Implementing Agentic AI into product development

Agentic AI is rapidly emerging as the next major shift in software engineering. For product teams, the challenge is no longer simply whether to adopt these capabilities, but how to integrate them in a way that is effective, secure and genuinely valuable to customers. 

I recently had the opportunity to attend AWS Summit London 2026. As a Technical Lead within our Product team, I am always looking for emerging capabilities across AWS services that we can meaningfully embed within our InformedDECISION™ and InformedINSIGHT™ platforms. Alongside the usual benefits of these events, including great conversations with like-minded engineers and a few inevitable goodie bags, this year felt notably different.

While much of the exhibition floor still focused on the steady, incremental improvements we have come to expect from cloud infrastructure, the overall shift in direction was clear. We are moving beyond AI as a chatbot-only interaction model, where tools are used occasionally for code snippets or summaries, and into something far more transformative: the era of the AI agent.

 The rise of Agentic AI

A consistent theme throughout the keynote sessions and technical talks was the emergence of Agentic AI. AWS’s agentic development ecosystem, including Kiro, Amazon Bedrock and Amazon Bedrock AgentCore, featured prominently, alongside developer platforms such as OpenAI Codex and Cursor. However, the real story is not any single tool. It is the broader shift in how we think about AI in the development lifecycle.

This is not just conference momentum. McKinsey’s 2025 State of AI research found that 88% of survey respondents report their organisations now use AI regularly in at least one business function, while 23% are already scaling agentic AI somewhere in the enterprise, and a further 39% are experimenting with AI agents.

We are moving away from passive, point-based interactions and toward systems that can actively help achieve goals. These agents are not just assisting with isolated tasks. They are beginning to contribute across parts of end-to-end workflows, from interpreting requirements to scaffolding environments and iterating on early-stage prototypes.

In practical terms, this represents a shift from AI as an advanced autocomplete tool to AI as a proactive collaborator.

This could be a key step change in productivity. Evidence from AI-assisted development already points to the scale of the opportunity, with controlled research into GitHub Copilot showing developers completed a defined JavaScript coding task 55.8% faster when using the tool. While this is not a direct proxy for agentic development, it illustrates the productivity potential of applying AI thoughtfully to software engineering tasks.

The focus is no longer just on writing code faster. Instead, we are exploring agentic workflows that can orchestrate complex development activities. This has the potential to significantly reduce the gap between a good idea and a production-ready application, provided this is paired with robust engineering, testing, security and assurance practices.

Embedding innovation into product delivery

Within the Informed Product team, we are embedding Agentic AI into our product delivery approach in a considered and assured way.

Our approach is not about adopting new technology for its own sake. It is about applying the same engineering discipline, quality standards and critical thinking to our internal product development as we do when delivering for our clients.

By integrating Agentic AI into our development cycles, we are creating a structured environment to explore, validate and refine these approaches. We are identifying where agent-based workflows provide the most value, whether that is automating infrastructure setup, accelerating prototyping, or supporting iterative feature development.

The opportunity is significant, but so is the need for discipline. The evidence shows that value does not come from adoption alone, but from embedding AI into the right workflows, controls and measures of success. For product teams, the differentiator will not simply be access to agentic tools, but the ability to embed them into well-governed delivery workflows: clear requirements, secure environments, human review, automated testing, auditability and measurable outcomes.

This allows us to de-risk adoption. Rather than reacting to industry hype, we are building practical experience that informs how and where these capabilities can be applied safely, effectively and at scale across our product and consulting work.

The “so what?”: Innovation with assurance

For us, the implications go beyond productivity gains.

We operate in highly regulated environments where trust, governance and accountability are essential. As we explore the potential of Agentic AI, we must do so in a way that is secure, transparent and aligned with the needs of our clients and their customers.

This matters because public trust is not moving at the same pace as technical capability. EY’s 2026 UK AI Sentiment Index found that 74% of UK respondents had used AI in the previous six months, but only 14% were comfortable relying on fully autonomous, agent-led systems.

This is where AI assurance becomes critical. Our ISO/IEC 42001-certified AI Management System provides a structured framework for governing the responsible design, development, deployment, and improvement of AI systems. It helps ensure that as we push forward with innovation, we do so with the right controls, oversight, and ethical considerations in place.

Being at the forefront of this technology is important, but it must be balanced with a thoughtful and disciplined approach. Our goal is not only to unlock the benefits of AI for our partners, but to do so in a way that builds confidence and delivers real, lasting value.

This is what we are known for, and it is what we will continue to prioritise as these technologies evolve.

Looking ahead

Personally, I value the opportunity to help shape our growing product business while continuing to develop my own skills and perspective in this rapidly changing space. It is an exciting time to be working at the intersection of AI, data and product development.

We are actively building and sharing AI-driven innovation across our platforms and continuing to grow our teams across Informed. Our work spans impactful sectors, including healthcare, the environment, and emergency services, where we use AI and data for good.

We have already seen the impact this kind of product-led innovation can have. InformedDECISION™ has supported significant improvements in NatureScot’s casework, including a 50% reduction in case triage time and a 40–60% reduction in processing times for routine statutory consultations, helping teams make faster decisions without compromising quality.

If you are interested in joining a growing team tackling meaningful challenges with modern technology, we would be keen to hear from you.