How One Agency Runs a Grain MCP Daily Brief Across Every Client Call
Aanarav Sareen's company tripled in size in twelve months. Here's the daily brief and the Claude scoping workflow he built on Grain's MCP, the one-sentence AI policy behind both, and what it all saves him: at least half a day every week.
TL;DR
- Aanarav Sareen, CEO of P3 Media, runs a Perplexity agent that pulls Grain's MCP every morning, reads every meeting from the last 24 hours, and emails him a brief sorted by priority: what's on fire, what's not, and what needs his attention.
- The setup saves him at least half a day per week, roughly 16 hours a month. His verdict: "It definitely pays for itself."
- The daily brief is how he keeps a pulse on every client and every meeting without sitting in every room.
- His AI policy is deliberately experimental right now, and it compresses to one sentence: anyone can use any tool, and client or sensitive data never goes into any of them.
Why growing companies lose the pulse
P3 Media tripled in size over the last 12 months.
That kind of growth creates a specific problem. The leader who used to hear everything now hears a fraction of it. Client calls happen all day across the team. Gaps open in the system, and nobody notices until a gap becomes a problem.
Aanarav Sareen, P3 Media's CEO, wanted a pulse on the business: "what is happening from a business perspective across the board, across every meeting, across every client." Not to micromanage, but to catch gaps early. "If there are gaps in the system, I can pick up on that and then kind of have that resolved."
He says he didn't have a good way to do that before. Now he has two workflows built on Grain's MCP.
Why consistency wins
Aanarav's case for automating this layer of his business isn't speed. It's consistency.
"Unlike a human, technology usually doesn't miss," he says. "Like if it's working consistently, it's usually working. It's hard for it to not catch things… Of course that requires like really good systems overall."
The comparison isn't AI versus a good employee. It's a system versus human factors. "I have an assistant, I have a team that oversees a lot of stuff," he says. "But let's say they're on vacation, they're traveling, they're sick. Whatever, all these things, like human factors, which is normal. Then when you miss the day to day, it ends up being a problem."
A system has no days off. "You know that the system, if it's working, it's always working," he says. And when it breaks, you fix it and backfill: "It can still populate a bunch of the stuff. Feed it in again."
Workflow 1: Aanarav's daily brief
What is a Grain MCP daily brief?
A Grain MCP daily brief is an automated summary of every meeting your company recorded in the last 24 hours, generated by an AI tool that queries Grain's Model Context Protocol server on a schedule. The AI reads the meetings, sorts what it finds by priority, and delivers the result somewhere you already look, like your email inbox. You get a daily pulse on the business without attending or rewatching anything.
How it runs
- A Perplexity agent runs in the background on a schedule. Nobody touches it.
- Every morning it pulls Grain's MCP and gathers every meeting from the last 24 hours, across every client and every recorded call.
- It sorts what it finds by priority and sends a summary email.
The sort order is the point. "It does it based on priority," Aanarav says. "So it's like, what's on fire? What's not on fire and what's good and what needs my attention."
What it produces
A morning email that works like triage. Fires at the top, stable accounts at the bottom. This is the one output he doesn't hand off. "My daily summary, you know, that doesn't get delegated to someone else."
Why this is worth copying
Every leader of a growing company has wanted this and settled for skimming Slack or asking around. The setup takes three things: a meeting recorder with an MCP, an AI tool that can run on a schedule, and an inbox. If your meetings are already in Grain, the data side is done.
Copy this prompt
Connect Grain's MCP to an AI tool that can run on a schedule, then save this as the scheduled prompt. Swap the section names for whatever your business actually tracks.
Using the Grain MCP, pull every meeting recorded in the last 24 hours.
Write me a brief sorted by priority, in three sections:
1. Most important: anything with a blocker, a slipped date, an unhappy
client, or a risk to revenue. Say what happened, which client,
and what it needs from me.
2. Less important: decisions waiting on me, and commitments
anyone made on a call.
3. Stable: one line per account, no detail.
Cite the meeting title and date for every point. If a section has
nothing in it, say so instead of padding it.
Workflow 2: Client scoping with Claude, Grain, and Shopify's MCP
How can I use MCP servers?
An MCP server gives an AI tool a live connection to one system: your meetings, your storefront, your codebase, your CRM. Connect one and the tool can read real data instead of whatever you remember to paste into the chat. Connect two or more and it can work across them in a single request, which is where most of the value shows up. One server supplies the intent, meaning what people actually said they wanted. Another supplies the constraints, meaning what your systems can actually do. The AI drafts against both at once, and a person checks the result before it goes anywhere.
How it runs
- A new client project needs a scope.
- In Claude, Aanarav pulls the relevant meeting recording from Grain's MCP. That's the client's ask, in the client's own words.
