Rethinking Estimations in the Age of AI

Rethinking estimations in AI-driven processes has become essential in the fast-paced world of IT and software development. Estimations shape project timelines, budgets, and expectations, yet traditional practices often lead to inefficiency and frustration. In a previous post, I explored the #NoEstimates movement and its call to move beyond traditional estimation techniques. Here, I’ll dive deeper into how historical data and AI can revolutionise estimations, making them more reliable and less burdensome.

The Core Question: Why Do We Estimate?

Estimation practices serve multiple purposes, each critical in its own right. However, understanding these purposes is key to evaluating whether traditional methods are still relevant or if new approaches can address the same needs more effectively.

Predictability

In my experience, one of the most common reasons for estimation is to provide predictability. Organizations I have worked with rely on estimates to plan timelines, allocate budgets, and manage resources effectively. Predictability builds trust with stakeholders, ensuring they feel confident in delivery schedules and can make informed decisions.

Alignment

Estimates foster alignment by encouraging collaboration and discussion among team members. They surface assumptions, clarify scope, and provide a shared understanding of priorities and challenges. While this may not be the main goal of estimation, in my experience, it is a huge benefit when done in a group setting. Activities like Planning Poker not only create alignment but also stimulate productive conversations and team bonding. Read more about Planning Poker here.

Other Reasons for Estimation

• Budgeting and Resource Allocation: Estimations help determine funding and staffing needs for projects.
• Feasibility Assessment: Teams use estimations to evaluate whether goals are achievable within constraints like time and cost.
• Prioritization: Estimations guide decision-making by highlighting the effort required for different tasks.
• Risk Assessment: By identifying high-effort or complex tasks, estimations help teams mitigate risks.
• Contractual or Stakeholder Requirements: Clients and external stakeholders often demand estimates to validate project scope and commitments.
• Motivational or Commitment Tools: Teams often use estimations as anchors to create a sense of urgency and accountability.
• Transparency and Communication: Estimations help communicate complexity and effort to stakeholders, fostering trust and clarity.

The Role of Historical Data in Estimations

The reliance on gut feelings and individual judgment in traditional estimation methods is fraught with issues, including bias and inconsistency. Historical data offers a more objective and reliable alternative.

Examples of Historical Data

• Velocity: The amount of work completed in a sprint or iteration.
• Cycle Time: The time it takes to complete a task or feature.
• Estimated vs. Actual Hours: Comparing planned effort to actual effort provides insights into accuracy and trends.
• Defect Trends: Patterns in defects can reveal bottlenecks or quality issues.

How Historical Data Improves Estimations

By analyzing past performance, teams can identify patterns and make evidence-based predictions about future work. For example, if a team’s average cycle time for similar features is well-documented, this data can be used to forecast completion times without the need for speculative estimates.

Consider a team that tracks recurring bug fix rates or delivery trends for standardized feature rollouts. Instead of assigning story points, they analyze past performance data such as the frequency of similar tasks or average team throughput. This approach allows them to forecast delivery timelines effectively, bypassing the need for subjective estimations like story points.

Shifting from Gut Feel to Data-Driven Predictions

Traditional estimation methods often rely on human intuition, which can be flawed due to biases like over-optimism or anchoring. Data-driven methods, supported by AI, provide an opportunity to improve both accuracy and trustworthiness in estimations.

Pitfalls of Human Estimation

• Bias: Overconfidence or pessimism can skew estimates.
• Inconsistency: Different individuals may provide widely varying estimates for the same task.
• Lack of Accountability: Without data, it’s difficult to validate or learn from past estimation errors.

How Data-Driven Methods Address These Issues

• Objectivity: Historical data removes emotional and subjective factors from the equation.
• Scalability: AI tools can analyze large datasets quickly, identifying trends and making predictions that would be impractical for humans.
• Accountability: Data-driven methods create a feedback loop where teams can compare forecasts to outcomes and continuously improve.

AI’s Role in Estimations

AI’s ability to analyse historical data at scale offers unprecedented opportunities for improving estimation accuracy and reliability. By identifying patterns and predicting task durations, AI can provide confidence intervals or ranges, rather than single-point estimates, enhancing predictability. Moreover, AI helps highlight potential bottlenecks or risks early, enabling teams to address them proactively. Recognising and leveraging this potential is a transformative step for organisations aiming to reduce inefficiencies and increase trust in their estimation processes. AI can process historical data at scale to:

• Identify patterns and predict task durations.
• Highlight potential bottlenecks or risks.
• Provide confidence intervals or ranges, rather than single-point estimates, enhancing predictability.

For example, tools like Jira Advanced Roadmaps can analyse historical sprint data to forecast project completion timelines. Similarly, Forecast.app, which uses machine learning algorithms, evaluates past project metrics to predict resource allocation and task durations with high accuracy. These tools not only improve accuracy but also free teams from the repetitive effort of manual estimations, allowing them to focus on value-driven tasks.

Conclusion

Estimations have always been a cornerstone of software development, but their traditional practices are ripe for transformation. By shifting from human intuition to data-driven methods, teams can achieve greater predictability, alignment, and efficiency. Historical data and AI are not threats but tools to enhance decision-making and reduce unnecessary guesswork.

The future of estimations lies in embracing these innovations and challenging the status quo. It’s time to rethink how we estimate—not just to meet today’s demands but to build a foundation for greater reliability and trust in the processes that drive our industry.