Agile was the last revolution: AI-DLC is the next one

Looking back over the last 25 years of software development, a clear pattern presents itself. Every 25 years or so, the way we build software changes completely.

The first era: From the 1960s through 1980s, waterfall ruled. We wrote extensive requirement documents first, got sign-offs, with the working software delivered in 12 to 36 months. Given code was created with punch cards and expensive computer lab time, it made sense to have gating and phasing in place.

By the 1990s, rapid application development compressed that timeline to 60 to 90 days, and began to change processes: Prototyping replacing documentation and an emphasis on visual tools which could garner faster feedback.

The second era: Then came the Agile Manifesto in 2001, development’s first enormous leap forward. Borrowing heavily from Lean concepts (and the Toyota Production System), Agile changed everything. How teams worked, what roles people played. Continuous improvement, two-week sprints, business owners working directly with the teams doing the building.

By the mid 2010’s, Agile had gained significant traction across the industry. Then Development Operations (DevOps) extended it across the full product lifecycle, tearing down the wall between the people who write the code and the people who deploy and monitor it.

Each of these transitions felt radical at the time. Each became the new normal within a decade. We are now almost exactly 25 years from signing of the Agile Manifesto. And if the pattern holds, we are standing at the next inflection point.
‍

What is the AI Development Lifecycle?

The Artificial Intelligence Development Lifecycle (AI-DLC) is a complete rethinking of how software gets built, enabled by AI agents that work alongside engineers across the entire process, thus creating a new process: A “bolt” which takes you from “inception” and “construction” to operation.

The closest analogy: what Agile did to waterfall, AI-DLC does to Agile. The methodology does not disappear, it gets unleashed.

But before we start building using AI-DLC, in January 2026 we published our responsible AI policy, as it was important to set our design principle, our governance and our way of working.

“AI use at Benevity is guided by principles that ensure it remains human-centered, privacy-respecting, fair, secure and accountable, while supporting social and environmental responsibility." 

This principle matters. The mistake early adopters make is treating AI as a replacement for engineering thinking. It is not. The teams that see the biggest gains are the ones where human judgment about what to build, why to build it, how to define the problem, remains firmly in the loop. AI handles the construction velocity, while humans own the direction.

This is what "human in the loop, always" means in practice. Not a governance slogan, it is our daily design constraint.
 

What we did — and how fast we moved

Moving with such speed required sequencing and the same change management rigor that any leadership team would apply to a structural shift in how the business operates.

Without the board-level sponsorship and leadership support, our AI-DLC would have been no more than a gradual side experiment. Without that alignment, the velocity gains we saw in our best teams would have been impossible because the hardest part of AI-DLC is not the technology. It comes down to the organizational willingness to change and invest.
‍

What the data showed — and what surprised us

Early velocity data from our first AI-DLC bolts showed a wide range of outcomes. Our best-performing team achieved roughly six times the normalized velocity of a comparable traditional sprint. Our worst-performing team underperformed against that same baseline.

In our highest-performing bolt, the team was small, the work was well-defined and the scope was contained. Every person in the room understood what they were building and why. In the bolt that struggled, the Product Requirements Document and the architecture document conflicted. The team spiralled, forgetting that code generation is now the fastest part of the process; not everything is a one-way door, and the AI-DLC teaches us that the bolt’s code is disposable if it doesn’t come out right the first time. 

Our instincts from the beginning were being proved out in real time: what one builds has become more important than ever. AI-DLC compresses the development timeline dramatically. That means poorly defined initiatives collapse faster. Strong ones move faster than you thought possible.

If anything, the definition and inception phases layered with clear specifications have become the most important investment in the entire process. Quality and effectiveness of the feature starts with critical human thinking. How an organization scales the AI-DLC is what really matters.

We believe a 60% improvement in velocity is only the beginning. As our teams develop fluency with the AI-DLC operating model, the ceiling moves higher.
‍

What this means for social impact leaders

AI-DLC is changing how Benevity supports CSR leaders and their programs.

Every capability Benevity ships — Grant Summaries, Match Assurance, the agentic administrator portal, conversational analytics — will run on the foundation of AI-DLC.This is the engine behind the promise that 'what used to take months now takes days.' It is the result of a deliberate transformation in how we build.

For your program, this means capabilities will become available faster. Compliance changes that used to require multi-quarter release cycles can now respond timely to a shifting regulatory environment. Features that would have competed for engineering attention for years can be accelerated when the underlying development model operates at a fundamentally different speed.

This will also change how we collaborate with our clients, from building the specifications to how we pilot and beta test. Of course, speed is only part of the story — quality, trust and reliability remain non-negotiable.

