Software development appears poised for a paradigm shift, with low-code tools promising to put the power of programming and into the hands of non-technical staff. But with low code still in its infancy, better tools are needed to equip this army of so-called “citizen developers” with the skills they need to build apps and systems that add value to their organisations. AI-powered solutions could hold the key to further growth of the low-code movement.

low code and AI
Low code helps staff without deep programming knowledge build applications. (Photo by Best-backgrounds/Shutterstock)

Analysts are predicting big things for low code. Gartner expects the low-code market to grow 22.6% in 2021, to a value of $13.8bn. “While low-code application development is not new, a confluence of digital disruptions [and] hyperautomation has led to an influx of tools and rising demand,” said Fabrizio Biscotti, research vice president at Gartner.

Low code is proving popular with developers, with Forester Research predicting that 75% of software development teams will have deployed low-code platforms by the end of the year. For these professional coders, low-code platforms are ready to use as a time-saving tool. But for enterprise organisations, the bigger prize on offer is equipping non-developers with the ability to create digital systems. Realising this potential may be more complex.

What is low code and who is using it?

As the name suggests, low-code platforms allow users to build software applications and tools without having to write long and complex code scripts. “Low code is something that lets you move fast and build things, often without coding, but which will require some technical knowledge,” says Nile Frater, who runs NoCode.tech, a platform for novice programmers. “So you might be able to drag and drop an interface together, but you’d still have to write some code underneath that to make it all work.”

Low code is being used across a range of industries and is particularly popular with financial services companies. “Insurance companies, banks and other financial institutions are grappling with large estates of technology and looking at ways to innovate,” says Nick Ford from Mendix, one of the largest low-code platforms. “More interactions with customers are happening remotely, and Covid-19 had driven a new narrative where technology and software are the bridge between people. A couple of years ago low code wouldn’t be on the tip of anyone’s tongue, but now it is.”

The approach is not without its risks. “You’re giving tools to potentially non-technical people to be able to make experiences happen for customers,” says Stef Lewandowski, a software developer and long-time low-code advocate who oversees low code infrastructure at On Deck, an online education platform. “Anything could go wrong in there, of course, so I think there has to be a fair degree of trust both ways.”

The upside is that teams using low code can move more quickly to improve processes and test new functions, Lewandowski says. “The velocity we’re able to maintain as a team is higher than you could with a centralised product function,” he says. “I think that’s one of the main selling points of the approach, and why other teams are going to start adopting it. It means your development team isn’t a bottleneck – if you observe something that a particular customer or user type needs, you can develop a solution without having to schedule work with the product team and fight for resources against everyone else.”

Low code and AI: the role of automation

Despite its promise, using low code is still not yet a reality for most businesses, with only 41% of companies having active “citizen developer” programmes, according to Gartner.

Rob ten Kroode runs Low Code Plaza, a community and consultancy business helping companies with low code deployments. The spread of low code has been “slower than I expected”, he says. This is perhaps due in part to the hype that surrounds the technology. “These days everybody says they want to get involved [in low code] and talk about doing these big programmes, but when it comes to the actual adoption it all slows down,” says Ten Kroode.

For it to become more accessible, “more intelligence needs to be built into low-code tooling,” he adds.

This is where AI can play a role. AI assistants for programmers have been emerging in other markets, and this week GitHub, the ubiquitous code repository owned by Microsoft, announced GitHub copilot, an AI for users of the Visual Studio Code programming platform. The system monitors a developers’ code as it is being written and makes suggestions for appropriate lines or functions which the user can add.

This intelligence will be even more important in the growth of low code, Ten Kroode says. “You want people who are not technical to be able to say ‘I have different pieces of information that I need to store’ and then for the system to be able to figure out how all that data fits together,” he says. “Most of the systems out there assume you can create your own data model, and I would expect this issue to have been addressed by now, so that we would have systems that make suggestions or come up with a skeleton of a database model. This would make citizen development go so much faster.”

Such systems are missing because low code platform companies have been focusing on the needs of professional programmers, rather than citizen developers, Ten Kroode says. “Experienced developers are interested in low code because their managers are asking them to do things faster,” he explains. “Their needs are being addressed by a lot of the low-code platforms [which] promise to help companies work more efficiently.

“At the moment the needs of citizen developers are being served by smaller companies in the market, and for them, these are difficult technological challenges that take time to solve.”

Low code, robotic process automation and connecting APIs

Lewandowski believes that there is also currently a “missing class of tools” aimed at connecting different low-code systems, and AI is likely to end up playing a role in solving this problem. “Within our organisation, we create thousands of automation rules, and it becomes hard to keep track of what’s going on, which rules apply to which services and who is in control of them,” he says. “Tools are going to emerge to help with this and they are likely to have an AI component.”

He says low code and robotic process automation (RPA) could be a potent combination for this task. “There are a bunch of RPA tools which have done very well in the last few years in plugging into legacy systems and automating repetitive tasks,” he says. “I suspect there’s an application for RPA on top of these low-code systems that will emerge.”

Marrying low-code systems and RPA may be particularly necessary for large organisations wanting to plug newly developed applications or functions into existing systems, Lewandowski says. “For enterprises who want to adopt low code, there are bound to be orchestration and monitoring and maintenance tools which need to be plugged in,” he adds.

Frater agrees that automatically facilitating connections between different databases will be key to low code’s future, and says being able to synchronise APIs that provide information from different systems will be vital for citizen developers. “If you look at the consumer side of IT there’s now an API for pretty much everything,” he says. “The interesting thing for low code over the next few years will be how that comes to the enterprise environment. At the moment it’s still quite hard to get those APIs to connect into whatever low-code platform you are using, but the more those connections can be made the more you’ll see low-code systems transform from something that isn’t that useful to something