The deal gets the headlines. The integration decides whether the deal was worth it. In Adobe's M&A PMO I ran a reconciliation track across engineering teams, and the work was almost entirely about reconciling things two companies did differently and had no reason to think about until now.
The first surprise is how much is implicit. Each side has systems, access models, and assumptions that were never written down because everyone internal already knew them. Integration forces all of it into the open, fast, under a clock.
I learned to run integration as a set of explicit reconciliation decisions, not a vague "align the teams." For each area: whose system wins, what migrates, what gets retired, who owns the cutover. Vague ownership is where integrations quietly rot.
The other lesson was sequencing under uncertainty. You rarely have full information when the clock starts. You decide anyway, you record what you assumed, and you revisit when better data shows up. Waiting for certainty just means missing the integration window with nothing to show.