Hacking GenAI: from ideas to organisational impact

Recently, myself and Harry (one of our Senior Account Directors) travelled up to York to visit our clients at STEM Learning.

The reason was an exciting one. We'd be attending a hackathon along with their other digital partner, AWS, to explore how STEM could harness the power of Generative AI (GenAI) to improve problem areas, optimise efficiency and enhance the commercial offering across their organisation.

We started working with STEM Learning in January 2024 with the ambition to accelerate their digital maturity by significantly transforming their digital ecosystem. We're all about collaboration, coming up with new ideas and adding value through digital acceleration - and a hackathon is a perfect blend of all of these things.

But... what is a hackathon?

If you don’t know what a hackathon is, it’s when a group of multi-skilled stakeholders get together for a condensed working session. Working through from ideation, to solution, test and build. The idea is you leave with an understanding of whether a solution is viable and should be taken forwards into further development.

Fuelled by excitement and award-winning carrot cake, even some bad weather and loss of internet connection couldn’t stop us. It was extremely insightful and there were several learnings that I personally took away around it that I wanted to share around what to consider when implementing AI from ideation to productisation.

What is GenAI (and what it is not)?

You've likely been living under a rock if you've not heard of GenAI by now, but just to cover all baes... GenAI is the tech that's quietly revolutionising everything from art to customer service. It refers to AI models that create content like text, images, audio or video and can be used in a diverse range of applications.

However, before you start thinking about how you could use it, it is good to understand what it can do and what it can’t. GenAI excels at creating realistic, context-aware content like text, images but lacks deep reasoning or true understanding; it’s a powerful tool for automation and creativity, but not a replacement for human judgment or expertise.

But why were STEM Learning invested in understanding more about GenAI? One of the things they wanted to understand is how the ways in it could support the CPD (continued professional development) for teachers. Could it help speed up the development of new course structures based on existing materials? What about teaching materials? Could GenAI play a role in creating high quality teaching CPD content, at pace, and find new ways to help deepen students’ knowledge of STEM subjects?

During the hackathon, we went through some typical use cases for GenAI. Including, content generation (e.g., creating blog articles or educational materials), summarisation to extract key points and metadata for enhancing the accessibility of legacy content, improving search by interpreting diverse data to deliver accurate insights from unstructured sources, and enhancing productivity through automation across tasks such as coding and business intelligence.

Key takeaway : GenAI's real power doesn’t just lie in what it creates, but also how it can inspire new ways to solve problems, making hackathons the perfect environment to explore its potential and push the boundaries of its capabilities through rapid experimentation and innovation.

Where do you start?

The short answer to this question is not with the technology. There is always a temptation to get your hands on the tools and start tinkering. Whilst this is great for getting to grips with the tooling and its general capabilities, you should first identify the problem you are trying to solve and then explore how GenAI could help.

For the team I was in at the hackathon, this involved lots of collaborative discussions around a specific use case, writing on whiteboards and flip charts, and an application walkthrough with an internal subject matter expert to understand the broader context of the problem(s) that currently exist, then narrowing it down to a specific problem area and exploring potential solution’s around it.

Key takeaway: Once you have a potential use case for GenAI, start by focusing in on a small discrete task and use a subset of available data to prove the concept, learn from the results, refine, and iterate to improve.

Rapid and accessible prototyping

Once you’ve generated some ideas that address key pain points, the next step is to validate these ideas quickly. Having a toolset that will enable you to quickly create a proof of concept is essential, especially under the time constraints of a hackathon.

An example is AWS PartyRock, a user-friendly, low/no-code tool which we were provided access to at the hackathon. It is underpinned by Amazon Bedrock, a fully managed AWS service that provides access to various foundation models (FMs), including Large Language Models (LLMs), through a common API.

PartyRock provided us with the ability to privately upload the publicly available data we had into the Model and create prompts to guide the models to deliver accurate and contextually relevant outputs as either text, image or chatbot.

Using these types of tools lowers the barrier to entry, enabling teams to rapidly create prototypes using intuitive and easy-to-use user interfaces without requiring any programming experience and the underlying infrastructure or APIs are all taken care of.

Key takeaway: using a series of focused prompts with concrete instructions will produce more accurate results than a single larger, more complex one.

Great idea. Concept validated. What’s next?

Implementing GenAI applications into production requires comprehensive and careful planning and execution. And there is a lot to consider.

Begin by establishing a clear AI strategy that outlines how GenAI will be utilised across your organisation. This should identify business objectives, address key pain points, highlight opportunities for added value, and ensure alignment among all stakeholders including technical teams, business leaders, and end users on its implementation, application, and usage.

Developing an integration plan is also critical so that GenAI can be seamlessly embedded into existing workflows, tools, and processes for maximum impact.

The plan must ensure AI applications are ethical, secure, and scalable, with other key considerations such as observability, controllability, fairness, transparency, and sustainability included.

Key takeaway: whilst defining the AI strategy, starting with an internal pilot can be a practical first step to validate use cases, refine ideas, and assess feasibility before scaling up.

Bonus top tip: struggling to get started with ideas on how GenAI could help – try asking GenAI!

The hackathon was a success and is a great approach to getting started with under understanding how GenAI could help your organisation. If you would like help facilitating an internal Hackathon or guidance on defining an AI strategy, then drop us a line gary.trimnell@greatstate.co.