Strangler Approach for Replacing an Old System
Replace functionality piece by piece without a big-bang switch.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Plan the gradual replacement of [OLD_SYSTEM]. Its main functions: [FUNCTIONS]. Users and integrations: [USERS_AND_INTEGRATIONS]. New platform: [NEW_PLATFORM]. Constraints: [CONSTRAINTS]. Identify seams where traffic can be routed to new code. Rank functions by value and risk, choose the first slice, and describe how requests are routed, how data is kept consistent, how results of old and new are compared, and how to roll back each slice. Return a slice table, a routing diagram in text, a data synchronisation approach, a testing method with parallel runs, a decommission checklist for the old parts, and the points where the plan could fail. Include signals that mean stop and reassess. Provide a first-slice example: choose a low-risk read-only function, build it in the new system, route one percent of traffic to it, compare responses, and increase gradually. Describe how to handle shared databases during the transition, with options such as synchronising data in one direction or having one system own each table. Add a risk register with at least six entries, and a plan for training the support team on both systems while they coexist. Explain the exit criteria for each slice and what evidence is needed before shutting down the old function. Include a note on keeping a record of decisions as the plan changes, since long migrations lose their history. Estimate effort as a range per slice with the assumptions, and warn that the last ten percent of functions often take longer than the first ninety. Add a section for stakeholders outside engineering: a plain-language timeline, the effect on users at each slice, what support will say, and a monthly status format with three indicators. Include a rule that no new features are added to the old system once a slice has moved, apart from critical fixes. Add a short list of what must be decided before the first slice moves: who owns the routing layer, how incidents are handled when two systems are involved, and who has authority to stop the programme if the exit criteria are not met.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.