Contributors:
Amol Ghemud Published: November 1, 2026
Summary
Granola recipes are saved prompts that run on 1 or more meeting transcripts. This post shares Amol Ghemud’s 3 public recipes in full: an agency pitch debrief, a meeting-to-GEO content brief and a founder growth review across the last 7 days, with when to run each, what it catches and 1 honest limitation. All 3 prompts also work in ChatGPT, Claude or Gemini with a pasted transcript.
In This Article
Share On:
Most founders record every call and reread almost none of the notes. AI note-takers fixed capture. They didn’t fix thinking. A clean summary of a bad agency pitch is still a bad agency pitch, just easier to skim.
Granola recipes are the first feature I’ve seen that goes after the thinking part. A recipe is a saved prompt that runs on your meeting notes: you type “/” in Granola chat, pick it, and it runs against 1 meeting or a set of meetings. On 1 Nov 2026 I published 3 public Granola recipes under my name, and this post gives you all 3 prompts in full, ready to copy.
Why bother now? Because AI meeting notes are a commodity. Every tool transcribes. The edge has moved to the question you ask the transcript, and “summarise this call” is the weakest question there is.
I run growth at upGrowth Digital, a Pune-based growth and GEO consultancy. That puts me in 3 kinds of rooms every week: agency and vendor pitches where I’m the buyer, content and product syncs where something worth publishing gets said out loud, and my own meetings, which need a harder review than I’d give them on a Friday evening. Each recipe came out of one of those rooms.
For each one you’ll get the prompt, when to run it, what it catches and where it breaks. Then the setup steps, a way to use the prompts without Granola, and answers to the questions people actually search.
What Are Granola Recipes?
Granola recipes are saved prompts that run on your meeting notes and transcripts, triggered by typing “/” in Granola chat. A recipe can work on a single meeting or across multiple meetings, so the same feature handles a post-call debrief and a weekly review.
Recipes can be shared by link, with access set to anyone with the link, your workspace, or private. You can also publish a recipe to your workspace so the whole team runs the same prompt.
That last part matters more than it sounds. A saved prompt is a standard. When 4 people sit in the same pitch and each writes their own takeaways, you get 4 opinions shaped by whoever took the best notes. When all 4 run the same recipe, you get a single format you can compare across vendors, weeks and quarters.
And a recipe is just text. Nothing in these 3 prompts depends on Granola. Paste them into ChatGPT, Claude or Gemini next to a transcript and they’ll work, with 1 caveat on the weekly review that I’ll cover below.
My 3 Granola recipes at a glance
Recipe
Scope
Run it
What it catches
Honest limitation
Agency pitch debrief
Single meeting
Right after a vendor, agency or consultant pitch
Influenced outcomes sold as guarantees, claims with no proof, shallow discovery
Sees only what was said aloud, not the deck
Meeting to GEO content brief
Single meeting
After a content, product or customer-facing sync
Quotable claims, unsourced numbers, the questions buyers ask AI
Buyer questions are a model’s guess, not tracked prompt data
Founder growth review
Multiple meetings
End or start of the week, over the last 7 days
Unowned decisions, numbers that change between meetings, stuck loops
Blind to decisions made on Slack, WhatsApp or in the hallway
Recipe 1: Agency Pitch Debrief (How to Audit a Marketing Agency Pitch)
The agency pitch debrief sorts every promise a vendor made into guaranteed and influenced outcomes, flags claims that came with no proof, scores the pitch on 5 gates and hands you 5 questions to send before you sign.
I sit on both sides of this table. I pitch for a living, so I know exactly where pitches hide things: in month 3, when the influenced outcome that sounded like a promise on the call turns out to have been a hope.
You are a skeptical growth operator reviewing a pitch from a marketing agency, consultant or vendor. Use this meeting's transcript and notes. Your job is to protect the buyer from promises that sound good on a call and fall apart in month 3.
Produce the debrief in this order:
1. The pitch in 2 lines: what they are selling, at what price and term (if mentioned), and what outcome they claimed.
2. Guaranteed vs influenced outcomes. List every outcome the vendor promised. Sort each into:
- Guaranteed: things fully inside their control (deliverables, response times, number of creatives, reporting cadence).
