#4 GTM AI Insider: How to use Custom GPTs in a workflow
Video overview from Jonathan Moss on using multiple GPTs in one workflow and a step by step guide.
Custom GPTs act like digital workers tailored to your specific needs. They’re trained on your data and help reduce repetitive tasks, save time, cut costs, and boost consistency. You can use them for things like creating ICPs, building outbound email sequences, preparing for calls, or competitive analysis. This guide walks through how to create and use custom GPTs effectively.
Step-by-Step Guide
1. Know What You Want
Start with clarity. What are you trying to achieve? Write it down. Maybe you want to automate research, streamline email sequences, or prepare better for calls. The clearer the goal, the better the result.
2. Map Your Process
Outline the steps you follow to reach that goal. Be specific. If it’s creating an ICP, for instance, list every step: gather data, define firmographics, identify pain points, etc.
3. Write a Prompt
Take your process and turn it into a clear instruction for the GPT. Speak naturally; you don’t need perfect wording. If you’re stuck, ask the GPT to help you refine it into a prompt.
You can use prompting frameworks to help you such as GRACE one that Coach K made:
• Goal/Role: Clearly state the goal or the role that the AI needs to adopt. For example, “Generate a content strategy targeting healthcare CIOs interested in AI solutions for reducing operational costs.”
• This is direct and leaves no ambiguity about the purpose.
• Rules: Set clear parameters for how the AI should approach the task. “Ensure each piece of content highlights the cost benefits of AI, includes a CTA, and adheres to a professional tone tailored to C-suite executives.”
• By setting specific rules, you ensure the output aligns with strategic KPIs or messaging guidelines.
• Advanced Context: Provide as much background information as possible. “Our AI platform has helped reduce costs by 15% in hospitals and healthcare networks by automating back-office operations. We need to emphasize this success in the content.”
• This advanced context gives the AI enough information to tailor the content to your specific goals.
• Clear Instructions: Outline the exact structure or deliverable you’re looking for. “Provide a 5-part blog series. Each blog should focus on a specific operational pain point in healthcare and how our AI solution addresses it. Include real-world examples.”
• This ensures that the AI produces a complete and structured output that fits the required format.
• Examples: Finally, provide a sample or reference to guide the AI. “For reference, check our existing case studies for tone and structure. See the case study on ‘AI in hospital administration’ as an example of the detail level we’re aiming for.”
• Examples help the AI to mimic or improve upon previous outputs.
An example of this in action is:
Goal or Role:
"I need to streamline GTM workflows by using custom GPTs to automate repetitive tasks, improve team efficiency, and enhance decision-making."
Rules:
Focus on workflows across sales, marketing, customer success, and operations.
Provide actionable steps that are practical and easy to implement.
The GPT should output tailored processes based on specific goals like lead qualification, content personalization, or competitive analysis.
Outputs should be detailed and ready to use, including necessary data inputs, steps, and expected results.
Ensure the GPT is adaptable for small teams or large-scale operations.
Advanced Context:
"The user works at a mid-sized SaaS company with a GTM team of 15 people. The main challenges include inconsistent ICP creation, lengthy outbound sequences, and limited bandwidth for account research. The team relies on a CRM and has access to tools like Zapier and Google Sheets. They want to prioritize workflows that reduce manual work, improve response quality, and increase pipeline velocity. Examples of desired outputs include automated ICP templates, tailored email sequences, and discovery question lists based on account data."
Clear Instructions:
"Create an actionable guide for automating GTM workflows with custom GPTs. For each example workflow, provide:
Examples
1. Workflow Name: ICP and Persona Creation
Inputs Needed:
Customer data (CRM records, firmographics, buyer personas)
Existing ICP templates
Steps for GPT Setup:
Train GPT using uploaded CRM data and ICP templates.
Write a prompt to extract relevant customer insights.
Configure GPT to output ICPs with structured fields like company size, industry, challenges, and goals.
4. Add Knowledge
Uploading a document to a custom GPT mirrors the workings of a Retrieval-Augmented Generation (RAG) system. Here's a breakdown of how this process works and why it's similar:
1. Document Ingestion
When you upload a document (e.g., PDF, Excel, or other formats) to a custom GPT, it becomes part of the knowledge base the GPT references. This step is akin to the retrieval component in a RAG system, where external knowledge is indexed for easy access.
In RAG: Documents are pre-processed and indexed into a vector database, enabling efficient retrieval based on similarity to a query.
