Optimization

Agile meets AI

By February 6, 2021No Comments

Atom Planner is a cloud-based system that offers intelligence as a service to optimize the use of shared and constrained resources.

From the day it was first conceived, the team understands that such optimization can be applied to predictive projects and agile approaches, although we have initially focused on traditional projects.

This presentation was first conceived for quick interaction with author Scott Ambler (Choose your WOW) to demonstrate how RSO (Resource Scheduling Optimization) would interact with Disciplined Agile.

Artificial Intelligence in projects is a way to seek organizational outcomes with the integration of methods, math, and tools for optimum performance. The core function of Atom Planner is the application of RSO, sometimes called RCO (Resource Clustering Optimization), and this presentation starts with one of the focal points of Ambler´s book: “Better decisions lead to better outcomes.”

Principles of RSO

We argue that RSO is a great fit for the Lean Change Management Cycle (illustration above), as it provides insights and options for the teams to develop their Minimum Viable Change (MVC).

To understand how RSO will be applied to Agile, it is necessary to understand a few Resource Scheduling Optimization principles.

  • All work is performed by resources.
  • A resource can only work one task at one time
    (does not mean it cannot multi-task)
  • The order in which multiple tasks with multiple resources are executed may increase or decrease the total work performed at one time-box.
  • Tasks are defined by 8M
    manpower, machine, material, method, milieu,
    measurement
    , momentum, and money.

RSO and Deliveries

To illustrate RSO’s application to Agile, we will use two DELIVERIES that will be affected by resources and their work: D1 and D2.

A big point of attention to understand the value of RSO is:

Depending on efforts needed and resources applied to Delivery 1 (task A+B) and Delivery 2 (task C+D), starting from D1 instead of D2 or vice-versa may drive deliveries to fit or not a certain time-box.

Resource Leveling

One of the stages to reach the Resource Scheduling Optimization is to generate resource leveling to tasks (work) expected to deliver.

That starts when we see that parts of a Delivery may take more time than others, as illustrated below.

What is been showing here is that Delivery 1, which has two building blocks (design and development; or development and testing; or ideating and executing; etc.), each block has a size of 4 days.

For Delivery 2, the first part (TC) is smaller. It will take only one day, while the next one will take 4 days as the rest of the tasks.

In this exercise, we have also defined that SCOTT will do the blocks that are starting points of each delivery (TA and TC), and Peter is doing the other blocks (TB and TD)

Once we remember that no resource can do more than one task at one time, Scott must finish TA to execute TC. Once TA is released, Peter can work on TB, but he cannot work on TD immediately as he is still working on another task. In this scenario, once TB is finished, Peter will then execute TD.

For this example (considering a time-box of 10 days), when breaking the tasks flow, executing the work will not fit on the team goals. This means that even if the size of activities is estimated and they seem to fit the required time-box, the sequence that will happen once Scott starts from TA will not work for Peter on time.

If we run optimization procedures for the resources available and necessary tasks, there is actually a combination of task flow that will empower the team to conclude the tasks under the expected 10 days.

In other words: Although teams may have their set of tasks already estimated to fit a certain time-box, the way tasks are flowing between the team members and also external factors, such as meetings with customers, can prevent the team from achieving their goals.

AI Resource Scheduling Optimization offers alternatives for teams to figure out ways to improve their deliveries through the realization of a project.

The structure of self-managed teams have great value for the development of Agile projects. Still, to certain conditions, they may not have the condition to realize certain gains/losses in the work network, or task flow, without the assistance of analytic data calculations, simulations, and scenarios built by AI Resource Scheduling Optimization.

Comparing results (plain resource leveling and optimization)

As illustrated before, depending on what Delivery the team starts an iteration, more or less work may be executed during the expected time. To be able to use AI for finding such opportunities, it is necessary to consider:

  • Backlog tasks must be defined for each time-box through a set of parameters, such as business value, effort, resources required, logical dependencies, priority, due dates, and so on.
  • With that, AI Resource Scheduling Optimization provides the right strategy for effectively achieve organizational outcomes.

Going back to our original goals:

And our final conclusion is:

Comments

After the presentation given to Scott Ambler, author of Choose your WOW, his feedback brings important insights about AI and Agile.

Author

I am Peter Mello, CEO of Atom Planner, and I would like to opportunity to “Thank S. Ambler for watching the original presentation and giving me a path to follow through.”

Peter Mello (peter.mello@atomplanner.com)

 

Leave a Reply

twelve + four =