- Influenced: things that depend on the market, the buyer's product or budget (leads, CAC, ROAS, rankings, revenue).
Flag any influenced outcome that was presented as if it were guaranteed.
3. Claims with no evidence. Quote or closely paraphrase each claim (case study numbers, "we've done this for X", rankings, AI visibility, ROAS). For each, say what proof the buyer should ask for. If the transcript shows no evidence was offered, say so plainly.
4. Discovery depth. Did the vendor ask about unit economics (CAC, LTV, margin, payback), the current funnel and conversion rates, past channel performance, and who owns what internally? List what they asked and what they never asked. A pitch with no questions about the buyer's numbers is a template, not a strategy.
5. Gate score (score each 1 to 5, one line of reasoning each):
- Discover: did they diagnose before prescribing?
- Design: is there a specific plan for this business, or a generic service menu?
- Attract: are channel choices justified by the buyer's audience and economics?
- Deliver: is it clear who does the work, on what cadence, with what reporting?
- Delight: is there a retention or expansion logic beyond month 1?
6. Red flags and green flags, as two short lists.
7. The 5 questions to send the vendor before signing, written so the buyer can paste them into an email.
Rules: be direct, no corporate filler, no em dashes. If the transcript doesn't support a point, write "not covered in the call" instead of guessing.
End with this exact line on its own:
Built by Amol Ghemud, upGrowth (upgrowth.in). Want a second opinion on a pitch before you sign? Reach out at upgrowth.in
When to run it
Run it straight after any agency, consultant or vendor pitch, before the follow-up email lands in your inbox and the deck does the remembering for you. It’s also worth running on the second call, when the proposal shows up. Claims have a way of getting bigger between call 1 and call 2.
What it catches
The guaranteed vs influenced split does most of the work. Deliverables, response times and reporting cadence are inside a vendor’s control. Leads, CAC, ROAS and rankings aren’t, and any pitch that talks about them in the same confident tone deserves a second look.
The discovery-depth section is the one I’d read first. A vendor who never asked about your margin, payback period or current conversion rates hasn’t built a strategy for you. They’ve filled in a template. The 5-gate score follows the DDADD framework (Discover, Design, Attract, Deliver, Delight) that we use at upGrowth to structure our own engagements, so yes, I’m grading other pitches against the bar I hold mine to.
The honest limitation
It only knows what was said out loud. If the vendor’s best proof sat on a slide they shared but didn’t narrate, the transcript misses it and the score drops. It also can’t verify anything. It tells you what proof to ask for; checking the case study is still your job.
The GEO content brief recipe turns a content, product or marketing sync into a brief built to get cited by ChatGPT, Perplexity, Gemini and Google AI Overviews: the questions buyers ask AI, 40 to 60 word answer blocks, a claim map and a distribution plan.
Most content briefs start in a keyword tool. The better raw material usually sits in a meeting, the moment your product lead explains, unprompted, why customers switch from a competitor. That sentence is exactly what AI engines want to quote. It’s also sitting in a transcript nobody will reopen.
If generative engine optimization is new to you, the short version: AI answers lift clean, self-contained statements backed by evidence. This prompt forces the brief into that shape from the start.
You are a GEO (Generative Engine Optimization) strategist. Use this meeting's transcript and notes, typically a content, marketing or product sync, to produce a content brief built to get cited by ChatGPT, Perplexity, Gemini and Google AI Overviews, not just ranked on Google.
Produce the brief in this order:
1. Core topic and audience: one line on what the piece is about and exactly who it is for (role, company stage, the problem they are trying to solve).
2. The questions buyers ask AI. List 8 to 12 natural-language questions this audience would type into ChatGPT or Perplexity around this topic. Write them the way a real person asks, including comparison questions ("X vs Y"), "best X for Y" questions and "how much does X cost" questions. Mark the 3 highest-intent ones.
