Its a major paradigm shift that is not usually achievable in my experience. The time taken to estimate/re-estimate could be used to deliver the new product or system change. The preference for hypothetical/abstract prediction [the "Linus blanket"" of crystal balling, magic wands and other quasi-religious ceremonial artifacts] compared to actual business value realising delivery. I support the principle of #noestimates movement, which is to shift organisations to the lean priniciple of maximising value creation, and reducing waste. Yes the argument can be made that there is value to be be derived from estimations and plans, but this needs to be carefully considered as to the actual level of effort required versus the perceived or objectively quantifiable "value"[cost saving? increase in efficiency/profitability?] or is just for "peace of mind"or to secure contractual parameters that in the end prove unreasonable to hold a vendor hostage in an untenable situation.
Again I have witnessed first hand the success of moving business organisations to the truly lean agile approach and maximising delivery of actual value where the correct stakeholders actively participate in the project with a medium term [6month to 12month] high level roadmap and then short term timeboxed [sprints - 2weeks] plan driven by a prioritised backlog, and continuous delivery/integration approach to evolve the final project product deliverables.
In addition the project team needs to achieve lean flow and empirical techniques such as cumulative flow measurements, and velocity, to create metrics on delivery [as opposed to hypothetical metrics in planning such as planning poker]. Extrapolating the evidence based data allows for projections for completion which replace traditional up front estimation techniques. Allow the work to emerge and commit budget to tranches of work. Again this is usually too much of a radical shift from the safety blanket of "knowing everything upfront".
Author: Richard Smith - for: Knowledgesmith