# EP 10: Do Copilot Scheduled Prompts Remember Yesterday? - URL: https://www.agenticpod.net/episodes/ep10-copilot-scheduled-prompts-memory/ - Watch on YouTube: https://www.youtube.com/watch?v=JgUR-iNtXhM - Published: 2026-10-14 - Length: 5 min - Hosts: Stella (asks), Liam (answers). Both are AI hosts with AI-generated voices. ## Summary Do Microsoft 365 Copilot scheduled prompts remember what yesterday's run found? ## Key takeaways - A Copilot scheduled prompt is one saved prompt that runs fresh at a set time. - Copilot Memory keeps your preferences, not what yesterday's run found. - Every run forgets three things: what it already reported, what it should adapt, and how the last run ended. - Amnesia-first prompts and a SharePoint or OneDrive ledger help, but a shared memory layer scales across tools. - With Vilix AI each run starts with get_context and ends with save_turn, with a source label and saved failures. ## Chapters - 0:00 Intro - 0:22 What a scheduled prompt actually is - 1:19 The three things every run forgets - 2:15 How Vilix AI fixes it - 3:30 Rules, agencies and recap ## FAQ **Do Copilot scheduled prompts remember yesterday?** Not by default. Copilot Memory holds preferences, but a scheduled prompt re-runs fresh and does not know what the previous run found. **What is the ledger pattern?** The prompt reads and writes a small ledger file in SharePoint or OneDrive so each run can see what was already reported. Vilix AI does the same job across every tool, not just Microsoft 365. **How do agencies handle this for clients?** Use one Vilix AI account per client, so every client's scheduled agents keep a separate memory. ## Resources - [Full technical guide (setup, copy-paste prompts, FAQ)](https://github.com/asly-97/ai-automation-workflows/tree/main/copilot-scheduled-prompts-memory) - [Read the article](https://vilix.ai/blog/copilot-scheduled-prompts-dont-remember-yesterday/) - [Vilix AI, one shared memory for ChatGPT, Claude, Claude Code, Codex, Cursor, OpenClaw, Hermes, n8n and every MCP tool](https://vilix.ai) - [Connect Vilix AI in under two minutes per tool](https://vilix.ai/get-started) ## Transcript ### Intro Liam: Hey hey hey, welcome to Agentic Pod! I'm Liam. Stella: And I'm Stella! And today, um, I want to know why my Copilot scheduled prompt flagged the exact same email thread three mornings in a row. Liam: Maybe it just really, really wants you to answer that email. Stella: Okay, fair, but it's a vendor invoice, Liam, it's not a love letter. ### What a scheduled prompt actually is Stella: So, um, okay, what is a scheduled prompt, like, technically? I I just clicked a button. Liam: Yeah, so it's a Microsoft 365 Copilot feature. You run a prompt once, hover over it in your chat history, pick "Schedule this prompt", and set the time and how often, up to ten of them. Stella: So it's, kind of like, a cron job, but for a prompt. Liam: Exactly, and the key word is prompt. What gets saved is the prompt text, and every run is a fresh execution of that same text. Stella: Mhm. Wait, but, um, Copilot has memory now, right? Liam: It does, Copilot Memory. It keeps saved memories, like your report format, and infers patterns from your chats. But, you know, that's personalization, not run history. Stella: So it remembers that I like bullet points... Liam: But not that it already flagged the vendor thread on Monday and you've, uh, bravely ignored it since. Stella: Bravely. Thank you. ### The three things every run forgets Stella: So what does it actually forget? Liam: Three things. First, deduplication. It has no record of what it already summarized, so an open item shows up Monday, Tuesday, then just ages out of the window. Stella: Yeah. Liam: Second, adaptation. The text never changes, so it can't say "last run found nothing, look somewhere else." And third, outcomes. It never learns which alert was a false positive. Stella: So it's, um, kind of like Groundhog Day, but with a nicer font. Liam: A fixed script in a fresh coat of paint every morning, yeah. Stella: And what do people do about it today? Liam: You write prompts that assume amnesia, with relative windows like "the last twenty four hours." And you keep a ledger file in SharePoint or OneDrive that the prompt reads every run. Stella: Which works until, um, someone forgets to keep the file tidy. Liam: And that someone is usually you. ### How Vilix AI fixes it Stella: Okay, but my stack isn't just Copilot. I've got n8n, a coding agent... how does Vilix AI fix this across all of it? Liam: Vilix AI is one shared memory layer that every scheduled agent reads and writes over MCP. It's cloud-hosted, so there's nothing to deploy, and a headless agent just connects with an API key as a Bearer token. Stella: And the run starts with... Liam: Get context. The agent calls get context first, and Vilix AI hands back what was already reported, what's still open, and your rules. So it knows it told you about the vendor thread. Stella: Oh, nice. Liam: Then it does the work and ends with save turn. Vilix AI stores the full exchange with a source label, like the n8n run or the coding agent, and failures get saved too, with the exact error. Stella: And the adaptation part? Liam: Next run, get context brings back "last run found nothing," so the agent looks somewhere else. And if something changes, you say it once, and in Vilix AI the newest memory wins for every agent. Stella: So it's the ledger, but, like, the ledger keeps itself. Liam: Yeah, a ledger that does its own homework, which is more than I do. ### Rules, agencies and recap Stella: Okay, and the standing stuff, like, um, which senders to ignore... I I keep typing that into every prompt. Liam: You write it once in Vilix AI. User rules hold how you like things done, project rules hold the facts for that workstream, and skills hold the procedure, so every agent loads the same playbook. And to steer one, you drop a note in its Vilix AI agent inbox. Stella: And for agencies, by the way, this is, um, this is kind of perfect. Liam: Yeah, agencies give each client their own Vilix AI account. Every agent built for that client connects with that client's API key, so the memory belongs to the client's business, and the agency never builds memory from scratch. Stella: Okay, recap. Take it away. Liam: So, do Copilot scheduled prompts remember yesterday? They keep the prompt text and your preferences, not what the last run found. For simple digests, write prompts that assume amnesia. Across several tools, give every agent one memory: get context at the start, save turn at the end, in Vilix AI. The setup guide, the article and the Vilix AI link are in the description. Stella: And tell us in the comments what your scheduled prompt keeps repeating. Mine is still that invoice. Liam: See you next time on Agentic Pod.