There is an irony in the technology industry: we adopt agile structures to gain flexibility and, over time, we turn these very structures into rigid rituals. Scrum has become a rite, and Kanban, in some cases, has turned into a wall of purposeless post-its. And the agility that should arise from uncertainty and adapt to chaos has been diluted in ceremonies and velocity metrics.

It is a fact that planning more honestly about uncertainty costs less than pretending it doesn’t exist. Jim Highsmith, creator of Adaptive Software Development (ADS), tested this premise in over 100 commercial projects before publishing the framework in 2000. More than two decades later, projects have become more complex, requirements more volatile, and technological environments more unpredictable. 

In this context, Adaptive Software Development comes back to the forefront with renewed strength, and not by chance. The emergence of Generative Artificial Intelligence tools applied to software development has created a radically new work environment.

What is Adaptive Software Development (ASD)?

Adaptive Software Development is based on the following premise: complex software cannot be planned with precision. It emerges from a collaborative, interactive, and, above all, adaptive process. ASD does not promise predictability. It proposes something more valuable: the ability to respond well to the unexpected.

While traditional methodologies try to control uncertainty through extensive planning, ASD accepts it as a fundamental given. Teams operating in complex domains, such as platforms and data architectures, financial products, and solutions with embedded AI, know that the scenario changes while you are still drawing the map.

The philosophical basis of ASD comes from chaos theory and complex adaptive systems. In practice, it translates into short learning cycles, collaboration with stakeholders, and a permanent attitude of reviewing assumptions.

What are the three phases of the Adaptive Software Development cycle?

The adaptive software development cycle consists of three phases that repeat iteratively, each carrying a specific intention.

Speculate

This is the phase of adaptive planning. The term is not by chance: instead of “planning” with false precision, ASD proposes that the team speculate on what will be necessary, defining directions without intending certainty. Here, the cycle’s objectives, priority functionalities, and value hypotheses are established. It is a planning that recognizes its own limits.

Collaborate

It is in this phase that the work actually happens. The team develops the functionalities of the cycle with an emphasis on continuous communication, both internally and with the client or user. ASD sees collaboration not as a status meeting but as a mechanism for real-time learning. Decisions are made based on evidence that emerges from the work itself.

Learn

This is the phase with the great differentiator of ASD and the most neglected in other methodologies. At this moment, the team analyzes the results of the cycle not only to check what was delivered but to review what it learned about the problem, the product, and the process itself. The learning feeds the next speculation, closing the adaptive loop.

This cycle has an essential property: it does not seek to eliminate errors, but to shorten them. An error made early, learned quickly, and corrected in the next iteration is worth more than the perfect planning that ignores signals from the real world.

How does Adaptive Software Development adapt to the new development context with AI?

The integration of Artificial Intelligence (AI) into software development is a change in the nature of work. When a developer uses an AI agent to generate code, review architecture, or explore implementation alternatives, the process ceases to be linear. It becomes emergent.

And here, ASD offers something that classic Scrum was not designed to provide: a framework that operates well in environments where the output of the next sprint depends on what you learned in this one, not just on what was planned in the backlog.

Generative AI

In projects using generative AI, teams face constant variability: hallucinations in responses, non-deterministic behavior, model latency, dependence on external APIs. Planning in two closed sprints can become obsolete in days. The adaptive speculation of ASD, combined with short learning cycles, allows the team to recalibrate without methodological disruption.

Moreover, the learning phase of ASD creates formal space to review not only what the team did but also what the AI did, and how the team can use that learning to improve prompts, workflows, and integrations in the next cycle. It is a loop of improvement that makes sense in contexts where the technology environment itself evolves while the product is being built.

What is the difference between Adaptive Software Development, Scrum, and Kanban?

The comparison between the three methodologies has direct implications on the choice of work model for specific teams and projects.

Market numbers help to understand the scenario. According to the 16th Annual State of Agile Report, nearly 9 in 10 corporations adopting Agile claim to use Scrum. Kanban appears in half of them. ASD does not appear as a measurable category in the same reports, and this is not a coincidence.