In Custom GPT: The uploaded document is tokenized, stored, and linked to the GPT for use in generating responses. Instead of a separate vector database, this knowledge is typically embedded directly into the GPT’s framework for structured access.
2. Query Matching
When you ask the custom GPT a question, it searches the uploaded document for relevant context. This process resembles how RAG systems retrieve chunks of text from indexed documents based on semantic similarity.
In RAG: The system searches for relevant passages from the document using vector similarity (e.g., cosine similarity in embeddings).
In Custom GPT: The GPT compares the query against the document content, identifying relevant sections to include in its response.
3. Response Generation
Once the GPT identifies the relevant content from the uploaded document, it integrates that information into the generated response. This is similar to how the generation component in a RAG system works.
In RAG: Retrieved passages are fed as context into a language model (e.g., GPT-4) to generate a coherent, context-aware response.
In Custom GPT: The extracted information is used as direct context for the response, maintaining relevance and grounding the output in the uploaded document.
Key Features Shared with RAG
Grounding Responses in External Data
Both methods ensure that the AI does not "hallucinate" answers by anchoring its responses in the document's factual content.
Dynamic Knowledge Updates
Just as a RAG system can refresh its database for new content, custom GPTs can use newly uploaded documents to update their knowledge base without retraining the entire model.
Contextual Relevance
Queries are matched with document content, ensuring that responses are directly tied to the uploaded materials.
Efficient Retrieval
Both use methods to retrieve the most relevant chunks of data, minimizing irrelevant or extraneous information in the final output.
Differences Between Custom GPT and RAG
Infrastructure
RAG often involves separate tools for retrieval (e.g., Pinecone for vector search) and generation (e.g., GPT-4). Custom GPTs handle both internally.
Scalability
RAG systems are designed for massive datasets and can handle millions of documents. Custom GPTs typically handle smaller-scale document uploads.
Customization
Custom GPTs streamline the process for non-technical users, while RAG often requires setting up pipelines, databases, and embeddings.
Practical Use Case
Example:
You upload a company FAQ document to a custom GPT.
Query: “What’s the refund policy for premium users?”
Custom GPT: It retrieves the section from the document containing the refund policy and generates an answer grounded in that content.
In a RAG system, the FAQ would be indexed in a vector database, and the query would retrieve relevant sections from the database for the language model to generate the same grounded response.
Uploading documents to a custom GPT effectively mimics a RAG system by combining retrieval (finding relevant data) with generation (creating coherent outputs). It simplifies the process while maintaining the benefits of dynamically grounded, accurate, and context-aware responses.
5. Build It
Go to the GPT builder.
Start with the “Create” option.
Write a long-form prompt that guides the GPT’s behavior.
Enable any features you need, like web search or integrations.
6. Test and Adjust
Run a quick test. Ask the GPT to do a task. Does it give the right output? If not, tweak the prompt or add more context. Keep refining until it’s on point.
7. Use It for Tasks
Here’s how I use custom GPTs:
GTM Strategy: Create ICPs, personas, or workflows.
Research: Find key account details or executives.
Email Sequences: Build personalized sequences with subject lines and even video scripts.
Call Prep: Generate tailored discovery questions based on personas or accounts.
Competitive Analysis: Develop battle cards with SWOT analyses and objection handling.
8. Automate Actions
Want it to pull from live data or external apps? Connect it to tools like Zapier, Make or APIs. This lets it access CRMs, spreadsheets, or other resources.
Coach K will be showing a bit more of this in the coming year.
9. Practice and Refine
If you’re preparing for a meeting, use roleplay GPTs to simulate the call. Practice with voice or text. Use it to handle objections, test scripts, or refine your delivery.
10. Rinse and Repeat
Keep improving. Add new data, update prompts, and explore additional use cases. The more you iterate, the better the results.
PROMPTS
I had a request from a subscriber to make a prompt on scoring or analyzing calls with Command the Message framework in mind.
For you:
Here’s a detailed prompt based on your request, incorporating the Force Management Value-Based Conversation structure and the context you provided. Both of these below use the GRACE framework:
Goal or Role:
"I need an analysis of a sales call transcript to evaluate how effectively the sales representative engages in a Value-Based Conversation framework, identifying strengths and opportunities for improvement."
Rules:
Focus Areas:
Evaluate the sales representative’s ability to explore the Before State, Negative Consequences, After State, Positive Business Outcomes (PBOs), Required Capabilities, Decision Criteria, Metrics, How We Do It, How We Do It Better, and Proof Points.
