From hype to horizon: A CTO's roadmap for AI-powered software delivery

There's no shortage of voices and headlines telling you what AI can do for software development. At CGI, we also have a perspective on this, built from our own adoption and client experiences, which you can explore through our From AI to ROI podcast series and our enterprise software delivery hub.

What I have often seen missing is a practical roadmap and horizon view that includes not only what to prioritize when it comes to AI, but when to do so. What’s needed is a roadmap that factors in not only the capabilities of the technology but also the people and organizational side.

In my conversations with clients and CGI practitioners  across the globe, the same patterns keep emerging. What follows is a synthesis of what separates organizations building durable value from those still chasing quick wins.

The trap of easy answers: the “red herring” of efficiency

The marketplace is flooded with simple narratives: buy this copilot, deploy that agent, translate everything into productivity percentages. Many of the gains are real. Within CGI, our fully ramped-up programs have shown efficiency improvements of 30 to 40 percent  across the software life cycle. But the efficiency story, alone, can be misleading.

Research based on 44.97 million lines of code found that AI coding assistants produced output patterns statistically like "rogue developers,” with the AI assistants demonstrating “distinctive activity patterns and quality metrics correlated with historically problematic development behaviors.”  ^[1] If your organization measures AI's impact purely through speed and volume, you may be accumulating hidden liabilities in quality, security, and technical debt.

Julie Godin, Executive Chair of the Board of Directors, and François Boulanger, President and CEO, go more deeply into this topic in their point of view, Why the digital puzzle can’t solve itself. You’ll also find throughout this blog quotes from experts across CGI that expound on the topic.

"Even in legacy environments, teams can code better, faster, and more sustainably if they rely on strong engineering practices. Without them, we've seen AI have the opposite effect, amplifying existing challenges and creating new risks." — Kevin Beaugrand, Director, Consulting Expert, AI Technologies Lead, CGI Western & Southern Europe

Three horizons of AI maturity

AI within the software development life cycle isn’t just about coding efficiency; it’s also about reimagining how we can deploy value across the entire organization. Here are some recommendations to consider over the next five years:

Horizon 1: Build the foundation (Now – 12 months)

Most organizations are at this stage. The immediate priority shouldn’t be AI tooling only; it should be on building strong foundations, including an honest inventory of your digital estate, AI governance with executive ownership, and AI literacy beyond just for developers.

This is also the moment to resist premature standardization. The landscape shifts fast; organizations that design for modularity are adapting the fastest.

"AI doesn't fail because the technology isn't ready, it fails because organizations aren't. True differentiation comes from the maturity to govern, educate, and continuously evolve around AI." — Luuk Filart, Director, Consulting Expert, CGI SmartLab Founder, CGI’s Scandinavia, Northwest and Central-East Europe Operations

Legacy applications are also often a focus here. According to our 2025 CGI Voice of Our Clients research, 41% of executives say legacy systems pose a significant challenge to implementing their digital strategy.  

In nearly every client conversation I have, there's at least one legacy application where the original developers are long gone, documentation is sparse, and critical logic is buried in the code. AI now makes it possible to recover that knowledge and create a foundation for modernization. You can't chart a modernization path if you don't know your starting point.

"We needed to upgrade components on an outdated framework with no documentation— like trying to put a 2025 engine into a 1990 car. Leveraging AI with our own expertise, we made the upgrade where it had been nearly impossible before." — Jan Rauchfuss, Director, Consulting Expert, CGI’s Germany Operations

Horizon 2: From assistance to orchestration (12 – 36 months)

In this stage, AI moves from helping individuals work faster to orchestrating delivery processes across previously siloed phases, and the ambition expands.

"Leading organizations are using AI to build the right thing—not just to build the thing right." — John Davis, Vice President, Consulting Expert, AI Software Engineering, CGI’s UK & Australia Operations

The organizations that pull ahead won't be the ones that deploy the most agents; they'll be the ones that can connect strategy, architecture, and execution into a continuous, intelligence-driven flow.

"AI is no longer just assisting development, it is becoming an intelligent orchestration engine. When you connect business architecture, IT architecture, and implementation through AI, you fundamentally change what's possible." — Guillaume Brincin, Director, Consulting Expert, CGI’s Canada Operations

Measurement must evolve, too. Developer productivity and sprint velocity become insufficient. Executive teams need frameworks connecting delivery performance to business outcomes. (Read more here: The AI value paradox: Measuring what matters (not just what’s easy).)

