Tree, a software specialist at leading research and development (R&D) tax credit consultancy ForrestBrown, scrutinises the world of artificial intelligence (AI), arguing that UK companies could be misinterpreting the eligibility of AI and non-AI software and digital projects when approaching the government’s R&D tax incentives scheme. Here, he explains why companies shouldn’t necessarily exclude a trigger-driven relational database or rule engine from a software R&D claim.
AI has long been portrayed as both an opportunity and a threat to humankind. We’ve seen AI medical image processing identify a large volume of eye-patients’ macular degeneration, and we’ve seen deep learning trawl astral data searching for habitable exoplanets. Yet the negative hype around AI would have you believe that we’re heading straight for a Terminator-style apocalypse. The reality of course, is far less extravagant.
Identifying and demystifying AI
Today there are plenty of systems that are AI-led, like convolutional neural networks for example, which can be used to identify faces in a picture. Such AI services, from the likes of tech heavyweights Amazon, Microsoft and Google, are fully integrated into applications and do all of the leg work for the end-consumer. As a result, understanding around how these systems work has remained minimal, and AI has become very commoditised.
Perhaps this is one of the reasons that there’s a heavy reliance from companies to label their applications and solutions as artificial intelligence. Right now, there are plenty of ongoing ‘AI’ projects that software companies are undertaking around the world, and this includes everyone from the smallest SMEs up to the largest multinational companies. Take Google and its Dialogflow solution for example. It enables users to create conversational experiences across devices and platforms – and it’s highly effective at helping users to build solutions jam-packed with logic and conditioning.
Is it highly innovative? Definitely. But is it solely AI? I wouldn’t be so sure.
Put simply, AI – or machine/deep learning as it’s also known – has become a somewhat used and abused term. It’s really just another part of a software developer’s toolkit, allowing them to create the best applications they can. There’s no doubting its usefulness to perform tasks that would normally require human intelligence – such as decision-making, speech recognition, visual perception, and the translation of languages.
Since entering the sector, and through my work in the gaming industry during the 1990s, I’ve found the name ‘artificial intelligence’ to be quite the misnomer, as ultimately, the intelligence you need to apply to software applications or solutions to make them workable needs to be initially created by a human. With many software solutions of this type, we’re essentially aiming for ‘apparent intelligence’, with reliability and believability already built in and tested by humans, instead.
And while AI, machine-learning and deep-learning might be today’s buzzwords, ‘apparent intelligence’ is nothing more than an illusion created by a developer, an algorithm and good old-fashioned logic. That’s because at its core, AI is simply a system that controls behavioural algorithms. In gaming, this was about creating an enjoyable experience for players, while in other applications – such as big data – it’s about predictions, gleaning insight, and detecting anomalous activity.
There’s become an expectation over time that AI is unique in its ability to pick up insight every time it’s used, much like a feedback loop, and that it will improve itself over time. This is, of course, true – except many are forgetting that any system has the capability to do this.
What’s more, AI isn’t the only way that developers can create software solutions. In many cases, developers can simply incorporate logic and monitoring into their applications, rather than using specific technology. It’s quite common for there to be an element of compromise for developers when they set about creating workable solutions, as a great deal of their work is challenging and complex.
And it’s within these challenging and complex areas of the sector that software R&D comes into play.
Unlocking real value
First introduced two decades ago, the Government’s Research and Development (R&D) tax credit incentive rewards innovation and fuels growth. The eligibility criteria for claiming is purposefully broad. Whether companies are creating new products, processes and services or adapting existing ones, R&D tax credits are a valuable source of cash for them to invest in accelerating their R&D, hiring new staff and ultimately growing.
Companies that handle complexity are prime candidates for R&D, and a lot of software development work qualifies for the relief. If you ever wonder about your company’s eligibility for R&D tax credits, my advice is: If you can’t Google it, whether that be in terms of your solution or how you’re trying to build it into your current technology and make it work efficiently, then there’s a good chance that the work you’re doing is R&D.
It’s in the detail – the company’s key developments and failures in the project – that R&D claims hold their merit, and there are plenty of reputable advisers out there that take businesses along every step of the journey.
Successful software
One of the main issues seasoned developers face when building new software isn’t whether the solution is viable or not, it’s trying to understand how it can be created in an efficient, clean, reusable, scalable manner. R&D holds the key because even R&D that’s failed commercially can be included, providing a safe space to test and learn.
We’ve helped many successful clients that have used AI to access R&D incentives, including Zenotech, who develop computational fluid dynamics (CFD) modelling techniques and use AI to predict computing service availability. Then there’s EAI Systems, who use a range of AI techniques to create an internationalised medical diagnostic tool.
However, there are also plenty of companies with non-AI solutions that have benefitted from R&D tax relief too. This includes The Engage Hub – who built a processing engine that consumes millions of data points each day for customer journey insight, and used highly optimal classification techniques that scale instead of AI. We’ve also helped RFIdentikit, which developed an increasingly sophisticated platform for managing trackable tags and badges. The company’s R&D needed to be specific to their sector and as a result, continues to be extensible.
While all of the solutions I’ve mentioned sound quite high-tech, I would still discourage developers from shying away from solutions viewed as ‘low-tech’, such as rules engines or trigger-driven relational databases. These technologies often need to be applied in specific solutions, and can offer value, efficiency, and simplicity when problem-solving. Above all, they can still be eligible for R&D tax credits.
It’s crucial for the sector to collectively dispel the myth that applications, solutions or programmes need to be high-tech and embody the futuristic ‘AI’ image commonly seen in sci-fi films to be innovative – as it simply isn’t true. There’s an abundance of companies developing unique software solutions to address the nation’s digital problems, and there are plenty of untapped opportunities before them.
Now’s the time to seize these opportunities and unlock the value that companies have worked so hard to earn. A word of warning though – just don’t call the solution AI if it isn’t.
