Talk of agentic AI and its possibilities are everywhere at the moment. Forrester has called it “the most exciting development” on its list of Top 10 Emerging Technologies For 2024. PwC says the “central question isn’t whether to adopt this technology, but how swiftly organizations can integrate it to stay ahead of the competition.”
As we look towards the future, this blog looks at what agentic AI is and how it works. We also look at where agentic AI opportunities lie, with a particular focus on the software development lifecycle. We finish by considering what you need to think about now so you’re ready to embrace agentic AI’s possibilities.
It aims to give you the primer you need to start thinking about its possibilities for your teams.
What is Agentic AI and How Does it Work?
We can think of agentic AI as the third generation of AI (after predictive AI and generative AI).
To help understand what agentic AI is and what its potential is, it helps to understand what we mean by ‘agentic’. It’s someone (or something) that has agency – the ability to self-organize, be proactive, be self-reflective, and be self-regulating.
So, unlike generative AI, agentic AI doesn’t rely on human prompts before making a move. Instead, it’s autonomous. It can take data inputs, make decisions based on those inputs, execute its decisions, and learn from the outputs.
It works in four looped stages. The first stage is to gather data from all the inputs available to it, including inputs from other agentic AIs. Then, using that data and a large language model, it formulates a plan for what’s needed. It then executes the plan. The final step is to reflect on the process and use its new knowledge as one of the inputs as it starts the first stage again.
For those worried about the possibilities of negative consequences, it’s important to know that guardrails can be put in place to limit the power and autonomy of agentic AIs. It is also important to know that agentic AIs are orchestrated and managed by humans.
And while there are undoubtedly downsides to consider, the upside possibilities of agentic AIs are hugely exciting, not least in the software development lifecycle.
From Generative AI to Agentic AI in Software Development
To understand the practical potential of agentic AI, let’s consider how it could be used in software development.
Software developers have already been quick to embrace what generative AI has to offer – 46% of software engineers already use it, and 80% think its ability to automate simpler, more repetitive tasks will free them to focus on higher-value-adding tasks.
Further, a recent study found programmers using AI could code 126% more projects per week, which opens up huge possibilities for the future productivity of software development teams.
It is only getting better too – in software test automation, for example, tools such as Eggplant GAI are giving developers testing super powers.
So, if software developers are already reaping the rewards of generative AI, imagine the potential of agentic AI, where they can be even more hands-off.
In the world of generative AI, developers still need to author code. And while generative AI will find script errors, developers still need to fix them.
But agentic AI has agency, so it can author, fix, and evolve code without being told. In the analysis and planning phase of the software development lifecycle it can undertake business story modeling, requirements engineering, create executable specifications, and more. In the coding phase it can generate code, translate code, and more. In the testing phase it can design, generate, execute, analyze, report, maintain, offer insights, and more.
Plus, in the software development lifecycle, as in all business processes, the benefit of agentic AI is its ability to run multiple workflows at scale. It’s the equivalent of a near-infinite team of junior developers that gives you an unparalleled ability to expand your reach and coverage while simultaneously turbo-charging productivity. It’s why, in its look at agentic AI as a top strategic technology trend for 2025, Gartner called it “an extension of the workforce that doesn’t need vacations or other benefits.”
The Importance of the Human in the Loop
Of course, when we understand agentic AI’s ability to work autonomously, the most obvious question is: will we still need human developers?
The simple answer is yes. The human in the loop is still essential. However, their role changes. Instead of doing, they’re steering and adding uniquely human insights at every stage. And because agentic AI can do so much more of the day-to-day junior work, experienced developers have the time to focus on deeper-level improvements, whether driving technological advances or using their uniquely human abilities to improve user interfaces and enhance the user experience.
Ultimately, the combination of human and agentic AI has the potential to deliver enhanced efficiency, enhanced user experience, and enhanced quality. In other words, it helps you achieve the holy grail of better, faster, and cheaper.
Preparing Your Organization for Agentic AI
The rewards of agentic AI are clear to see. But, as always, the biggest rewards go to those who properly prepare. So where should you be focusing your efforts?
Assessing Your Data
The modern enterprise accumulates vast amounts of data in many formats and many places.
Some data points conflict with others. The problems with this are already emerging with generative AI. For example, when one telecoms firm tried to train a call-center generative AI assistant, it found 37 standard operating procedures.
It is also true that some data is better quality than others. As always, garbage in equals garbage out.
While these issues are always a problem in every organization, no matter how advanced its use of technology, they move to the next level when you add agentic AI to the mix. It therefore pays to consider undertaking a robust data audit and quality assurance process.
Assessing Your System Quality and Security
A perennial issue with new technology is the way it interfaces with legacy technology. And as always when introducing new systems, weaknesses in existing tech will likely come to the fore. An inability to cope with the volume of information requests or the opening up of security vulnerabilities will impact on your ability to realize the opportunities of agentic AI.
How well do you understand your organization’s tech stack preparedness for the addition of agentic AI systems?
Assessing Your Workforce Strategy
The widely held assumption is that AI will reduce headcount as it takes on more of the tasks that humans have previously fulfilled. But McKinsey believes top-growing occupations between now and 2030 include software developers, computer systems analysts, and data scientists, often due to the deployment of automation systems themselves.
But there will undoubtedly be considerable workforce shifts. We’ve seen how agentic AI could fulfil the work of junior developers. It also seems clear that the more experienced and senior developers will be the ones who will bring the most valuable human-led insights that agentic AI requires to achieve its maximum potential.
But without a steady inflow of junior developers, how will your organization develop the next generation of highly skilled and experienced developers? And what is true in software development also holds true for the junior and senior roles across an entire organization.
Get Ready for the Future
Agentic AI offers huge potential, even more so than generative AI. It is therefore vital to consider how and where it could add value in your organization. It is also vital to move quickly. Generative AI was hailed as the future of business in November 2022 with the launch of ChatGPT 3.5. Just two years later, the future of business is already shifting to agentic AI. Standing still, or even waiting to see, is not an option with such seismic shifts taking place so quickly.