The Impact of Generative AI on the IT Services Lifecycle

The Hype

The launch of OpenAI’s ChatGPT in 2022 catapulted GenAI into the spotlight, breathing new life into AI adoption, which had been lagging other technologies. Technologists and domain experts quickly began exploring its potential and, according to a Gartner report, the hype around GenAI reached its peak in 2023.

However, in reality,

· GenAI is still in its infancy and will see exponential (~10X) growth in the coming 3-5 years fuelled by the ability to process different data types like text, images, audio and videos.

· The underlying dynamic nature of GenAI technology has never been experienced before. The very fact that it can evolve itself stands out compared to every other earlier form of technology evolution, which was static once released.

· These factors are still in the process of being appreciated by enterprises, and the true unlocking of value is possibly still 3-5 years away.

There are a few areas where Gen AI is set to impact (personalisation, predictability and productivity, as quoted by HFS) but I believe that in the short and medium term, most impact will be on personalisation and productivity movements.

Fully realising the benefits of predictability is still 2 to 3 years away. This is because the technologies, as they stand today, need to mature to process data from disparate sources and provide meaningful output.

IT Services Industry

A similar buildup has been created around the possibilities of GenAI automating almost every task in the IT Services industry, many of which are based in India. The Indian IT services industry, until recently, flourished primarily due to the cost arbitrage achieved by moving work to offshore.

This story has changed due to Covid-19. Remote work has brought the possibility of cutting costs even further. This key advantage of cost arbitrage will be questioned once GenAI technologies become the primary means to develop and maintain software.

CFOs and CIOs love the term “productivity improvements,” which essentially means realising the true power of GenAI. This is not just about achieving 10% to 35% productivity gains in certain phases of IT lifecycle, but about increasing the efficiency of developers from anywhere between 2x to 5x in the entire IT lifecycle stages of development, maintenance, and operations.

However, there are challenges that need to be addressed before enterprises can reap the true benefits of GenAI. While there are numerous products that predominantly act as wrappers around LLMs (like Copilot for GitHub from Microsoft, AWS Code Whisperer, Cody), they do not address all stages of the IT lifecycle. It is a journey starting from the very basics of writing simple applications and progressing towards addressing more complex and real-world scenarios, including supporting frameworks, other platforms, and complex integrations within an enterprise.

Where are we today

GenAI tools come at a cost. Current GenAI technologies are designed to perform specialised tasks very well but are isolated in their functionality and do not integrate easily with other systems or areas. For instance, SAP provides tools specifically for ABAP development but restricts the use of many third-party tools within its ecosystem, limiting integration with external technologies.

Similarly, tools like Microsoft Copilot are good for custom code development but struggle to understand and work with the complex frameworks typically used by large organizations.

Although there are numerous tools available that excel in specific areas, using them all can be expensive and complicated to manage and justify a ROI.

A sample list of our analysis of various tech stacks, tools, and benefits is in the table below and the verdict is clear. The technology today is useful for building new applications and supporting code with simple logic and limited scenarios.

However, many of these tasks could previously be accomplished even using Google search and developer assistance websites like Stack overflow. The difference is these GenAI tools run within enterprise's ecosystem, ensuring your code does not leave the enterprise boundaries. Also, the fact that these tools integrate well with the Integrated Development Environment (IDE) of the developer, makes them more user-friendly and effective.

*Scenarios are limited and for simple logic as of today

The journey forward:

If we plot the maturity of using GenAI into IT lifecycle stages and how an organization can navigate through the same, it will look something like below:

Today, tools and technologies are not compatible with all levels and scenarios within the IT process. However, enterprises who are prudent in selecting the right tools are seeing some advantages that justify the expense of these tools. Despite this, these benefits are not significant enough to provide a positive return on investment (ROI) for the overall business.

So where is the reduction in workforce in IT services industry coming from?

The IT services industry is cutting the flab they have been maintaining over the years. Called the “bench,” where individuals in the company do not have billable work from a client, it was maintained to be ready due to continuous and fast paced growth seen in the industry, to quickly start new projects without waiting for weeks or months to hire new talent.

As the growth rate in the industry tapers down, the “bench” has reduced over time from about 28% to about 15% of the workforce.

The end game:

The true potential of GenAI in the IT services industry will be realized by using “Agentic” workflows, where multiple GenAI tools (or agents) work together to achieve the desired results.

Agents will work in their areas of specialization (such as code generation and synthetic data generation), review code generated by other agents (against enterprise-defined standards and optimized to the framework within an enterprise), and stitch it all together for final application delivery.

We will achieve true nirvana when business analysts discuss business workflows, which the GenAI agents will assimilate to generate the necessary application code. The agentic workflow will review, fine-tune, test, and deploy this code, as well as manage operations. In this ideal situation, the entire IT team would consist only of Business Analysts and SREs. The SDLC workflow in future would look something as the diagram below, with agents in action and a minimal team of Dev and Ops (SRE) engineers supporting the same. 

By Ashish Agarwal, Vice President at ITC Infotech