How VSM helps IT teams deliver faster, improve processes more precisely, and use AI without adding chaos.
11.12.2025 This publication is an adapted translation of the "2025 State of AI-Assisted Software Development" report by Google Cloud and DORA. The original is available via the link. Reading time: 7 min. This chapter was prepared with the participation of Rob Edwards, Application Delivery Lead, Google Cloud.
All organizations are expected to move faster today.
Everyone is adopting AI, automating processes, building platforms, and shipping new features at breakneck speed.
But are we actually getting better?
Or are we just building features with no value faster, burning out faster, and increasing complexity faster?
The main risk now is not falling behind, but investing major resources in chaotic activity that does not produce results.
For more than ten years, DORA research has been based on a core idea: the most effective organizations do not just adopt tools, they become experts in the value delivery system.
They know how to "get better at getting better"
: understand your workflow, see the real bottlenecks, and direct resources deliberately. This year we confirmed that the ability to analyze and manage the value stream is exactly what separates chaotic activity from focused improvement. According to a 2025 study, teams that pay attention to understanding their value stream spend significantly more time on work that creates real value. And most importantly:
(VSM) is the amplifier that turns AI investment into a competitive advantage. It helps direct AI toward solving the right problems instead of creating more chaos.
Value Stream Management (VSM) is the practice of visualizing, analyzing, and improving the flow of work from idea to customer.
It is not a complicated bureaucratic process, but a set of four principles that help create clarity and focus improvements where they really matter.
A detailed step-by-step guide to value stream mapping is available in the DORA Value Stream Management guide.
It is hard to understand a complex system.
It is impossible to keep every detail in your head all the time, and that prevents you from seeing the big picture.
When the team visualizes the system together, all those details move out of individual heads and into a shared space.
The structure of the system, and its hidden patterns, become visible.
It makes the conversation about what works and what does not much simpler. In essence, that is exactly what VSM is about.
The practice is to map the entire software delivery lifecycle, from idea to customer.
It covers everything: product discovery, design, development, testing, deployment, and operations.
A shared visualization helps the team build a collective understanding of the process and spot bottlenecks and inefficiencies faster.
The goal is to map the entire system from concept to customer. But you do not need to do everything at once.
The key is to start where the impact will be greatest.
Before you start, look at the process as a whole and identify the main limiting factor so you do not optimize a part that is not the real bottleneck.
If your team's main difficulty is at the product discovery stage, that may be the right place to begin.
A good starting point, used by DORA for many years, is the path from commit to production.
It is the easiest to standardize and optimize.
Even more important, this is where teams have the greatest opportunity for fast, visible improvement.
This differs from discovery stages, where the main goal is to improve effectiveness rather than speed.
Successfully improving the core process creates quick wins, builds trust, and creates momentum for changes in broader areas such as product discovery and user feedback collection.
Example of a value stream map: the path of a task from backlog to production.
Once the stream map is in place, the key task is to ensure predictable, smooth flow of work.
To do that, stop optimizing isolated parts and start improving the system as a whole. Start by measuring what matters.
Track the metrics: lead time, process time, and the share of value relative to waiting time.
These metrics show the system's real constraints and provide a baseline for understanding whether improvements are working.
This systems view helps identify where new tools or technologies will have the greatest impact. For example, a team may see on the map that code review is the bottleneck.
In that case, it makes sense to use AI to speed up that process rather than to generate more code, which would only increase overload.
The main gain is not speeding up a single step, but removing the real bottleneck.
That is the point of focusing on flow: solving the problem of the whole system instead of speeding up one stage. "And there is also the complexity of the process itself: how much work will it take? how long will it take? how do you even estimate it? There is always uncertainty and a feeling that you are not in control. Will we be able to make the needed changes quickly, as we go? Or will we have to go through a two- or three-week process every time just to change a firewall rule, for example? Something like that.
There are too many ways this process could be slowing our work down. I think the more experienced people understood this especially well, because they had already encountered it in other large companies."
A fast and sustainable flow of work is impossible without a solid technical foundation, most often a well-designed internal platform.
By giving developers "paved paths"
for testing, delivery, and other key practices, such a platform removes unnecessary complexity and makes effective work scalable. The link between a strong technical foundation and organizational performance is covered in detail in part 6.
For many years, DORA has recommended using VSM practices to enable fast and stable flow of work.
But the question remains: is this approach still relevant now that AI is being adopted everywhere? This year, our goal was to test our long-standing hypotheses about the value of VSM.
The results confirm:
Teams that regularly analyze and improve their value stream show significantly higher performance. VSM increases the share of "valuable" work.
Such teams spend much more time on tasks that truly matter to the business and customers. VSM improves product quality.
A focus on the value stream leads to better product outcomes, which is ultimately what matters most.
The data shows a simple fact:
A shared understanding of the process helps focus effort on what matters and turn clarity into concrete improvements.
VSM lets you be strategic: instead of local optimizations, you remove the system's real bottleneck.
This observation led us to the key hypothesis of the 2025 study: VSM turns AI into an organizational advantage. We assume that VSM strengthens the link between AI adoption and organizational performance growth. Teams that use VSM direct AI's benefits toward solving systemic problems, ensuring that targeted improvements truly lead to company-wide results.
Without VSM, AI can create local optimizations that get swallowed by downstream bottlenecks and add no value to the organization as a whole. VSM shapes AI's impact on organizational performance.
Our analysis confirms this hypothesis.
While AI adoption on its own has a moderate effect, that effect becomes much stronger in organizations with a mature VSM practice.
This confirms that VSM is a critical mechanism for realizing real returns on AI investment.
When gains in productivity at the team and individual level are directed at key system constraints, VSM turns local improvements into meaningful organizational outcomes.
This research is not just an observation, but a challenge.
The first step out of the cycle of chaotic activity is simple, but not easy: ask your team, "Can we draw our software delivery value stream on a whiteboard?"
If the answer is no, or the diagram raises more questions than it answers, then you've found the starting point. And that conversation is where the path begins to getting better at getting better.
To understand why AI's impact on teams differs so much from expectations and from company to company, it is important to look not at the tools, but at the system in which they operate.
The next chapter explains this key principle: AI acts as a mirror of organizational processes, amplifying strengths and exposing weaknesses.
We examine what actually determines AI's effect and which conditions turn isolated acceleration into a meaningful shift in the whole system.
