Staying Across a Global Project Without Staying Up For It...

I sit on a project that touches business, tax, technology, and legal, all in the same workstream, all moving at different speeds, and all involving people spread well beyond my own time zone. Being based in EMEA means a fair share of the calls that matter happen at hours I am not always available. The obvious answer is to ask someone to fill me in. The less obvious, and better, answer turned out to be one I did not need to build, just configure properly. Microsoft 365 Copilot is grounded, by default, in my email, my Teams chats, the SharePoint sites I have access to, and meeting recaps where transcription was switched on. A project like this rarely lives in one channel. Legal raises something in a chat thread, technology posts an update to a SharePoint folder, tax picks it up over email, and business discusses it live on a call I was asleep for. Copilot reading across all four in a single pass is the actual value here, not any one source on its own. The prompt I actually use I test manually before scheduling anything, and I keep refining it rather than treating the first version as final. "Search my emails, Teams chats with [named colleagues across business, tax, technology, and legal], and SharePoint updates from the last 7 days related to [project name]. Summarise progress, key decisions, open risks, and outstanding actions by workstream. Include anything from meeting recaps for calls or workshops I did not attend live, particularly any held outside EMEA hours." A few things I have learned to be deliberate about. - I name people, not just workstreams, since a general project keyword pulls in noise and naming the specific person in legal, the specific person in technology, narrows both the chat and email search to what is actually relevant to me. - I ask for it by workstream, explicitly, because a single undifferentiated summary blurs a legal risk into a technology delay into a tax position, when really I want to see each one separately and judge for myself what needs my attention this week. - I have had to be honest about what depends on other people entirely. Copilot can only summarise a call after the fact if transcription was turned on during it, and that is not automatic, an organiser has to enable it. For the calls I am structurally unlikely to attend, the ones running late into an EMEA evening, I have had to ask whoever is chairing them to keep transcription switched on as a matter of course. Without that, there is nothing for the summary to draw on for exactly the meetings I need it most for. Once the manual version reads close to something I would write myself, I schedule it, weekly, Monday morning, so I am reading it fresh rather than trying to process it on a Friday when my attention is already elsewhere. Managed from the Scheduled Prompts tab in the Copilot panel, easy enough to adjust as the project's stakeholders or emphasis shift. What this is not, and where I still check it myself Worth being precise about what this actually is, given the ground it covers. This is not an agent, nothing is taking multi-step action or writing anywhere on my behalf, which matters on a project with legal and tax exposure, I would want more scrutiny before letting anything act autonomously in that space. It is not a Project either, no dedicated instructed workspace sitting behind it. It is native grounding plus a schedule, and for the specific problem of staying current across time zones without staying up for them, that turned out to be exactly enough. I do not treat the summary as settled fact for anything that touches legal exposure or a tax position. Two things I watch for specifically: a decision the summary states as agreed when it was genuinely still contested on the call, and a workstream update attributed to the wrong meeting entirely. Both are easy for a model pulling across four different sources to get subtly wrong, and on a project like this, subtly wrong is worse than obviously missing.