When Software Creation Becomes Everyone's Job

Piramal Finance embarked on a significant transformation, moving from nearly zero non-engineer involvement in shipping software to a widespread adoption of AI across the organization. This initiative transitioned the focus from individual developer productivity to a collective, organization-wide approach to software creation.

Historically, software development was confined to engineering teams, with business needs often managed through spreadsheets and dashboards. The integration of AI tools marked a pivotal shift, enabling various departments to actively participate in live software production, reflecting a profound change in the company's operational model.

Gurleen Kaur noted that if asked two years prior, non-engineer software output would have been "nearly zero," highlighting the dramatic scale of this transition.

The widespread adoption of AI tools has significantly blurred the traditional boundaries separating business, product, engineering, and operations teams within Piramal Finance. AI now functions as a comprehensive software creation layer, extending its utility beyond a mere tool for individual developers.

This allows individuals with innovative ideas to bypass the previous necessity of waiting for an engineer or coder to implement their vision. The direct engagement of non-technical staff in software development accelerates innovation and responsiveness to market demands.

This shift has empowered a broader spectrum of employees to contribute directly to the software development lifecycle, fostering a more collaborative and efficient environment.

This strategic shift redefined the development paradigm from a 'Business-Engineering-Product' model to a 'Business-Platform-Application' model.

Gurleen Kaur emphasized the BYOT initiative as the strongest transition in the company's story, as initial trials confirmed that AI could drastically reduce the lead time for software delivery.

This operating system was developed to manage critical aspects such as taking applications to production, handling Release Candidates (RC), and monitoring for application errors.

Cria OS addressed the secondary challenges that arose from the democratization of software creation, allowing non-engineers to manage the complete software delivery lifecycle. This system ensures that even with widespread building, applications maintain stability and quality through structured deployment and error resolution processes.

This initiative, named after the Indian urgent train ticket system, aims to drastically reduce wait times for new features from months to weeks, with an eventual goal of days.

Keerthivasan Devaraj noted that a different approach was needed to solve the issue of slow delivery, leading to the creation of this new experiment squad. The 'Tatkal' model represents a commitment to agile development and responsiveness, ensuring that critical features reach users much more quickly.

Engineering teams at Piramal Finance faced an insatiable demand for new tools and features, creating a significant bottleneck in software delivery. This challenge spurred a shift in perspective, moving from an emphasis on what engineers could do to exploring how business teams could build the tools they needed directly.

This strategic pivot led to the creation of the BYOT (Build Your Own Tool) initiative, designed as a controlled platform. It empowered non-engineers to develop internal tools while adhering to established standards and guidelines.

This new approach aimed to decentralize development, putting creation capabilities in the hands of those who best understood the requirements and urgency of their own tools.

Gurleen Kaur articulated this change in philosophy by asking, "What if people who need the tool build the tool?", capturing the essence of the BYOT program.

This eliminated common issues like node installations and dependency management, ensuring a consistent development environment for all users.

Gurleen Kaur stated that the first thing they removed was setup, making it easier for everyone to begin building immediately. The platform also automated deployment with a single button, maintaining engineering standards through built-in guardrails, which simplified the transition from development to production.

It aims to unify various tools and workflows, providing a standardized environment for all stages of software development.

This system ensures that every part of the SDLC is managed efficiently, supporting the organization's expanded software creation capabilities and promoting consistent practices across teams.

This approach mandates that no code is written until there is full agreement on the idea, supported by approved requirement documentation and mockups.

This method significantly reduces development costs by ensuring that resources are only expended on validated requirements, thereby preventing rework. The 'shift left' strategy also contributes to higher quality output by embedding early validation into the development process, aligning all stakeholders before coding begins.

This shift means engineers are no longer spending time repeatedly solving the same problems in applications, but instead focus on creating foundational solutions.

This new focus includes developing a 'Piramal skills repository' to standardize AI behavior and integrating MCPs (Managed Control Planes) for real-time monitoring and deployment. Engineers now concentrate on building 'starter kits' and secure defaults, which automate compliance and governance across all software initiatives.

This strategic repositioning allows engineering to address systemic challenges and ensure that all new applications adhere to organizational standards and security protocols efficiently.

The rapid increase in software creation led to a new bottleneck in quality assurance, prompting the development of QAIDLC (Quality as a Software Development Life Cycle). This system automates testing cycles for every application, ensuring that quality checks are not dependent on individual developers remembering to write them.

QAIDLC establishes integration, regression, and smoke tests as mandatory Continuous Integration (CI) gates, rather than optional suggestions. This ensures that every application undergoes rigorous testing automatically, maintaining high quality standards across all deployed software.

This proactive approach embeds quality assurance directly into the development pipeline, preventing defects from reaching production and ensuring consistency.

The role of engineers has transitioned from building individual applications to creating and managing scalable platforms for internal users. Even as AI improves at writing code, complex problems related to core systems, scalability, security, and data persist.

Keerthivasan Devaraj explained that engineering efforts are now directed towards architectural problems where reliability and fundamental system design truly matter. This shift allows engineers to focus on building stable and robust foundations that support the entire organization's software ecosystem.

They tackle challenges that AI cannot solve alone, ensuring the integrity, performance, and security of critical infrastructure.

Applications developed on the platform, such as Crea OS, are only permitted to call services exposed through an approved API catalog, thereby preventing them from executing arbitrary actions.

Gurleen Kaur highlighted that the platform itself prevents technical risks by disallowing unapproved access, rather than relying solely on human review. While the platform handles the coding and technical enforcement, human judgment remains essential for making policy and ethical decisions that guide the AI's operation.

This dual approach ensures that technical safeguards are built-in, while strategic oversight maintains human control over critical business decisions.

Gurleen Kaur emphasized that quality begins with the requirements themselves. This involves defining test cases before any application construction begins and documenting quality expectations through 'shape notes.'

Reviews are conducted continuously throughout the development loop, ensuring that quality is built in at every stage rather than being an afterthought, leading to more robust and reliable software.

Gurleen Kaur emphasized that even the best platform cannot succeed without changing someone's belief system about their capabilities.

The process begins with solving tiny, real-life problems to demonstrate immediate value and build initial success. Through successive small achievements, individuals gain confidence in their ability to use and contribute to software creation.

This approach acknowledges that mindsets cannot be simply installed; they must be nurtured and grown over time through practical experience and positive reinforcement, enabling broader participation in software development.

Keerthivasan Devaraj highlighted that simply choosing an agentic AI software or tool is not enough; comprehensive guardrails must accompany it.

Authentication systems serve as fundamental constraints, ensuring that only authorized users can access and utilize specific tools. Role-Based Access Control (RBAC) further refines this by granting different access levels to various departments, preventing unauthorized tool usage and maintaining data integrity and security.