Vaibhav Gujral works at Capgemini and is a 6x Microsoft MVP. He writes about cloud architecture, agentic AI, and technology leadership.
getty
In the past 18 months, the core question in software architecture has shifted from “should we adopt AI tools?” to “how do we rewire our entire development process around them?” Most enterprises are still stuck on the first, but the real transformation lies in the second.
I’ve watched this play out repeatedly. Teams get excited about AI-assisted coding, run a pilot, see productivity gains and then hit a wall—because the underlying operating model was never designed for agents. The tooling changed, but the process didn’t. That mismatch is where most enterprise AI initiatives stall.
What Agentic AI Actually Does To Your SDLC
The most common question I hear from engineering leaders is some version of: "How do we actually use AI agents across our development life cycle, not just for code generation?" It’s the right question, and the honest answer is that agents don’t just accelerate your existing process—they expose every weak point in it.
Take something like cloud transformation. Tools like GitHub Copilot modernization agent or Azure SRE agent can compress work that used to take weeks into days. Legacy code analysis, dependency mapping, migration scaffolding and infrastructure remediation—agents handle the mechanical parts faster than any team can. But here’s what I’ve noticed: The projects that benefit most aren’t the ones with the best AI tools. They’re the ones with the clearest human intent going in. When requirements are fuzzy, agents produce fuzzy output—just at much higher speed and volume. Garbage in, garbage out has never been more true or more expensive.
An AI-native operating model flips the traditional SDLC on its head. Instead of humans authoring everything and tools assisting at the margins, intent becomes the primary human output. Engineers define what needs to happen, agents propose how to do it and humans govern the result. That’s a real shift in where expertise needs to live.
The Job Displacement Argument Gets It Wrong
Here’s where I’ll push back on the narrative that dominates most conversations about AI and the workforce: the fear that IT jobs will simply vanish. I don’t buy it—but I do think the people saying “jobs are safe” are also missing something important.
The more accurate picture is that current job definitions are becoming irrelevant, and new ones are emerging that don’t yet have names. Someone has to decide what the agent builds. Someone has to catch what it gets wrong. Someone has to translate between what a business stakeholder wants and what an AI system can faithfully execute. These are human jobs. They require judgment, context and accountability that agents fundamentally can’t carry.
What’s actually disappearing is the entry-level role as we’ve known it—the junior developer whose job was to implement well-scoped tickets, write boilerplate code and learn the craft through repetition. Agents do that work now. Most senior engineers today got good by doing a lot of unglamorous work first. That opportunity is disappearing.
The Training Problem Nobody Is Planning For
That’s what worries me more than the automation itself. Senior engineering judgment—the ability to spot a bad architecture, understand why a system will fail under load, know when a business rule has been encoded incorrectly—has always been built through years of junior-level struggle. You learned what good code looked like by first producing a lot of bad code and having someone senior tell you why.
This isn’t just my observation. A February 2026 paper titled "Redefining the Software Engineering Profession for AI" in Communications of the ACM makes the same case—agentic AI is creating an “AI boost” for senior engineers while imposing an “AI drag” on early-career developers who lack the judgment to steer it. The path most of us took to get good at this craft is closing, and the industry hasn’t figured out what replaces it.
If agents absorb the foundational work, we lose the natural pipeline that produces senior engineers. The next generation will enter the field directing AI systems without ever having built anything from scratch themselves. They might be highly capable with the tools and still lack the instinct to govern them well. That’s a dangerous gap to carry into production systems.
Enterprises need to get ahead of this deliberately. Structured mentorship programs, architectural simulation exercises, a deliberate code-review culture and exposure to legacy systems and to why they’re built the way they are—these become essential, not optional. The industry can’t assume that engineering judgment will develop organically as it used to. We need to design for it.
What Leaders Should Actually Do
If you’re a technology leader navigating this shift, focus on three core actions:
First, invest in clearly defining intent. High-quality requirements and architectural decisions are foundational and set the ceiling on what your agents can accomplish.
Second, build a structured accountability layer: Specify who reviews agent output, identify when human intervention is required, and establish how AI-generated outcomes trace back to business needs. Treat these as engineering imperatives, not merely policy decisions.
Third, start thinking now about how you grow senior engineers in a world where the traditional path is closing. The teams that figure that out early will have an advantage that compounds over time.
The shift happening in software development is real, and it’s moving faster than most road maps account for. The organizations that will come out ahead aren’t necessarily the ones moving fastest. They’re the ones building the human infrastructure—the judgment, the governance, the deliberate training—to match the machines' pace.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?