Years from now, we'll look back and see that the organizations that changed social impact were the ones that started investing and building differently with AI today.
 

What it really takes: 5 lessons for SaaS leaders considering AI-DLC

This is an organizational transformation and here is an honest account of what is required.

1. Accept that everything is different now

The AI-DLC rethinks daily ceremonies, makes traditional velocity metrics nonsensical and nearly demands a restructured development team. It does away with story points; everything we know about estimating software no longer applies. One of our teams thought a change would take three days based on their current knowledge of the AI-DLC. It was completed that same morning. ‍

2. Executive alignment is integral

Given the as-yet-unknown nature of estimating, we can no longer use the same metrics we’ve relied on for the past 25 years, and that is uncomfortable. Applying new success metrics takes time, experimentation, executive alignment - all elements that are critical to any adoption of the AI-DLC. 

3. Invest in more focused, capable teams 

AI-DLC rewards teams where every person is a driver, not a passenger. The model does not reward adding more resources to a struggling bolt; it rewards defining the work more clearly.

4. Make room for early failure 

We failed early. Our first teams spent three weeks on inception and construction phases that should have taken days. We course-corrected, adapted the methodology to fit our context and moved forward. The willingness to fail fast; and the organizational transparency to support it - mattered as much as any technical decision we made. We’ll fail again, but faster than we ever could have with Agile. 

5. Put human judgment at the center 

The engineers who thrive in this model are those who know when to trust AI output and when to override it. Who can work out loud, cross-pollinate and speak up when over confident LLMs assume too much. Developers and their knowledge have not been replaced, it’s that where they spend their time and exercise their expertise has changed.

‍

The next 25 years start now

Each era of software development has fundamentally changed how we design, code and build. Waterfall gave way because shorter feedback cycles, improved adaptability and better collaboration were possible with Agile. And now Agile will give  way to the AI-DLC for the same reasons. 
‍
We’re making a bold organizational decision to rebuild our engineering team from the inside out. That is what it means to be AI-native. Not a feature list. Not a marketing claim. An organizational decision to rebuild from the inside out with human judgment at the center, AI accelerating everything around it. Governance first. Infrastructure first. People first.

At Benevity, we are not adopting AI-DLC simply to build faster. We are adopting it because the clients and programs we support: the employee giving, the grantmaking, the nonprofit partnerships, the communities behind every transaction deserve a platform that moves at the speed of the world they operate in. The people who depend on us deserve a platform that moves at the speed of the world they're navigating — and we're committed to delivering it.

AI-DLC is the beginning of a revolution in software development, the next chapter and we at Benevity are ready. 

Join us at Benevity Live! 2026 to witness how we are becoming an AI-native enterprise impact platform.
‍

Frequently asked questions

What is the AI Development Lifecycle (AI-DLC)?

The AI Development Lifecycle is a specification-driven software development methodology where the process pulls context out of the engineering team in a structured way, rather than engineers feeding information into the AI using ‘vibes’. In our opinion, it does not replace Agile; it unleashes it. Human judgment remains central: engineers define direction, review outputs and govern decisions. AI handles the construction of the code; the result is significantly faster delivery of higher-quality software than traditional sprint-based development.

How does AI-DLC differ from using AI coding tools like GitHub Copilot?

Copilot and similar tools add AI assistance at the individual code-writing stage. AI-DLC rethinks the entire process, with organizational process and scale naturally following. Teams using individual AI coding tools inside see incremental gains; we believe teams operating on AI-DLC will see structural transformation.

What is the biggest risk in adopting AI-DLC?

Beyond transformation, the biggest risk is not the technology, it is having the right specification quality early. AI-DLC dramatically accelerates construction. That means poorly defined work will ensure teams struggle and trash. 

What does this mean for Benevity clients?

AI-DLC is the mechanism behind the commitment at Benevity to faster, higher-quality innovation. Capabilities that previously required long development cycles can now be designed, built and delivered significantly faster. For social impact program leaders, this translates to an AI-native platform that responds to the regulatory environment, client feedback and emerging program needs at a pace that was not previously possible.

What does it take to adopt AI-DLC at an organizational level?

Executive sponsorship from the board level down, a genuine reason to change (urgency, not just interest), and a more empowered engineering team, tolerance for early failures, and a culture where human judgment remains at the center of every consequential decision. And while many of our developers are excited about the new pace of our ability to deliver, this type of success and collaborative way of working is not for everyone, nor is the rate of change one that everyone feels comfortable with. It is a transformation, not an upgrade. Organizations that approach it as a pilot or a side experiment are unlikely to see the full benefit.

{{divider}}

‍Benevity powers social impact programs for leading companies around the world — from employee giving and volunteering to grants management and nonprofit disbursements.