Github Copilot vs Other Harnesses
I look after AI delivery in my organisation. We currently have Github Copilot as our enterprise AI software development harness, adoption has been slow and the engineers haven’t really built a practice around structuring the repos to be AI optimized , creating agents that do certain tasks well, etc..
The head of software development in our org is now making a push to move to Claude Code and transforming into a full on AI-SDLC. He suggested a rather weird set up (imo) where we build agents using bedrock and agentcore , and use the claude agent SDK.
I believe we can reach 90% of the value of AI assisted coding with copilot github as opposed to other harnesses , and should focus our effort on training engineers so that they build an AI friendly engineering practice/context/repo structure etc..
Is my instinct wrong here? I have been following this space for a while, and while Copilot 365, Copilot Github, and Copilot Studio have a bad rep for being the worst AI offering, Microsoft has been really improving their value over the past few months.