Specifically assess whether the rep explores quantitative impacts of pain points and negative consequences when the prospect discusses challenges.
Timestamping:
Highlight and timestamp areas in the transcript where:
Pain points are mentioned but not probed further.
The rep misses an opportunity to quantify or explore the negative consequences.
Positive opportunities (e.g., future state benefits) are identified but not expanded.
Value Alignment:
Ensure the analysis includes mapping conversations to key value drivers like Accelerate Financial Close, Control Dynamic & Complex Financial Operations, Enable Data-Driven Insights & Decisions, or Ensure & Support Regulatory Compliance.
Conversation Evaluation:
Use the Differentiators (e.g., Auditable & Traceable, No Code, Unified Platform, etc.) to assess whether the rep leverages unique selling points effectively.
Advanced Context:
"The transcript involves a sales representative pitching to a prospect in the financial operations industry. The prospect's pain points include inefficiencies in manual reconciliation processes, regulatory compliance challenges, and a lack of actionable data insights. The goal is to identify where the rep could improve by diving deeper into these pain points, quantifying their impact, and aligning solutions to the company's value drivers and differentiators."
Clear Instructions:
"Please analyze the provided transcript using the Value-Based Conversation framework. Provide the following deliverables:
A summary of where the rep excelled or missed opportunities in addressing:
Before State, Negative Consequences, After State, and PBOs.
Decision Criteria, Required Capabilities, Metrics, and Proof Points.
Timestamped excerpts of the transcript for missed opportunities to:
Probe deeper into pain points and negative consequences.
Quantify the business impact of challenges.
Recommendations for how the rep could improve in future conversations, using relevant value drivers and differentiators.
A mapped evaluation of the conversation's alignment with the company's core value drivers and differentiators."
Example Output Section:
1. Observation Category:
(What aspect of the conversation is being analyzed, e.g., missed opportunities, strengths, or overall alignment)
Summary of Observations:
(Provide a high-level summary of the key observations for this category, e.g., where the rep excelled or missed opportunities.)
2. Timestamped Excerpts with Analysis:
Timestamp: [HH:MM:SS]
Transcript Excerpt:
(Copy a brief, relevant excerpt from the transcript where the observation occurs.)
Analysis of Rep Response:
(Describe how the rep handled the situation, what was done well or missed, and why it matters.)
Missed Opportunity (if applicable):(List specific missed questions or follow-ups that could have deepened the conversation.)
Repeat the structure for additional timestamps:
Timestamp: [HH:MM:SS]
Transcript Excerpt:
(Copy another relevant excerpt.)
Analysis of Rep Response:
(Analysis of how the rep engaged.)
Missed Opportunity (if applicable):(Identify unasked questions or areas to probe.)
3. Mapping to Framework:
(Align the observation to key aspects of the Value-Based Conversation framework.)
Framework Component Observed Alignment Improvement Opportunity Before State (Describe how the rep explored the current state.) (Suggest how this could be better explored.) Negative Consequences (Was the rep probing deep enough?) (Identify areas for probing.) After State / PBOs (Describe how the rep positioned the future state.) (Suggest more compelling positioning.) Required Capabilities (Highlight any gaps in identifying what is needed.) (Provide examples of missed questions.) Proof Points (Evaluate whether these were effectively used.) (Recommend specific proof points.)
4. Recommendations for Improvement:
(Provide actionable suggestions tailored to the observations.)
Probing Questions:
(List key questions the rep could ask to explore negative consequences, quantify impacts, or deepen understanding.)Positioning Value Drivers:
(Suggest how the rep can align responses more effectively with specific value drivers, e.g., Accelerate Financial Close.)Using Differentiators:
(Recommend leveraging differentiators like Robust Yet Flexible or No Code for specific client pain points.)
5. Conclusion and Next Steps:
(Summarize the overall performance and key actions for improvement in the next conversation.)
Another prompt for MEDDPICC
Prompt for MEDDPICC-Based Sales Call Analysis
Goal or Role:
"I need an analysis of a sales call transcript to evaluate how effectively the sales representative adhered to the MEDDPICC framework, identifying areas of strength and improvement."
Rules:
Focus Areas:
Assess how well the sales representative navigated the MEDDPICC framework components:
Metrics, Economic Buyer, Decision Criteria, Decision Process, Paper Process, Identified Pain, Champion, and Competition.