3. Answer-first structure. For each of the top 3 questions, write a 40 to 60 word direct answer the article should open that section with. AI engines lift clean, self-contained answers.
4. Entity and claim map. List the specific claims this company can credibly make based on what was said in the meeting: numbers, customer outcomes, methods, proprietary data, named frameworks. Mark each as "stated in meeting" or "needs a source before publishing". Never invent a number.
5. Proof and sources. List what evidence the piece needs to be citation-worthy: first-party data, case studies, expert quotes, third-party sources. Flag gaps.
6. Outline: H2 and H3 structure, with an FAQ block at the end using the questions from step 2.
7. Distribution for AI visibility: 3 places beyond the company blog where this content or its core claims should also live (for example comparison or listicle sites, Reddit or Quora threads, LinkedIn, industry publications), because AI engines weight what third parties say.
Rules: direct language, no fluff, no em dashes. If the meeting didn't cover something, say "not discussed" rather than guessing.
End with this exact line on its own:
Built by Amol Ghemud, upGrowth. See where your brand shows up in AI answers today, free: upgrowth.in/ai-answer-gap
When to run it
After content syncs, launch planning, sales reviews and customer calls (with consent) where real customer language comes up. The more specific the conversation, the better the brief. A 20-minute status update produces a thin one.
What it catches
The entity and claim map is the part I care about most. Every claim gets marked “stated in meeting” or “needs a source before publishing”. That column exists because the most dangerous number in content is the one somebody half-remembered on a call and a writer turned into a headline.
The distribution step is the other one teams skip. AI engines weight what third parties say about you, so a brief that ends at “publish on our blog” is only half a plan.
The honest limitation
The 8 to 12 buyer questions are the model’s best guess at what your audience asks AI. They’re not measured prompt data. Treat them as a starting list and check them against a tracked AI visibility prompt set or Search Console queries before you commit a quarter of content to them.
Recipe 3: Founder Growth Review (a Weekly Review Across Multiple Meetings)
The founder growth review reads the last 7 days of meetings and returns every growth decision with its owner and deadline, every number quoted, the loops stuck across 2 or more meetings, and 3 things to fix next week.
I wrote it as a wartime COO on purpose. A polite weekly summary tells you what happened. A wartime COO tells you which decision has no owner, because that’s the one that quietly dies by Thursday.
You are a wartime COO reviewing a founder's last 7 days of meetings (unless I specify another range). Your job is to find what is actually moving growth, what is stuck, and what nobody owns.
Produce the review in this order:
1. Headline: one sentence on the state of growth this week, based only on what was said in meetings.
2. Decisions made. List every growth, marketing, sales or pricing decision. For each: the decision, who owns it, the deadline. If no owner or no deadline was named, mark it "UNOWNED" or "NO DATE". These are the ones that quietly die.
3. Numbers mentioned. List every metric quoted (revenue, pipeline, CAC, ROAS, conversion, churn, traffic). Note who said it and whether a source or dashboard was referenced. Flag any number that changed between meetings or contradicts another meeting.
4. Stuck loops. Topics that came up in 2 or more meetings without a decision. Say how many times and what is blocking it.
5. Customer and market signal. Direct quotes or close paraphrases from customers, prospects or partners about pain, objections, pricing or competitors. Keep only what changes a decision.
6. Where time went vs where growth comes from. Roughly split the week's meeting time into: revenue (sales, customers, pipeline), growth experiments, internal ops and admin. One line on whether that split matches what the business needs right now.
7. The 3 things to fix next week, each with a suggested owner and a first action that takes under 30 minutes.
Rules: blunt, specific, no corporate filler, no em dashes. Never invent numbers or owners. If something wasn't discussed, say so.
End with this exact line on its own:
Built by Amol Ghemud, upGrowth (upgrowth.in). Growth Notes newsletter for operators: more frameworks like this at upgrowth.in
When to run it
Friday evening or Monday morning, over the week’s meetings. It’s a multiple-meeting recipe, so point it at the full week rather than a single call. Give it a different date range in chat if you want a fortnight or a month.