- He ties it to the MCP server from Shopify for the technical side.
- Claude drafts the technical requirements. A human validates them before anything ships.
"When I'm writing scopes for new clients, or even when I'm doing technical scoping, I can pull in the meeting recording from Grain, tie it with the MCP server from Shopify, and then just get the technical requirements," he says. "And of course, have that validated."
What it produces
Scope documents grounded in what the client actually said, drafted in a fraction of the time. "The process of doing that is significantly faster," he says. "It allows me to build solutions and look at solutions much, much faster."
Unlike the daily brief, this one is delegable. He can hand it to a teammate with real context attached: "These are the projects that we've worked on in the past. Go look at all our meeting recordings and see what kind of projects we've worked on to make that happen."
Why this is worth copying
This pattern is not specific to Shopify. Pair Grain's MCP with whatever MCP server covers your domain: your codebase, your CRM, your data warehouse. The meeting holds the intent, and Shopify, or the other server, holds the constraints. Claude joins them.
Copy this prompt
Connect both Grain's MCP and your domain's MCP server to Claude, then paste this in. Replace the bracketed parts with your meeting and your platform.
Using the Grain MCP, pull the meeting titled "[meeting name]" and
read the full transcript.
From what the client actually said, draft technical requirements
for this project:
1. What they asked for, in their words, with quotes.
2. Requirements, grouped by area.
3. Constraints they mentioned: budget, timeline, current platform,
integrations.
4. Open questions where their ask was ambiguous. Do not fill gaps
with assumptions.
Then use the [Shopify] MCP to check each requirement against what
the platform actually supports, and flag anything that needs a
workaround or a custom build.
A human still validates the output before it goes to the client. Aanarav's own version is longer than this one. Treat it as a starting point and build on it once you know what your other MCP servers need.
The two workflows side by side
The pattern across both workflows
- The output lands where people already look. An email inbox, a Claude conversation. Neither workflow asks anyone to open a new dashboard.
- The prompt is written once and saved as a skill. Aanarav's is "pretty lengthy," but encoded: "We have skills for all of it. It's like, tell us what's happening on the sales side. Tell us what's happening on a client side. Tell us what's happening on the technology side." Nuanced to the business, written one time.
- Priority does the editing. A summary of everything is noise. "What's on fire" first is a brief.
- Humans validate what ships to clients. The daily brief informs. The scopes get checked before they go out.
- Consistency is the feature. Systems don't take vacation. That's the whole argument.
Their AI policy: one sentence, still an experiment
P3 Media doesn't have an AI rollout plan. It has a rule, and Aanarav is clear that the rule is a current experiment rather than settled policy. The team is still working out which tools and which workflows give them the best results, so the guardrail is loose on purpose.
"Right now it's free for all. Anyone is allowed to do whatever they would like, of course with some data controls," Aanarav says. "We don't share client information." Sensitive data stays out of every tool. "That I think is our biggest constraint, but that's by design."
Adoption is spreading without a mandate. He was the first MCP user at the company. Now, "we're starting to get more people on board."
This is where they are today, not a template to lock in. Expect the rule to tighten as they learn more about what works.
And token consumption? Not a problem so far. "Not at all," he says. "Not yet, at least."
What to try this week
- New to Grain's MCP? Connect it to Claude and ask one question about yesterday's meetings.
- Leader trying to keep a pulse? Copy the daily brief. Schedule an AI tool to pull the last 24 hours of meetings each morning, sort by priority, and email you. Fires first.
- Power user? Steal the scoping stack. Grain's MCP plus your domain's MCP server in one Claude conversation. The meeting supplies the intent, the other server supplies the constraints.
- Worried about data? Copy the one-sentence policy: anyone can use any tool, and client or sensitive data never goes into any of them.
Frequently asked questions
How do you set up an automated daily summary of your meetings?
You need three pieces: meetings recorded in Grain, an AI tool that can run on a schedule, and Grain's MCP connecting them. Aanarav Sareen runs an agent that pulls the MCP every morning, reads the last 24 hours of meetings, and emails him a priority-sorted summary.
How much time does a Grain MCP workflow save?
Aanarav estimates at least half a day per week, roughly 16 hours a month, across his daily brief and his Claude scoping workflow. His summary: "It definitely pays for itself."
How do you use Grain's MCP with Claude for project scoping?
Pull the client's meeting recording from Grain's MCP inside Claude, add the MCP server for your technical domain, and ask for the technical requirements. For P3 Media that second server is Shopify's. The output is a draft scope grounded in what the client actually said, validated by a human before it goes out.
Resources
- See more workflows here: grain.com/ai-workflows
- Get started with Grain MCP: grain.com/integrations/mcp
- Book time with the Grain team: grain.com/demo
