Few IT executives are as acutely aware of the pace of advances in artificial intelligence (AI) as Manu Narayan. Nearly nine months after taking over as the first chief information officer (CIO) at GitLab, a software development platform with more than $1 billion in revenue and more than 2,000 employees, Narayan is tasked with turning the company into a proving ground for the very technology its customers use.
“The AI field in general is changing so rapidly that we’ve had to constantly rethink our goals and what we want to accomplish,” he said in a recent interview with Computer Weekly.
Working with GitLab’s R&D team on product development, Narayan’s mission is primarily internal: modernizing business application stacks, user support, and data and analytics. But rather than bolting AI into existing workflows, his goal is to rebuild operations from the ground up.
“When we were rethinking our AI strategy a few months ago, the focus wasn’t how to deploy AI,” he said. “The focus was on leveraging AI to reimagine the nature of work within the company: thinking about processes from first principles and using agent AI to drive them.”
Using the example of a customer success manager (CSM), Narayan said that even though the purpose of the role is to build deep relationships with customers, CSMs spend hours on administrative tasks such as creating quarterly business review slides for customers, transcribing notes, and exploring context across customer relationship management systems, data warehouses, and chat channels.
GitLab wants to deploy AI agents to handle the grunt work, freeing up employees to focus on high-level strategy. “We want all of our team members to focus on what matters most: the core purpose of their role,” Narayan said. “We are leveraging AI for tasks that not only increase productivity by 10-15%, but also allow us to scale out in a more linear way.”
To manage AI deployments, GitLab has adopted a hub-and-spoke operating model. A central AI enterprise team handles governance, technology build, and guardrails, and dedicated “AI transformation owners” embedded within each department identify time-consuming, repeatable tasks that are ripe for automation.
This approach is already being applied to GitLab’s own internal employee support network. The company built an AI agent to assist its 120 internal support staff across IT, HR operations, and sales, enabling them to instantly derive the context they need or avoid routine tickets altogether.
Reject “tokenmaxxing”
As AI adoption increases across the enterprise, CIOs will naturally be tasked with managing and measuring costs. However, Narayan is wary of strategies such as “token maxing,” where developers and employees are incentivized to maximize the number of AI tokens they use.
It’s easy to reach 90% of applications developed in-house. The remaining 10% – role-based access control, auditability, and immutable logging – are what you need as a public company or a company with regulated customers, but they’re incredibly complex to build.
Manu Narayan, GitLab
“We have specifically avoided token maxing, and we don’t want to do that,” Narayan said. “While gamification can help drive results, I think it drives the wrong behaviors. We’re not looking for purely context-in, context-out as a measure of success. It’s very hard to know if someone is gaming the system. Are they just sending out too much content because they don’t know what they’re actually doing?”
Instead of tracking token writes, GitLab tracks daily active usage across the technology stack to ensure your employees are building sustainable habits. To calculate a rigorous return on investment (ROI), Narayan advocates anchoring AI deployments to traditional business metrics. For AI agents assisting sales development reps, success is measured by standard key performance indicators (outbound messages, scheduled meetings, sales pipeline conversion) rather than the number of prompts generated.
Build, buy, and the future of SaaS
There are suggestions that the days of off-the-shelf Software-as-a-Service (SaaS) applications are over, as AI lowers the barrier to building in-house tools. Narayan believes this is greatly overstated, especially from a governance and compliance perspective.
“We may see more custom interfaces and separation of interaction and recording systems,” he said. “However, the underlying governance controls for core SaaS tools are not widespread.”
Narayan also points out the hidden costs of bespoke software development, saying, “90% of the applications you develop in-house are easy to reach. The remaining 10% – role-based access control, auditability, immutable logging, etc. – are what you need as a public company or a company with regulated customers, but are incredibly complex to build.”
To ensure security across custom and supplier tools, GitLab has established AI governance based on strict data classification standards. While public data flows through self-service platforms, proprietary or customer data requires a more in-depth security review before interacting with language models at scale.
Despite strong management support and a budget, change management remains a challenge for Narayan. Bridging the gap between AI-forward employees and those slow to adapt requires a combination of departmental centers of excellence and internal AI hackathons.
But the biggest pressure for CIOs is the clock.
“What keeps me up at night is whether we are moving fast enough,” Narayan said. “In the AI era, decisions need to be made in days and weeks rather than months and quarters. But I still worry about whether we are driving the right initiatives that deliver the right long-term ROI.”
