The slower adoption in non-tech sectors isn't just cultural lag - the integration surface is genuinely harder. Legacy ERP systems, compliance review cycles for what data the model can touch, and change management overhead all front-load costs before any efficiency shows up in margins.
I've seen this in payment/API systems: the actual model integration takes weeks, but getting legal and security sign-off on the data pipeline takes months. Non-tech companies face the same pattern but with less internal tooling to manage it.
The margin signal might also be appearing at the wrong level. Gains in these sectors often show up first as headcount flatness or throughput improvements before they hit EBITDA. Measuring at the P&L level on a 2-year horizon is probably too early and too coarse - the operational metrics are moving, the accounting just hasn't caught up yet.
The one-way pattern is actually how Google uses it internally too, syncing outward from their monorepo to GitHub. Bidirectional gets messy because transforms (path remapping, file exclusions, header stripping) are easy to apply in one direction but can't always be cleanly inverted. When both sides have diverged, Copybara's baseline tracking starts producing confusing results because semantically equivalent commits generate different SHAs after transform.
One thing worth knowing: history "preservation" is actually cherry-picks with rewritten commits, not a true transplant. Git blame works because the file content and authorship carry over, but the SHAs are new. Copybara embeds the original SHA in a commit message trailer (GitOrigin-RevId), which is useful to know if you ever need to correlate commits across repos after the fact.
Lately interested in how AI agents change software testing, integration reliability, and production verification.