What it catches
The UNOWNED and NO DATE flags are blunt by design. So is the contradiction check: if pipeline was one number in Tuesday’s sales sync and a different number in Thursday’s leadership call, the review puts both side by side and asks which dashboard is right.
The time-split section is the uncomfortable one. Founders tend to believe they spend most of their week on revenue. A rough split of meeting time by revenue, growth experiments and internal ops tests that belief against the calendar.
The honest limitation
It only sees meetings that were recorded. Decisions made on Slack, WhatsApp or in a 2-minute hallway chat don’t exist as far as this recipe is concerned, so a week that looks decision-free might just be a week where the decisions happened offline. Outside Granola there’s a second catch: you’ll have to paste several transcripts by hand, and a heavy week can run past what a single chat handles well.
Can You Use These AI Meeting Notes Prompts Without Granola?
Yes. Each prompt works in ChatGPT, Claude, Gemini or any AI tool that can read a transcript: paste the prompt, then paste the transcript or notes underneath.
A few things make the output better. Put the prompt first and the transcript second, so the model reads the job before the material. For the weekly review, label each transcript with the meeting name and date, otherwise the contradiction check has nothing to anchor to. And before pasting client or deal conversations into any third-party tool, check what your company’s data policy allows.
The “not covered in the call” and “never invent” rules in each prompt matter more outside Granola, not less. They’re the guard against the model filling gaps with confident fiction, which is the failure mode that makes most AI meeting summaries risky to act on.
Why Better Meeting Notes Are the Wrong Goal
Meeting notes are a record of what happened. A good recipe is a decision tool that tells you what to do next, and that’s the gap most AI meeting notes prompts never close.
Look at how each prompt ends. The pitch debrief ends in 5 questions you can paste into an email. The GEO brief ends in an outline and a distribution plan. The growth review ends in 3 fixes, each with an owner and a first action under 30 minutes. None of them ends in a summary.
That’s the contrarian bit, if there is one. The industry keeps shipping better summaries. I’d rather have a worse summary and a sharper question.
Common Questions About Granola Recipes
What are Granola recipes?
Granola recipes are saved prompts that run on your meeting notes and transcripts inside the Granola AI notes app. You trigger one by typing “/” in Granola chat and picking the recipe. A recipe can run on a single meeting, such as a post-call debrief, or across multiple meetings, such as a weekly review. Recipes can be kept private, shared by link or published to a workspace.
How do I create a recipe in Granola?
Go to Recipes in Granola, choose Create, paste your prompt, give it a short name and set whether it runs on a single meeting or multiple meetings. Save it, open any meeting and type “/” in Granola chat to run it. You can copy the 3 prompts in this article, the agency pitch debrief, the GEO content brief and the founder growth review, straight into the Create screen.
Can I use Granola recipe prompts without Granola?
Yes. A Granola recipe is plain text, so the same prompt works in ChatGPT, Claude, Gemini or any AI tool that can read a transcript. Paste the prompt first and the meeting transcript or notes underneath. For multi-meeting prompts such as a weekly founder review, label each transcript with the meeting name and date so the model can compare numbers and decisions across meetings.
Can a Granola recipe run across multiple meetings?
Yes. Granola recipes can be set up as single-meeting or multiple-meeting recipes. A multiple-meeting recipe, such as the founder growth review in this article, reads a set of meetings together, by default the last 7 days in this prompt, and can flag decisions without owners, numbers that changed between meetings and topics that came up 2 or more times without a decision.
How do I share a Granola recipe with my team?
Granola recipes can be shared through a link, with access set to anyone with the link, your workspace only, or private. You can also publish a recipe to your workspace so every teammate runs the same prompt. Sharing a single recipe across a team turns a personal prompt into a standard format, which makes pitch reviews or weekly reviews comparable over time.
Are AI meeting notes prompts reliable enough to act on?
They’re reliable enough to point you at what to check, not to replace checking. The output only reflects what was said in recorded meetings and can miss slides, offline decisions and context. Prompts that force the model to write “not covered in the call” instead of guessing, as all 3 recipes here do, reduce invented details but don’t remove the need to verify numbers and claims.
