Agile in the AI World Is More Relevant Than Ever
In the introduction to the series, I asserted that the Agile Values and Principles have held up nicely, while the practices needed a bit of a refresh. Before we get to the practices, I will first do a fuller review of the Agile Manifesto. I don't believe this document is forgotten, but it certainly is not referenced as frequently and is rarely in the center of the AI discussion.
For this evaluation, I draw from recent LightSight clients who we are helping build solutions that leverage AI. Anything Agile comes with multiple opinions and perspectives, so let me say up front that this review is based on what we are seeing in the market with our clients today. Others will have different views on what's happening "out there." That's what makes this Agile evolution so interesting.
For each Value and Principle, the question I want to get at is the same one. Does it remain relevant and leverageable in the emerging AI world?
The Four Values
I still love the framing of the Values in the Manifesto: "while there is value in the items on the right, we value the items on the left more." This was really brilliant positioning. It was not alienating traditional practices through a full indictment of the way things were done, but rather introducing the Manifesto as a preferred approach.
The four pairs:
- Individuals and interactions over processes and tools
- Working software over comprehensive documentation
- Customer collaboration over contract negotiation
- Responding to change over following a plan
Now, the core of the discussion. What, if anything, has changed with AI to recalibrate the use or importance of these?
Individuals and interactions over processes and tools
Maybe tools should be more important than people in an age where tools can reason? Can we use AI to reduce headcount? The short answer is no. We never recommend this approach because people with your business context can now do even more with AI. Instead, we always ask: what do you wish you could be working on? What are your bottlenecks and the work that you hate? Then our goal is to provide these domain experts with the ability to shape their work accordingly.
Agents are very good at closed problems like solving integrals and writing code to specification. But that's not the type of high-value problems most businesses are dealing with, nor the ones that people want to spend their time doing. Most high-value and interesting problems are relational, complex, and strategic. And in those domains, humans still win.
Agents do have a place inside that high-value work. The line we keep landing on is that agents reduce the cost (time) to answer questions and people use those answers to make decisions. When working with our financial clients, they have workflows that start with extracting specific information from similar documentation repeatedly, so we automate that part. Then they start asking questions about the extracted information. We optimize that through agentic automation and enrich it with external market intelligence, significantly reducing the time needed to answer questions.
We have seen countless Rube Goldberg AI creations that on the drawing board will replace the need for human "interference" in the pursuit of process perfection. But we haven't seen them work effectively without a human in the loop. With increasingly shiny tools like AI agents, it's becoming easier to lose the focus on individuals. The organizations that we see succeeding focus on people first.
With our clients, the use of agents is all about freeing the knowledge worker from routine tasks. When these agents are created to support people, they are removing their most basic lower value work and allowing people to amplify their expertise, liberated from the mundane to live in the strategic.
Working software over comprehensive documentation
One of the AI detractors is what we call “Work Slop.” The temptation to add AI-generated snazzy-looking .htmls and presentations, often with limited value, can simply slow the delivery process as the team cuts through a superfluous veneer to get to what the client really wants. On the other hand, where AI is very much helping the process is through prototyping, the newest dimension to Project Requirements Documents. Prototyping helps everyone visualize a solution to a user pain point with a tangible user interface that can be tested and refined. You follow the same process of user interviews and structuring the project, but instead use agents to visualize the UI/UX. Not only does this help ensure that what is being built aligns with user needs, it creates engineering documentation that reduces implementation back and forth and time.
We will get into more details on prototyping below and in future blogs in this series.
There is a second reason documentation matters more than it used to. Agents need context. When work passes from one agent to the next, or from an agent to a person six months later, the written record is what carries the intent forward. Agents can now write that record themselves. So the sequence matters more than the volume.
Prototypes are becoming the vehicle for "this is what I want" or "please build this." That 1500-page requirements document outlining everything plus the kitchen sink is heading toward obsolescence. And documentation in this new paradigm is often now written by the agents for the agents.
Customer collaboration over contract negotiation
At its heart, this Value centers on committing to adaptability. Structuring flexible agreements allows teams to build agility directly into project deliverables. A quarter-century later, as we navigate heightened uncertainty and ever-tightening feedback loops, this mindset is more vital than ever.
While the previous Principle addresses the artifact capturing intent, this one focuses on the deal structure itself.
Just recently, a prospective client shared a traditional waterfall proposal they received from another vendor. It featured a multi-hundred-page requirements document paired with an implicit "we'll get back to you when it's finished" stance. While such traditional approaches persist, they are increasingly becoming the exception rather than the norm. Modern stakeholders now demand ongoing validation and the freedom to pivot iteratively without getting bogged down by rigid scope control.
When we are building out a solution for our clients, we have our engineers engage immediately and directly with the client, elbow to elbow, discussing and building prototypes. This process accelerates the path to an end result dramatically. Trust climbs quickly when you can demonstrate the art of the possible in an hour or two for your client. This paves the way for rapid software delivery, ultimately creating value for the client quickly.
Responding to change over following a plan
Driven by an emphasis on the preceding Values, contract flexibility has steadily expanded over time. Outside of minor, fixed-scope projects, maintaining a dynamic backlog has effectively become the standard practice today.
In this new era, the pace of change frequently exceeds our capacity to keep track of it. What was logical yesterday often no longer holds true today. Ultimate success depends on developing sharp inspect-and-adapt capabilities rather than adhering blindly to a rigid plan.
So, what about the Principles? Stay tuned for the next blog where we’ll cover the twelve principles.