Highlight opportunities to better uncover and align with the prospect’s key drivers and decision-making processes.
Timestamping:
Highlight and timestamp areas in the transcript where:
The rep excelled in addressing MEDDPICC components.
The rep missed opportunities to uncover critical information (e.g., decision criteria, metrics, or identifying the economic buyer).
Quantification:
Evaluate how effectively the rep quantified the prospect’s pain points and aligned solutions to measurable outcomes (Metrics).
Champion and Competition:
Analyze the rep’s success in identifying and validating a Champion and addressing potential Competition.
Advanced Context:
"The transcript involves a sales representative pitching a SaaS solution to a prospect in a competitive market. The rep must uncover and align with the prospect’s decision-making process while addressing pain points and differentiating against competitors. Key objectives include securing a Champion, understanding the Economic Buyer, and navigating complex Decision and Paper Processes."
Clear Instructions:
"Please analyze the provided transcript using the MEDDPICC framework. Provide the following deliverables:
A summary of how well the sales rep addressed each component of MEDDPICC:
Metrics, Economic Buyer, Decision Criteria, Decision Process, Paper Process, Identified Pain, Champion, and Competition.
Timestamped excerpts for:
Strong alignment with MEDDPICC components.
Missed opportunities to explore critical information or validate assumptions.
Recommendations for improvement in:
Navigating complex decision processes.
Uncovering and quantifying pain points (Metrics).
Identifying and enabling a Champion.
Addressing competitive differentiators.
A final evaluation of the conversation’s alignment with the MEDDPICC framework and specific next steps for improvement."
Example Output: MEDDPICC Analysis
Observation Category: Metrics
Summary of Observations:
The sales representative successfully identified high-level pain points but failed to quantify them into actionable and measurable metrics. The lack of specific, tangible metrics limited the ability to build urgency and align the solution with the prospect’s business goals.
Timestamped Excerpt:
Timestamp: 12:30
Transcript Excerpt:
Prospect: "Our manual processes are slowing us down, and we’re not hitting reporting deadlines consistently."
Rep Response: "Our platform automates those processes, so you don’t have to worry about missed deadlines anymore."
Analysis:The rep acknowledged the pain but missed an opportunity to quantify its impact by asking:
"How many reporting deadlines are being missed each month?"
"What financial or reputational impact does this have on your business?"
"How do these delays affect your team’s ability to focus on strategic initiatives?"
Improvement Opportunity: The rep could have quantified the impact of these inefficiencies to create a compelling reason to act.
Observation Category: Economic Buyer
Summary of Observations:
The rep failed to confirm the presence of the Economic Buyer or validate the authority of the current contact. This leaves a gap in understanding who holds the budget and decision-making power.
Timestamped Excerpt:
Timestamp: 18:45
Transcript Excerpt:
Prospect: "I’ll need to discuss this further with my team before making any decisions."
Rep Response: "Of course, I can provide additional details for your team."
Analysis:The rep did not probe to identify the Economic Buyer by asking:
"Who else will be involved in the decision-making process?"
"Is there someone on your team who will make the final decision on budget approval?"
Improvement Opportunity: The rep could have navigated this more effectively by securing next steps with the Economic Buyer involved.
Observation Category: Identified Pain
Summary of Observations:
The rep successfully surfaced key pain points but did not explore their broader implications or the emotional impact on the prospect’s team and business.
Timestamped Excerpt:
Timestamp: 25:00
Transcript Excerpt:
Prospect: "We’ve had instances where errors in our reports led to compliance issues."
Rep Response: "Our solution ensures compliance with automated checks and balances."
Analysis:The rep acknowledged the pain but could have asked:
"What were the consequences of these compliance issues?"
"Did these errors result in financial penalties or damage to your reputation?"
"How much time and effort does your team spend fixing these errors?"
Improvement Opportunity: By exploring these areas, the rep could have positioned the solution as not just a fix but a critical enabler of business continuity.
Recommendations for Improvement:
Quantify Pain Points (Metrics):
Use probing questions to uncover specific numbers tied to the prospect’s challenges (e.g., lost revenue, hours spent, financial penalties).
Validate the Economic Buyer:
Always confirm who holds budget authority and set next steps to involve them directly.
Deepen Pain Exploration:
Go beyond surface-level pain points to understand broader consequences, emotional impacts, and urgency.
Champion Development:
Use the identified pain points to begin enabling a Champion who can advocate internally for the solution.