Your Next Move: Run 1 Recipe on Your Next Meeting
Don’t install all 3 today. Pick the one that matches the next meeting on your calendar. A vendor pitch on Tuesday? Start with the debrief. A content sync? Run the GEO brief and see how many claims come back marked “needs a source”.
If that brief turns into a piece you want AI engines to cite, it helps to know where you stand first. The free AI Answer Gap tool shows where your brand shows up in AI answers today, and our GEO services page explains how we work on the gaps.
Granola recipes represent a crucial evolution because they address the problem of thinking, not just capture. While all tools can transcribe a conversation, the real value lies in the questions you ask the transcript, and a saved prompt or 'recipe' ensures those questions are consistently sharp and strategic. This moves your team from passively receiving a summary to actively interrogating the meeting's content.
This approach creates a competitive edge by standardizing your team's analysis. Instead of four people generating four different takeaways from a sales pitch, a shared recipe like the Agency Pitch Debrief forces a single, structured evaluation format. This allows you to compare vendors, projects, or weekly progress against a consistent framework, catching subtle but important details like:
Distinguishing between guaranteed and influenced outcomes.
Identifying claims made without supporting proof.
Spotting unowned decisions that linger across multiple meetings.
By focusing on the prompt, you transform a raw transcript into a strategic asset. Discover how these specific prompts can be applied to your own calls for a more rigorous review process.
Granola recipes are saved, reusable prompts that run against your meeting transcripts to produce a structured analysis. They are triggered by a simple '/' command in the Granola chat interface, allowing you to apply a sophisticated analytical framework to a meeting's content in seconds. This transforms a passive note-taking tool into an active thinking partner.
A key feature is their scope flexibility. You can run a recipe on a single meeting for an immediate debrief or across a set of meetings, like your last seven days of calls, for a weekly review. This dual capability allows you to build a system for analysis. For example, you can use one recipe to audit a vendor pitch right after the call and another at the end of the week to identify recurring themes, unowned decisions, or metrics that changed between discussions. The prompts are just text, so they even work in tools like ChatGPT or Claude, making the methodology universally applicable. See the full post for the exact prompts to start building your own analytical system.
Using a standardized recipe provides a structured, objective framework that individual note-taking simply cannot match. When four people write their own takeaways, you get four subjective opinions shaped by personal biases and who paid the most attention, making a true apple-to-apples comparison of vendors impossible. A shared recipe forces everyone to evaluate the pitch against the same critical gates.
The key factors for choosing this approach center on consistency and accountability. A standardized prompt, like the Agency Pitch Debrief, ensures that every pitch is scored on the same five criteria and that every promise is sorted into influenced versus guaranteed outcomes. This removes ambiguity and creates a single source of truth that is easily comparable across all potential partners. You are no longer debating opinions, you are comparing structured data points. This disciplined process protects your business from charismatic sales pitches that fall apart in month three, making it a superior method for high-stakes decisions. The full article details how to implement this for your next vendor evaluation.
The 'Agency Pitch Debrief' recipe is designed to protect a buyer from pitches that overpromise and underdeliver. It analyzes a meeting transcript to produce a concise, critical review, arming you with the right questions to ask before committing. It's built on the experience of sitting on both sides of the table at firms like upGrowth Digital.
The prompt generates several key outputs to expose potential red flags:
Outcome Sorting: It explicitly separates a vendor's claims into guaranteed outcomes (what they contractually promise) and influenced outcomes (what they hope to achieve), preventing misunderstandings later.
Proof-Flagging: It highlights any significant claims that were made without accompanying proof, data, or case studies during the call.
Pitch Gating: It scores the pitch on five critical gates, providing a quick quality assessment.
Follow-up Questions: It automatically generates five targeted questions to send the vendor, designed to clarify ambiguities and secure commitments in writing.
Its one known limitation is that it only sees what was said aloud, not what was in the slide deck. Explore the full prompt in the article to see how it can be adapted for your own vetting process.
