This is a tricky issue. Relief and development organizations have well established models for delivering aid. Those models are traditionally based on the expert-consumer models: experts who have authority and experience decide how the aid should be "provided" to those in need (this applies to various areas of aid, including education). While this model stems from practical considerations of "transferring value" (here, the value is education), it does not help those communities expand their capacities and become independent. For example, if the expert education program does not engage the locals in the process, then there will be no more education once the initiative is finished or the funding is cut off.
This means that aid experts, managers and educators would have to be more flexible and sacrifice some of the strict planning and quality of their service for the sake of engaging with refugees and explore how they can collaborate with refugees wishing to be involved. This includes even channeling parts of the funding dedicated to employees in the initiative to locals who are willing to join (providing a much needed source of income and a sense of self-respect). This is not easy to do, because it means that expert initiatives would need to accept this change in roles, which might go against funding models that need frequent and reporting on progress with quick impact assessments. This might also go against traditional professional interests (e.g. as a professional I am interested that the need for me remains so that I keep being useful).. but I'm sure people are creative enough to work around this!
For more inspiration around this, here is Idilbi's remarkable first-hand observations about education in Syrian refugee camps.