the challenge is to make component decisions systematically, with a clear rationale and evidence to support the outcome.
In this session, we explore a practical four-step approach:
- Find: Identify what actually ships, map dependencies and understand which component functionality the product uses.
- Vet: Compare candidates against agreed criteria for sustainability, reliability, security and compliance.
- Replace: Establish a behavioural test baseline, compare before-and-after results, and investigate differences before accepting a change.
- Sustain: Monitor component health and reassess candidates as releases, advisories and requirements change.
The key question: How can you make dependency decisions repeatable, measurable and defensible — while reducing the effort required to maintain your software supply chain?
What you walk out knowing
A practical view of the workflow, controls and engineering decisions behind applying Agentic AI in a regulated environment.
How governance works
12 dimensions from workflow and architecture to coding and security
How the AI harness is built
Our own structured set of 100+ version-controlled files, in the same repo as application
What one feature costs
A full end-to-end workflow in ~80 minutes and under $10
Bring a real engineering challenge.
And we turn your questions into practical next steps.
Bring one challenge from your own environment: a codebase constraint, a delivery bottleneck, or a SOUP or legacy component that needs closer assessment. Together, we’ll explore the problem, identify relevant controls and discuss what a practical next step could look like.
Workshop format: Two hours, on-site or virtual.
Turn one connected-services workflow into a working AI prototype.
Whether it's fault triage, OTA adoption, or compliance reporting, Star designs and engineers the AI experience layer on your existing connected-services platform — no infrastructure rebuild required. Let's scope the first working prototype.
FAQs
Agentic AI uses AI systems to plan and carry out multi-step tasks within a defined workflow. In regulated software development, that workflow should include controlled access to project context and tools, automated verification, traceable outputs and human review at appropriate decision points.
