Using GenAI for better failure

Whether you’re bullish on the impacts of AI or not, it’s hardly a revelation to state that these technologies have implications for software design and development. One of the clearest impacts is acceleration.

We’ve been working with machine learning and artificial intelligence platforms, both open-source and licensed, for some time now. We’ve used GenAI to deliver customer-facing chatbots, assist contact centre staff, and to automate aspects of software development and testing.

But we wanted to share one key application for GenAI at Marino Software, in our rapid prototyping phase. By using GenAI to deepen working prototypes, we’re not focusing on speeding up development cycles. Instead, we are focusing purely on enriching conversations within our teams. Put simply, GenAI is helping us to fail faster.

GenAI: a way to faster failure

Using GenAI we are spinning up working versions for demonstration and discussion. We don’t see these initial “wrong answers” as setbacks. They’re the failures that get us to the right answer. These enriched prototypes expose and illuminate biases, prioritisations, dependencies, assumptions, a host of dynamics that need to be explored on any project.

Historical precedents in the tech industry show the merit of this approach. Working software modules provide a solid foundation for making failures real. The key change GenAI introduces is the speed component: it avoids an overinvestment of time and eliminates effort. This makes prototyping more viable as a project component.

GenAI and functional prototyping

Incorporating GenAI into our prototyping efforts helps us iterate and evolve project concepts. GenAI tools assist both our design and development teams in crafting and revising testable models and code, at speed. This means we don’t just end up with a concept or a semi-interactive mock-up; we achieve a working prototype that can be demonstrated, discussed, tested, analysed, improved.

The right context needs to be established for this to happen, of course. It needs to be clear to all in the room that “the working software is not the software that will work”. But that’s a matter of good working practices, and solid relationships with our clients and partners. Where the purpose of a piece of prototyped software is clear and understood: as a basis for conversation.

Software design and development as a social process

This use of GenAI plays into our understanding of software design and development as a fundamentally social endeavour. The stereotype of the lone programmer single-handedly pushing the boundaries of software innovation is a persistent myth.

That may be the spark of an idea, but the reality is that in our industry, good software is the result of collaborative effort. People agree to come together around a problem or an opportunity. Those people organise themselves in a way that ideally leads to the optimal solution, within the typical constraints of budget, time and quality. Technologies of all kinds, including GenAI, are the tools we use.

What we do is essentially a social initiative, based on a technical capability, using collective knowledge and diverse perspectives for the best result. It’s not often portrayed as social. As human. But that’s what it is.

The integration of GenAI into our rapid prototyping phase is just one application for these technologies. There are many others we’re exploring, such as the design and development of multi-agent tools. We have been iterating these in-house for the past 12 months, and they’re providing clear benefits for us. But we’d still see these efforts as exploratory. It’s clear to us however that GenAI tools may have applicability across the software design & development cycle. Producing prototypes that enrich conversations about software is a good start.

Our recent in-house GenAI day produced working prototypes of over a dozen tools and apps, including: a learning tool to master Data Structures and algorithms through daily practice, an application that predicts if you’ll get wet waiting for your bus, and a tool to recommend when to spread organic fertiliser (using soil moisture deficit data and county weather forecast data).

In fact, what we’ve seen is that GenAI is helping us to unlock developer creativity.