Scrum and Kanban dominated because they are accessible frameworks, with clear roles, well-documented ceremonies, and a training and certification ecosystem. ASD has never had this apparatus. It was built for environments where unpredictability is structural. Therefore, its adoption happens in a less visible and more organic way: in teams dealing with genuinely complex domains, such as fintech, data platforms, and, increasingly, development with embedded AI.

This context matters because the choice of methodology should be dictated by the nature of the problem.

Scrum

The Scrum is oriented to cadenced deliveries with defined roles (Product Owner, Scrum Master, development team) and structured ceremonies. It works well when there is a reasonably stable backlog, requirements that can be refined in advance, and a client willing to participate in periodic reviews. Its greatest risk is the rigidity disguised as process: teams that follow the ritual without internalizing the agile principle behind it.

Kanban

The Kanban is a flow management system, not for iterations. It does not assume sprints or fixed roles. Its focus is to visualize work, limit work in progress, and optimize continuous flow. It is especially effective for operations, support teams, and environments with variable and unpredictable demand. But Kanban, by itself, does not offer a structure for systematic organizational learning.

Adaptive Software Development (ASD)

ASD operates at a different level. It is not a task management system or a delivery framework. Its differential lies not in ceremonies but in posture: to speculate instead of planning with certainty, to collaborate instead of reporting, to learn instead of just delivering.

In practice, ASD is more suitable for high-complexity projects, exploration of new technologies, product development in emerging markets, and, as discussed, projects that incorporate AI deeply. Scrum and Kanban continue to be valid choices but for different contexts.

How to measure the success of a project using Adaptive Software Development?

This is one of the questions that reveals a team’s maturity: what do you measure when the framework does not promise predictability? In ASD, success is not measured by sprint velocity or Kanban lead time. The metrics need to reflect the values of the framework itself.

Quality of learning per cycle

This is one of the most valuable metrics: did the team manage to identify hypotheses that were confirmed or refuted? Was the learning documented and incorporated into the next cycle? Teams that learn quickly deliver better products over time.

Ability to adapt

Can be measured by the frequency and depth of directional changes. It is not about celebrating changes, but evaluating whether the team can make them without process collapse. A mature ASD team changes direction with low organizational cost.

Value delivered per iteration

This metric goes beyond the number of functionalities. It includes impact on business indicators, reduction of technical risk, learning about the user, and validation of product hypotheses.

Stakeholder engagement

It is an indirect but revealing indicator. ASD assumes intense collaboration. If stakeholders are distant or the team is operating in silos, the framework is not being effectively applied.

Time between problem discovery and response

This is perhaps the most honest metric of ASD: how long does it take to identify that a premise was wrong and adjust course? The shorter this interval, the more adaptive the team truly is.

Real agility is not a method, it is a posture

The Brazilian technology market has matured a lot in recent years. Teams talk about OKRs, product discovery, dual-track agile, and continuous delivery with ease. But agility, that which emerges in the face of genuine uncertainty and not just in front of a well-refined backlog, is still a challenge. 

Objective has consistently pursued this path since the agile manifesto and bases its transformation and AI project operations on the principles of agility. One of the examples is an initiative with RaiaDrogasil. The company needed to evolve the maturity of its development teams at scale. Objective acted as a partner, and the results were significant: an 88% reduction in Lead Time, a 486% increase in Throughput, and a 68% growth in estimation accuracy. The digital revenue of RD grew 11 times during the period.

For João Paulo Miranda, CEO of Objective, the case illustrates the difference between agile maturity of superficial framework adoption and discusses the next steps:

“We are insisting on digital literacy, enabling business and technology areas to take full advantage of AI’s potential. Our goal is to measure the efficiency generated and enhance the impact of this technology in our projects.” Before AI, it was agile maturity that created the foundation. The sequence is not a coincidence.

This vision connects to what ASD proposes in its essence: continuous learning as a business asset. A team that learns quickly about AI, about the user, and about the product itself is, in practice, operating with adaptive logic, regardless of the name it gives to the framework. If you are leading an initiative that needs to balance speed, uncertainty, and business outcomes, contact the experts at Objective.

Questions and Answers on the topic