"The future isn't one AI assistant per engineer. It's orchestrating federations of agents that plan, build, and test together. Success won't come to teams that automate the most, but to those that measure which agents improve velocity, quality, and reliability." — Steve Zemanick, Director, Consulting Expert in AI, CGI’s U.S. Commercial & State Government Operations

The proof points start to emerge here, not only in efficiency but in other measures also. In one CGI engagement, this approach reduced delivery timelines by nearly 30 percent while improving defect detection.

"Delivering for a leading Japanese bank proved what disciplined AI-assisted delivery can achieve. We reimagined the life cycle to deliver over three months faster—not by cutting corners, but by engineering reliability into speed." — Raghav Kumar P, Vice President, CGI Asia Pacific Global Deliver Centers of Excellence

Horizon 3: Adaptive, AI-native enterprises (2 – 5+ years)

This is where I get genuinely excited. The first two horizons are about doing what we do today—better and faster. Horizon 3 is about doing things that weren't previously possible. AI becomes ubiquitous and nearly invisible, much as cloud computing is today. The distinction between "building software" and "running the business" blurs as feedback loops connect customer behavior directly to product evolution. The delivery model itself becomes the differentiator: not which tools you use, but whether your approach compounds value across engagements.

"AI-powered modernization isn't about replacing your technology stack overnight, but about having a delivery model built on hyper-reusable component architectures where every solution strengthens the next. When each engagement compounds rather than starts from scratch, you don't just reduce the cost of transformation, you fundamentally change the economics of creating new digital services and products." — Jussi Mäkinen, Head of the Innovation Center of Excellence in CGI’s Finland, Poland & Baltics Operations

The honest timeline

Most organizations remain in the early stages. It will likely take two to five years before practices like automated evaluation, platform-native AI, and flexible governance become more widely adopted. That timeline is longer than many might expect, but this also means organizations building foundations now will compound their advantage. AI coupled with other technologies, like quantum, accelerate those possibilities even more.

"Find peace amidst the chaos. The hype of the latest and greatest isn't where to apply your strategy. Implement solutions appropriate for your maturity today. Stay modular, stay nimble, and you'll capture incremental ROI while minimizing tech debt. And keep your eye on the horizon. When quantum computing converges with AI, we'll have the capability to solve optimization, risk, and decision-making problems at a scale we've never had before. — Victor Foulk, Vice President, Emerging Technologies, CGI Federal

Not every company needs to live on the bleeding edge. However, building organizational muscle— in governance, measurement, and talent—will let you capitalize on what's coming in technology.

Six questions for the boardroom

1. Are we measuring what matters? Quality, stability, and business outcomes, or just activity and velocity?
2. Are our foundations ready for speed? Governance and engineering disciplines ensure AI-accelerated delivery doesn't become AI-accelerated risk?
3. Do we have accountability for AI outputs? Transparency, auditability, and values alignment?
4. Are we enabling people, not just equipping them? AI literacy, change management, and new organizational structures?
5. Do we have a build-or-buy strategy? Strategic differentiation versus where to partner? (Read more here: Choosing the right AI solution: Five build or buy mistakes to avoid (and what to do instead).)
6. Is user value at the center? Accelerating outcomes customers need, or just accelerating output?

Where we go from here

We’re already starting to see that enterprise software is being built through human-AI collaboration as the default: not as an experiment, not as an accelerator bolted onto existing processes, but as the foundational way work gets done. The organizations that will lead aren't necessarily the ones adopting AI fastest today. They're the ones building the governance, the talent, and the measurement discipline to sustain it.

"This is the least capable these systems will ever be." — Jade Hind, Lead Data Scientist, CGI’s UK & Australia Operations

The question isn't whether AI will transform software delivery. The question is whether your organization is building the maturity to make it sustainable, governed, and aligned with real business value. The roadmap is clear. The window to build your foundations is now.

Explore our enterprise software delivery hub

Listen to the From AI to ROI podcast 

Connect with our experts

[1] AI Coding Assistants: Statistical Twins of Rogue Developers (February 2025)