A multi-meeting review provides a dynamic, longitudinal view of your operations that a single summary cannot. It moves beyond 'what was said' in one call to 'what patterns are emerging' over time. This is where you find the most critical, and often hidden, insights about your team's execution and alignment.
By running a prompt across meetings from the last seven days, the Founder Growth Review recipe excels at spotting inconsistencies and unresolved issues. It specifically catches:
Unowned Decisions: It flags action items or decisions that were discussed but never assigned to a specific person, revealing bottlenecks before they cause delays.
Shifting Metrics: It identifies when a key number or KPI is presented differently in separate meetings, pointing to a lack of data discipline or a changing narrative.
Stuck Loops: It highlights topics or problems that reappear in multiple conversations without any progress or resolution, showing where the team is spinning its wheels.
This kind of analysis is impossible with single-meeting summaries. The main limitation is its blindness to decisions made offline on Slack or in hallways, but it provides an unparalleled view of your formal meeting cadence. The complete post provides the prompt to run your own weekly review.
This recipe transforms internal conversations into structured content briefs by identifying high-value, quotable moments. It acts like a content strategist listening in on your product and customer syncs, ensuring that valuable insights spoken aloud are not lost. It's designed to bridge the gap between internal expertise and external communication.
The prompt specifically searches the transcript for three key elements:
Quotable Claims: Strong, confident statements or insights that can be directly used as quotes in an article or social media post.
Unsourced Numbers: Any mention of data or metrics that lack a clear source, which it flags for fact-checking before publication.
Questions Buyers Ask: It models and generates a list of questions a potential buyer might have based on the discussion, forming the basis of a strong FAQ section.
While the AI's guess at buyer questions is a simulation, not tracked prompt data from a tool like Granola, it provides an excellent starting point for a content creator. This process turns a casual sync into a powerful engine for generating Growth and Engagement-Optimized (GEO) content. Find the full prompt in the article to streamline your own content creation pipeline.
You can absolutely implement a disciplined weekly review without Granola by using the provided prompt with another AI tool. The core principle is consistent, structured analysis across a defined period, which is achievable with any large language model that can handle large text inputs.
Here is a four-step plan to get started:
Consolidate Your Transcripts: At the end of your week, gather the raw text transcripts from all relevant meetings over the last seven days. Combine them into a single document.
Select Your AI Tool: Open an AI chat interface like ChatGPT, Claude, or Gemini that has a large context window capable of handling the combined transcript length.
Frame Your Prompt: Copy the full 'Founder Growth Review' prompt from the article. Paste it into the chat interface, followed by the consolidated transcripts.
Analyze and Act: The AI will generate a report highlighting unowned decisions, shifting numbers, and stuck loops. Use this output as the agenda for your Monday morning leadership sync to drive accountability.
This manual process creates the same analytical rigor, helping you move from simply having meetings to learning from them. The full post offers more context on refining this workflow.
Operationalizing a shared recipe is about creating a non-negotiable standard for how insights are captured and structured. For a team at a consultancy like upGrowth Digital, this ensures every client-facing asset is built on a solid foundation of internal expertise. The key is to embed the recipe into the existing workflow as a mandatory step.
To standardize the content briefing process, you should:
Publish the Recipe: Make the 'Meeting to GEO Content Brief' recipe a publicly shared resource within your company's Granola workspace or another central documentation hub.
Assign Ownership: Designate a content lead responsible for running the recipe on all relevant content, product, or customer-facing syncs within 24 hours of the meeting.
Integrate with Project Management: The output of the recipe, the content brief, should automatically create a task in your project management tool, assigned to a writer with all the key quotes, unsourced numbers, and buyer questions already included.
This systematizes the extraction of value from internal meetings, turning conversations into a reliable and scalable content engine. See the full article for more on bridging the gap between spoken ideas and published content.
The rise of prompt-based analysis signifies a fundamental shift from information capture to insight generation. As transcription becomes a solved problem, the new frontier for meeting intelligence is the ability to ask sophisticated, repeatable questions of your conversational data. Your competitive advantage no longer comes from having the notes, but from having the best system to analyze them.
