Agentic AI Playbook: How MedTech Teams Must Evolve

Agentic AI changes the job of everyone who touches a regulated software product, not only the engineers. This Playbook is written for the five roles whose work shifts most:

• Software engineers, moving from writing every line of code to defining the contracts that govern how code is generated
• Product Owners and Product Managers, moving from specifying screens to defining clinical intent
• SDETs, moving from writing automation scripts to governing behavioral contracts
• Quality System Engineers, moving from populating compliance documents to governing systemic risk
• Engineering and product leaders, who set the measures their teams work to

If your team has bought AI tooling and has not yet seen the return, it is written for you.

• How five roles change as agents take on more of the drafting, analysis and review: engineer, Product Owner, SDET, Quality System Engineer and leader
• What orchestration looks like in practice, defining constraints and acceptance criteria rather than producing every artifact by hand
• How to set the guardrails agents work within, from behavioral contracts to regulatory boundaries
• Where clinical and regulatory context has to come from people, and why that judgment matters more in an agentic environment rather than less
• How to redesign cross-functional workflows, and the measures leaders use, so teams are rewarded for orchestration rather than artifact volume
• How to review and audit machine-generated output before it moves further through the lifecycle

Most MedTech teams adopting Agentic AI buy the licences and provision the frameworks, then find months later that AI is being used as a glorified spellchecker and the expected return has not arrived.

The constraint is usually the operating model rather than the technology. When agents can draft code, requirements, tests and risk documentation, value moves from producing artifacts to directing and reviewing them. Teams that still measure themselves on artifact volume push the AI aside and stay where they were.

Our article How MedTech Teams Must Evolve for the Agentic AI Era examines that pattern in depth. The Playbook turns it into role-by-role guidance you can act on.

The Playbook works through the shift role by role, covering:

• What changes when agents take over first-draft production, and why tooling alone does not deliver the return
• The shift for software engineers, from writing logic to defining contracts
• The shift for Product Owners, from specifying screens to defining clinical need
• The shift for SDETs, from maintaining scripts to governing behavioral contracts
• The shift for Quality System Engineers, from producing documentation to governing risk
• The shift for leaders, from managing output to redesigning how the team works
• What stays with people: architecture, clinical context, acceptable risk, and the final call on whether the result is safe and effective

Larkin Lowrey, Chief Technology Officer, Orthogonal

Larkin Lowrey is a veteran software engineering leader with over 30 years of experience across IoT, medical devices, telecom and e-commerce. He developed a telematics platform later acquired by Verizon and now leads cloud-native, analytics-driven software development for regulated medical technologies. He holds 26 U.S. patents and applies Agile principles to build high-performing teams.

Why Orthogonal

Orthogonal has spent over a decade building software for Class II and Class III connected medical devices and Software as a Medical Device. We work under an ISO 13485-certified and IEC 62304 and ISO 14971-compliant quality management system, with FDA, EU MDR, UK MDR, DiGA and GDPR regulatory experience, and we apply Agile methods to medical device software and SaMD.

What is Agentic AI in MedTech?

Agentic AI describes multi-agent systems that can plan, reason and execute product-development tasks rather than simply completing text or code. In MedTech that means agents helping to generate requirements, draft user flows, write and test software, and produce supporting documentation, inside a process that still has to satisfy design controls.

How should MedTech teams prepare for Agentic AI?

Start with a bounded use case rather than a tool selection. Define the problem, the constraints, what success looks like and a review process matched to the risk, then expand based on results. Teams also need to define what each agent may do, what context it can access, and when a human must review or escalate.

Which product, engineering, quality, and leadership roles change most?

Five: software engineers, Product Owners and Product Managers, SDETs, Quality System Engineers, and engineering and product leaders. In each case the shift is from producing artifacts by hand toward defining constraints, supplying clinical and regulatory context, and reviewing what agents produce. Leadership determines whether the rest can happen, because it sets the measures teams work to.

How do design controls and human review apply to Agentic AI outputs?

AI-assisted work sits inside the same design controls that govern the rest of development. Requirements, risk controls, tests and evidence stay linked, and agent output is reviewed before it moves further through the lifecycle. Engineers, quality professionals and clinical experts remain accountable for architecture, acceptable risk, and whether the result is safe and effective.

Orthogonal helps MedTech teams design, build and validate AI-enabled products under full design controls. Explore AI-enabled medical device software development.

• Automating Compliance with AI in the MedTech SDLC
  How MedTech teams are making compliance, testing and traceability continuous outputs of the development lifecycle.
• Agentic AI for MedTech: From Requirements to Tested Software
  A recorded webinar on connecting product intent to working, tested software with agentic workflows.