An AI Assistant For The Entire Software Development Life Cycle? It’s Here

We’ve all heard accounts of the ways that generative artificial intelligence is transforming the software business. Tools like Bard, Google’s new chatbot, can write and debug code, while OpenAI’s Codex can generate working code in more than a dozen programming languages — all of which represents an impressive leap forward.?

But the reality is that coding is just the tip of the iceberg. For any company that makes software, uses of generative AI can increasingly span the entire development life cycle, from early testing to production deployment.?

That’s why the Harness team and I are excited to announce AIDA (our AI Development Assistant), a game changer for software delivery. We’re expanding the influence of AI beyond coding to all facets of the software development life cycle, which spans building and testing code, addressing security concerns, ensuring reliability, deploying changes and optimizing costs. We’re rolling out several groundbreaking features now, with more coming in the months ahead. (To learn more, check out our new AIDA page.)?

Imagine AI not just speeding up coding with autocomplete but creating test cases automatically, while swiftly debugging and fixing code. On the governance and security side, AI can explain and fix vulnerabilities. It can also craft cloud asset policies, ensuring prudent cost management and a secure environment.

The implications go well beyond software teams. In an era of cost-cutting, generative AI can help businesses achieve more with the resources they have — important in a world where engineering talent remains scarce. In the bigger picture, AI also promises to speed up the pace of innovation, letting companies rapidly add more services and features that consumers crave.?

As the founder of multiple software platforms used by developers, I can attest to the power of AI and automation to cut costs, make teams more productive, and turbocharge innovation. But incorporating generative AI comes with important considerations. Here's how to get the most out of it — and what to watch for.

The software development life cycle … in the AI era

It’s no secret that software development remains highly inefficient. Most engineers spend half or less of their working hours writing code, according to one estimate, with the rest consumed by tasks like testing, as well as waiting for a product to move to the next development stage.

Introducing generative AI tools can reduce the repetitive drudgery that plagues the industry, while also boosting efficiency — not just for engineers but for other key players in the software development life cycle.?

Let’s say a large financial institution wants to add a feature to its mobile banking app for transferring different currencies between accounts. Consumers are always clamoring for more features, leaving banks that fall behind in the app race at risk of losing clients to competitors.

AI can prove a powerful ally in each step of the software development life cycle:

• Creating a spec: Development typically starts with the project manager writing a spec, or brief, for what will be built, based on existing features. Doing so for the bank’s new currency feature could easily take a week. But generative AI can assume the lion’s share of the work — writing the document based on, say, the specs of the previous 50 features and yielding better results in less time.
• Writing the code: This is the area where most industry attention has focused, with tools like Github’s Copilot attracting wide attention. Using autocomplete, AI anticipates developers’ coding needs, offering suggestions in real time to speed up the task. AI automates the debugging process too, proposing solutions and even fixing problems automatically. Generative AI–powered programming helpers can now autocomplete up to 60% of the code, leaving it for the developer to edit. According to one survey, this approach is as much as 55% percent faster, with 75% of respondents saying they felt less frustrated while coding and could focus on more satisfying work.
• Building and testing: During this stage, AI can identify potential errors in the code even before a build starts. If the build goes on to fail, it compares the problem with known issues, suggesting possible solutions. AI’s role in build and deployment troubleshooting also includes generating code fixes based on what it’s gleaned from previous successful builds. For developers, who typically spend 20% to 30% of their time troubleshooting, swiftly resolving these problems means higher productivity.
• Securing the code: AI not only identifies security vulnerabilities in the code but also explains them and suggests fixes. For example, it can offer recommendations for resolving Common Vulnerabilities and Exposures (CVEs) and Common Weakness Enumerations (CWEs), two widely recognized categories of vulnerability.?
• Deployment: Next, the baton is passed to a DevOps engineer, who has to ensure proper deployment of the app feature to end users, including clearing all-important security controls and governance, risk and compliance (GRC) hurdles.?For a bank developing a finance app, this might embrace everything from considerations around cyberattack prevention and protection of cloud data to adherence to laws and industry standards. Importantly, these controls, policies, and checks and balances can all be supported with AI.
• Managing costs: Without proper management, cloud costs can easily spin out of control. In fact, unnecessary expenses will account for almost an estimated 30%, or $180 billion, of the almost $600 billion spent on cloud services worldwide this year. Using natural language queries, AI makes cloud policy easier by automating the creation of new policies and helping people understand existing ones. This approach not only speeds up the process, resulting in cost savings, but also democratizes policy by enabling anyone to become an expert. At the same time, it makes cloud governance and cost management more effective.

New tools can manage idle cloud time, provide customizable dashboards that make costs easy to grasp, and even integrate cost management into the development pipeline.

Platforms are evolving fast, but I would estimate that integrating generative AI into the software delivery life cycle can already improve efficiency by 30% or more. Considering that IT spending is expected to reach $4.6 trillion, the potential savings are staggering.??

Sold on AI for development? Read the fine print before diving in

Of course, using generative AI for software development comes with risks and caveats.

First, all AI must be trained on existing code. But leaning on this code may run afoul of copyright restrictions. The maker of GitHub Copilot is currently being sued, along with parent company Microsoft Corp. and OpenAI, for allegedly using copyrighted code. (The easy fix here: use generative AI that only trains on code with a permissive software license, which allows people to use it in other places.)?

Using generative AI for code review also has its pitfalls. Code shared with an external AI platform for review might not be protected or kept private. Just ask Samsung, which banned staff use of generative AI tools after employees accidently uploaded sensitive code to ChatGPT.??

That said, it’s unrealistic for most companies to try to develop AI capabilities “in-house.” Although going that route might give more control over customization and integration, the process is prohibitively expensive and could easily drag on for years.?

For a bank creating an app, or most any company seeking to integrate AI into the development process, the prudent course is to rely on an established outside platform. By choosing a vendor with a long track record, one that trains its models on permissive code, and one that safeguards proprietary data, companies can easily avoid these pitfalls. (Spoiler alert: Harness’ AIDA ticks all those boxes.)?

For developers, AI is a helper, not a job killer

AI and automation are often mentioned in the same breath as job loss. But in this instance, AI tools won’t be stealing developers’ jobs any time soon. In fact, in an industry that’s crying out for reinforcements, they can make software engineers more productive and efficient.?

That’s a big plus given that despite widespread layoffs in the tech sector, programming talent remains relatively scarce. By 2026, the US will be short some 1.2 million engineers and developers, the Department of Labor projects. Worldwide, the shortage could reach an eye-popping 85.2 million as of 2030, more than double the alarming deficit of 40 million for 2021.

The upside of using generative AI technology from conception to production deployment is huge. Every software engineering organization has a backlog of hundreds — if not thousands — of features its customers want. By helping clear the decks, generative AI can accelerate innovation — freeing up developers to tackle the next wave of tech challenges to come our way.

Harness’ AIDA is a groundbreaking first step toward this reality — a future where AI is infused into the entire software development life cycle. As a developer myself, I’m excited about the real potential to empower engineers and their leaders in new ways, creating a game-changing future for software development.

Thank you for reading! I'd like to know your thoughts in the comments below. For more insights from my experience as a serial entrepreneur and how we can harness the power of software to change the world, be sure to?subscribe?to Entrepreneurship and Leadership.