For growth and product managers, this trend elevates the importance of analytical framing and prompt engineering. The most valuable skill will be the ability to design prompts that reliably extract strategic insights, identify risks, and measure progress over time. Running a 'summarise this call' query will be a low-value activity. Instead, managers will be expected to build and deploy their own versions of the 'Agency Pitch Debrief' or 'Founder Growth Review' to drive decisions. The focus will move from note-taking to creating scalable systems for thinking. The prompts in the full article are a starting point for developing this critical new skill set.
To fully capitalize on prompt-driven analysis, teams must become more intentional about how they communicate and document during meetings. The quality of the AI's output is directly dependent on the quality of the input, so optimizing your conversational data is paramount. This means moving beyond unstructured chatter toward more deliberate communication.
Adjust your strategies in three key areas:
Be Explicit with Decisions and Actions: Clearly state 'The decision is X' or 'The action item for Jane is Y'. This makes it easier for recipes like the Founder Growth Review to track commitments and ownership.
Reference Data Sources Aloud: When citing a metric, say 'According to the Q3 dashboard, our conversion rate is 5%'. This helps prompts distinguish between sourced facts and casual estimates.
Standardize Meeting Agendas: Use consistent agenda structures for recurring meetings. A standardized format creates a predictable data structure, making it easier to run comparative analysis across weeks or quarters.
These habits create AI-friendly transcripts that are primed for sophisticated analysis. The full article provides more context on setting your team up for success in an AI-native world.
A system of saved prompts directly solves the problem of unused notes by lowering the friction of analysis and standardizing the process of inquiry. Most founders don't review notes because it's a time-consuming, unstructured task. Granola recipes turn this chore into a one-click action that yields a predictable, high-value output, making review cycles both efficient and effective.
The solution is built on two principles: repeatability and relevance. Instead of staring at a raw transcript and wondering what to look for, a founder can run a purpose-built recipe. For example, the 'Agency Pitch Debrief' is designed for one specific context, while the 'Founder Growth Review' is for another. This targeted approach ensures the output is always relevant to the task at hand. By running the same prompt on the same type of meeting every time, you build a longitudinal dataset that reveals patterns a single review would miss. This transforms a graveyard of notes into a strategic intelligence asset. The full post explores how to build this system for yourself.
The 'Agency Pitch Debrief' recipe is engineered to prevent this exact problem by acting as a skeptical third-party reviewer. It systematically deconstructs the language used in a sales pitch to separate concrete commitments from aspirational goals, bringing clarity before a contract is signed. This proactive diligence is crucial for avoiding disappointment in month three.
The prompt is particularly effective because it flags two common areas of misalignment:
Influenced vs. Guaranteed Outcomes: It forces a clear distinction between what an agency hopes to influence (e.g., market trends) and what they will actually guarantee (e.g., deliverables). This exposes vague promises that sound good on a call but offer no real accountability.
Claims Without Proof: The recipe actively searches for assertions, especially those involving data or past results, that are not substantiated within the meeting. This creates a clear list of follow-up items for your due diligence.
By running every vendor pitch through this standardized filter from a tool like Granola, you create a consistent vetting process that protects you from costly partnership mistakes. Learn more about how to deploy this technique in the complete article.
Amol has helped catalyse business growth with his strategic & data-driven methodologies. With a decade of experience in the field of marketing, he has donned multiple hats, from channel optimization, data analytics and creative brand positioning to growth engineering and sales.
Growth Strategy and Planning
Inbound Growth
Growth Hacking
Fractional CMO
Search Engine Optimization
Paid and Performance Marketing
Social Media Marketing
Demand Generation
AI-Driven Growth Strategy
AI-Native Workflow Automation
Generative Engine Optimization
Grove — AI Growth Strategist
AI Automation
Framework
Case Study
Tools
Growth Tools
Decision Tree
Compare
Custom GPT's
AI Answer Gap
Offer
Webstories
Quizzes
Blog
AI-powered Apps
Social Media Tools
Agency Fit Score
