Over the last 12 months, the information technology industry has made strides in applying artificial intelligence to software development work.

Some vendor models have demonstrated the ability to translate natural language instructions into program code. Other tools have used AI to automatically generate test scripts and low-code development platform providers are incorporating AI into their development tools as well.

Emerging technology for AI-augmented software development has the potential to combine with scripted automation techniques to automate up to 70% of the work currently done manually by software engineers.

Figure 1 (below) shows the potential. The seven bars indicate where effort is typically expended in software development today, and where machine learning could have an impact in the near future.

AI augmented software development tools will bring huge changes to the way software development must be managed. There are three approaches to ride the change, depending on whether you employ software engineers, sell their services or buy software development services:

Employing software engineers

In 2020, there were more than 25 million specialist software developers, according to some estimates. Some developers are innovative, highly skilled and productive. At the other end of the scale are “cargo cult” developers who rely on Googling for useful code fragments that they can paste into their work. AI will automate the work of the less capable developers first.

To prepare for AI-augmented software development, managers of software engineers might:

• Identify the developers whose work can most easily be automated, and then either retrain them to higher-value work or reassign them to different roles.
• Develop new career paths for junior developers, because the lower-skilled work that has in the past been assigned to juniors will increasingly be automated.
• Consider decreasing the size of development teams. For example, a scrum team of 10 that has included three developers and two testers may, with generative AI, be able to produce the same work with only six people.

Selling software engineering services

According to Gartner, the market for software development and application maintenance services is worth more than $300 billion, with many service providers currently growing at double-digit rates.   

Seventy percent of software development engagements are priced based on the effort needed to complete the job, so there is an opportunity to increase win rates if AI can be used to deliver a project at a lower cost than the competition. Conversely, there is the risk that an hour saved through automation is an hour less revenue that can be billed, making it harder for consulting and outsourcing firms to grow.   

Providers must also recognize that writing code is becoming a much less important part of software development and application maintenance as businesses look to assemble “composable” applications, or business systems from prewritten packaged business capabilities. The more composing is done, the less coding is needed.

Consulting services that focus on solving business problems have a much higher value than technology skills. After all, contextual domain knowledge of an industry or business process is not typically vulnerable to automation since its success hinges on such to trail machine learning models.

Managers of software development service businesses should prepare for AI-augmented software development by:

• Using “bench” developers awaiting their next billable assignment to quantify the effort savings from AI automation tools. 
• Using the results to update the cost-estimating tools used in proposal writing, to ensure that the project prices offered for new engagements are as competitive as possible.
• Redesigning the service delivery organization, evolving away from providing teams of designers, architects, developers and testers. 

Buying software engineering services

Firms that outsource their software engineering to an external service provider are also affected by the disruptions of AI-augmented software.

In the past, software engineering services have been dominated by using software engineers in nearshore or offshore locations to optimize costs. More recently, software engineering firms have evolved to provide entire scrum teams of highly trained developers. Neither type of supplier can be guaranteed to embrace the potential in automating their professional services.

Buyers of software engineering services must take the lead, making clear to their suppliers an intention to automate the development life cycle beyond the established scripted automation of test and CI/CD deployment toolchains.

Although there will remain a market for traditional custom software development services for building unique customer experiences, unique operational processes or unique product features, generative AI will play a larger and larger role in delivering these services.

Software engineering managers who use outsourcers should prepare for AI-augmented software development by:

• Requiring incumbent software engineering suppliers to propose AI-augmented automation innovations at least every quarter as part of a continuous improvement process.
• At the end of a contract when services are renewed or rebid, include questions in requests for proposals about the supplier’s automated development plans. 
• Revising the Quality Engineering plan and contractual Definition of Done to tighten up the criteria for AI-generated code to be accepted.

Neil Barton is a vice president analyst at Gartner Inc., covering the custom application development and maintenance services market. He wrote this article for SiliconANGLE. Barton and his analyst colleagues at the Gartner Application Innovation & Business Solutions Summit taking place virtually in the Americas May 26-27.

Image: geralt/Pixabay