Lean AI SDLC: how CI&T is reinventing agility in Ways of Working

Since 2024, the use of Generative AI in development teams has established itself as a powerful lever for productivity. Code assistants, copilots, and generative tools have begun to support different roles in the development cycle using agile methodologies.

However, even with clear efficiency gains, Artificial Intelligence was used to enhance individual capacity but not necessarily to transform how software was built.

“When we started to observe the impact of AI on teams, we noticed that the processes remained essentially the same. You increased individual performance, but the development flow, with its handovers, waits, and cycles, remained intact,” explains Gilson Gaseorowski, Head of Lean Digital Transformation at CI&T.

This realization led CI&T to cross a new frontier in software development, so that GenAI would not only be a productivity tool but also an element capable of redesigning the very development process.

---

Lean methodology in practice: eliminating flow waste In the traditional model

Each stage of development depends on handovers between specialists (analysts, architects, developers, testers, among others), and much of the total time spent on software construction is consumed not by the execution itself, but by the waits between these stages. Therefore, the opportunity for gain was not only in isolated activities but also in removing the spaces between them. In the Lean view, this is waste.

“When you connect these stages with AI agents, the end of one activity can immediately trigger the next. Instead of waiting for the next sprint or the availability of another specialist, the flow continues almost in real time,” says the executive.

This change marks the beginning of a new logic of software production methodology, where the traditional agile model augmented by AI (in which humans perform tasks with the help of tools) gives way to an agentic model of development, coordinated and orchestrated by humans.

In this new framework, professionals cease to act primarily as task executors and take on roles as specialists, reviewers, and orchestrators who ensure governance, ethics, and strategic alignment of the work automated by Generative AI agents.

Gilson emphasizes that “humans remain in charge, but now as pilots and orchestrators of the process. AI executes, connects stages, and removes the frictions that were once inherent to the working model.”

---

The end of development as we know it: the new foundations of operational efficiency

While in 2025 the mindset pointed towards the use of Generative AI as an individual acceleration aid, CI&T's new Way of Working places Artificial Intelligence at the core, as an intrinsic part of the entire process, defining a new dynamic for technology teams. This gives rise to three main changes.

01. Commoditization of code

According to Luiz Grecco, SVP at CI&T, in the last two years, there has been a noticeable increase in the level of trust that companies and individuals have in Artificial Intelligence, which has shifted from skepticism about safety and hallucinations to a pursuit of its unquestionable potential for results.

“As a consequence, code has become a commodity; that is, writing code is no longer the main technical challenge, and the focus has shifted to the orchestration of the agents involved in the flow as a whole,” states the executive.
The maturity achieved by CI&T has proven the thesis that code production could cease to be an artisanal and creative activity and become a systematic process.

According to Marcos Fernandes, Executive Director at CI&T, this change “is almost like we are experiencing the industrial revolution of software development.”

The maturity achieved by CI&T has proven the thesis that code production could cease to be an artisanal and creative activity and become a systematic process.

According to Marcos Fernandes, Executive Director at CI&T, this change “is almost like we are experiencing the industrial revolution of software development.”

02. Generation of multiple derivatives for selection

As code has become systematic production, the efforts to create a single perfect input give way to four or five versions of the same requirement or functionality created by AI. The human specialist then chooses the best option to advance the flow.

A/B testing also gains efficiency, expanding the capacity for much broader models, such as A, B, C, D, E, and F, allowing diverse variants to be tested simultaneously, at a much lower cost than before.

03. Removal of waits (handovers)

The flow of the agentic development model allows for another structural change: since the entire process is redesigned, it is possible to skip or invert complete stages of the traditional process while still ensuring the consistency of what is delivered.

“The sequential process no longer matters, and even the pauses in the flow cease to be failures and become intentional, with the purpose of reviewing, ensuring ethics, or compliance with standards,” explains Fernando Ostanelli, VP of Digital Solutions at CI&T.

---

A new culture for a new development model

When the way software is produced changes, organizational structures, roles, and even the work culture inevitably change as well.

For years, the software industry has moved towards a high degree of specialization, but in the agentic model, professionals begin to act more generally and are equipped with a vast array of AI tools that allow them to operate in various fronts of the chain.

Another change is the so-called continuous learning - or continuous upskilling. Gilson explains that “deep and lasting learning is replaced by frequent ‘knowledge bites,’ since knowledge has a short shelf life due to rapid technological evolution.”

At the same time, the process ceases to be a rigid path that must be followed regardless of the problem.

Instead of fitting any challenge into the same workflow, teams now have more autonomy to define the best process for each specific context.

“Before, regardless of the problem, we followed the same workflow. Now, teams decide what the most appropriate process is for that type of delivery,” says Marcos.

This change places people in a new role within the work model: less as task executors and more as architects of their own production process.

---

Business Impact: why CI&T's agility generates value?

CI&T's expertise in orchestrating generative AI goes far beyond the adoption of advanced solutions (such as CI&T Flow) and focuses on the real business impact.

By knowing how to combine the best and most specialized available AI technologies, the company ensures a perfect match between technology and the business needs of its clients.

This positioning translates into real value generation in the following ways:

• Budget efficiency and predictability: the partnership shifts from a model based on "allocated hours" to objective metrics of functional scope and value delivery.

• Time-to-market and opportunity windows: clients can test hypotheses and capture revenue streams much more quickly, leveraging the acceleration in building digital products.

• "AI first" competitive advantage: preparing traditional companies to compete with new players that are already operating natively with AI.

• Pollination of Way of Working: CI&T acts not only as a supplier but also as a reference and inspiration for the client to modernize its internal way of working.

“In recent years, CI&T has built a historical capacity to effectively apply AI, transforming technology and methodology into results – and listening to this feedback from our clients,” concludes Marcos.

---