The myth of little or no process or working documentation in agile delivery is often propagated based on the following from the Agile Manifesto:
At the same time, one of the principles of the Agile Manifesto states:
When we attempt to embed a strong culture of continuous improvement both repetition and communication are important. Formal documentation and overt use of team Patterns play key roles.
If we fail to document what we are attempting to improve or fail to improve upon a repeatable process, we will fail to capture reliable results. The relationship between our actions and results may be simply due to random influence.
We should focus an improvement idea on something that is repeatable. From here we can then adopt a standard process for implementing improvement ideas. Sometimes stabilising a process is a necessary first step for improvement.
A common approach for improvement systems is to adopt empirical methods. We inspect the current state, form a hypothesis for improvement, conduct an experiment, observe the results and adjust accordingly. This is consistent with the adoption of Scrum. 
Though implemented in various forms such a process resembles the Toyota Kata as described by Mike Rother (below). 
Key aspects to consider:
By documenting and communicating improvement ideas we can leverage knowledge and learning across an organisation. Importantly, we are sharing the thinking and principles of an improvement, not just the localised improvement. This allows for ideas to be applied in different contexts. 
It is important to create both the environment (physical and/or virtual) and events to support the documentation and communication of implemented improvement ideas. Systemic improvement requires systematic support.
1. Imai, M. (2012) Gemba Kaizen: A Commonsense Approach to a Continuous Improvement Strategy. Second Edition.
2. Sutherland, J. & Schwaber, K. (2020) The 2020 Scrum Guide (TM). https://scrumguides.org/scrum-guide.html
3. Rother, M. Toyota Kata. http://www-personal.umich.edu/~mrother/Homepage.html
4. Akers, P.A. (2015) 2 Second Lean. https://paulakers.net/books/2-second-lean
5. Ballé, M. (2020) Is Lean Scientific? https://www.lean.org/balle/DisplayObject.cfm?o=1299