Nine’s AI journey: from prompt engineering to context engineering and beyond

Does AI success at Nine just mean perfect prompts? As you can imagine, it does not

With widespread use of AI, here at Nine we have been enthusiastically experimenting with Generative AI for opportunities to amplify our capabilities. Empowering our engineers to focus on complex design, creative problem solving while leaving AI to help through mundane tasks and accelerate prototyping.

But have you stared at your AI co-pilot, wondering why its response was bizarre or generic? So you rephrased your prompt to be crisp and clear, ticked all those best-practice boxes and yet another superfluous reply? Turns out you can structure the best Prompt, but you also need to equip your chosen AI co-pilot with tools and the right context to land you in your desired destination.

Properly interacting with AI is crucial to deliver value outcomes

So we took a step back to reflect on where to begin and how do we empower our teams to interact with AI efficiently and with ease to maximise value delivery? This led to collaboration within tech and business teams across Nine, initially looking to streamline Prompt Engineering for scalability. 

We were keen to identify engaging use-cases that would inspire our engineers to explore. This sparked several initiatives, including a Gen-AI diagram translator assisting with extracting components from an architecture diagram and converting them to code. To widen our coverage of scenarios, we diversified our approach by prompting Large Language Models (LLMs) directly as well as through AI Agents. Our learnings from these AI explorations, culminated in a systematic approach to AI communication. This involved centralising re-usable artifacts into a Prompt Gallery and creating robust Prompt templates that define persona, task, and structured outputs.

Selecting the right inputs or context for our tools and AI solutions goes hand-in-hand with prompt engineering

And in our pursuit of more refined outcomes, we stumbled into the world of Context Engineering. The need for higher precision in some of our code generation use-cases, led us to evaluate Hashicorp Model Context Protocol (MCP) for Infrastructure as Code (IaC) and the Data Build Tool (DBT) MCP for data transformations. Context goes beyond Prompts and encompasses every input to Gen-AI models including but not limited to instructions, documents, articles, github code, memory, Model Context Protocols (MCP) and other internal tools. With context the possibilities seem endless, but the core engineering challenge is going to be about finding the optimal balance between latency, cost, and accuracy.

Tailored guidance on prompt and context engineering will lay the foundation for AI democratisation at Nine

While our initial focus was Prompt Engineering, our journey paved the way to an understanding of Context Engineering, agentic-AI and Model Context Protocol (MCP). Prompt engineering as a discipline did not stand in isolation, the Context weighed in equally. 

Safe & governed democratised access to agent-building capabilities will empower our teams to deliver transformative value with the help of AI

We have barely scratched the surface, but seem to have set in motion a domino effect here at Nine – defining prompt engineering standards is steering us towards context engineering and beyond. As we further explore AI and move towards leveraging AI agents, we are beginning to look at Prompt, Context, Workflow breakdown and AI Solution design more holistically. 

And while we are taking stock of our learnings to help Nine deliver value, we find it crucial to also invest time thinking through governance mechanisms, setting up guardrails for secure implementations, designing Human-in-the-Loop workflows to mitigate risk, observability, performance and cost early on.

There has never been a better time to join us!

We’re at the brink, defining the future of AI at Nine. What is the key to unlocking AI’s potential? Is it the prompt, the context or the Generative AI model itself? Under the hood, it is our engineers and their pioneer spirit with a deep-seated curiosity for solving problems. They are the strategists, the cognitive modelers designing and maneuvering our Gen-AI models of the future. 

If you’re passionate about AI, data, and solving creative technical challenges this is an exciting place to be. At Nine, we’re not just using Generative AI, we’re building the playbook to enable our teams to work with AI effectively and responsibly.