Creating Effective Prompts for Analyzing Energy Company Insights

We are trying to scan a multitude of different energy companies to develop insights for personalisation some are public, private or government owned so they have different areas where the insights are kept. most will have an about us section, details on sustainability or innovation initiatives, but many will have specific reports to download e.g. for public companies investor relations an annual report as a pdf, or for most companies a sustainability report as a pdf etc ... how can we create a clay prompt that analyse the company and holistically gathers up the information we care about based on keywords be it web content or documents? Whats the best practise to create prompts that cover different use cases to summarise the data accurately and make it comprehensive and efficient with credits?

15 comments

Channeled
APP
·
·
To create effective Clay prompts for company analysis, follow these key principles: 1. Set Clear Context and Structure: • Start with a system prompt that outlines the goal and context of the analysis • Provide step-by-step instructions for the AI to follow • Specify desired output format and structure 2. Organize Your Analysis Approach: • Information Summarization: Use prompts to distill complex company data into digestible insights • Information Extraction: Create prompts to pull specific data points like company mission, target markets, and initiatives • Content Analysis: Set up prompts to analyze various content types including web pages, PDFs, and reports 3. Configure Keywords and Categories: • Define specific categories you want to analyze • Create lists of relevant keywords for each category • Set up the output schema to specify exactly which fields you want to extract 4. Leverage Clay's AI Capabilities: • Use Clay to scrape and analyze PDFs and documents • Apply AI to summarize content across multiple sources • Extract targeted information without manual intervention Sources: We Made 24 Prompts So You Don’t Have To [Automated Outbound] - Clay University Format Categories Based on Keywords with Clay Prompt Engineering Crash Course [Clay 101] - Clay University Make the entire internet your research canvas | Clay.com
Mike B.
·
·
Use Clay to scrape and analyze PDFs and documents ... how detailed do we have to be to specify the type of document or can we create a generic prompt so clay scans all the documents available for download and then pulls relevant insights based on our goals and keywords? How do we make sure in the prompt we know where the information originated from?
Channeled
APP
·
·
Here's how to create an effective generic prompt for document analysis in Clay: 1. Start with a clear system prompt that: • Sets the context and goal of your analysis • Defines the types of insights you're looking for • Specifies the document types to analyze 2. Structure your main prompt with: • Step-by-step instructions for the AI to follow • Specific categories and keywords to look for • Required output format 3. Use the "Use AI" integration where you can: • Write your prompt manually or use the "Help me" feature • Include dynamic variables with "/" to personalize the analysis • Define specific output columns to organize the extracted informationClaygent Neon is particularly effective for this task as it excels at: ** Parsing PDFs and hard-to-find information ** Providing higher accuracy in responses * Delivering reliable and flexible answer formatting Sources: Use AI integration overview | Documentation | Clay University Prompt Engineering Crash Course [Clay 101] - Clay University Prompt Engineering Crash Course [Limitless Research] - Clay University Introducing Claygent Neon
Mike B.
·
·
Do we need the Claygent to find all pdfs on the target website so it can scrape the name to provide a specified document type for the AI to analyse? How do we manage this across multtiple countries where they may be named differently e.g. strategy report, annual report, sustainability report, innovation reprot etc ... and then enable the AI to then review the pdf
Channeled
APP
·
·
To effectively use Claygent for finding and analyzing PDFs across multiple countries, here's how to structure your approach: 1. Start with a clear system prompt that: • Sets the context for finding company reports • Lists all possible document variations (strategy, annual, sustainability reports, etc.) • Specifies the type of insights you want to extract 2. Use Claygent's preset templates as a foundation: • Navigate to the enrichment panel in your Clay workspace • Look under the AI section for Claygent • Modify the existing templates for document search 3. Structure your prompt to: • First locate and identify all downloadable PDFs on the company website • Classify documents based on content rather than just titles • Extract relevant information based on your specified categoriesRemember to specify the output format clearly in your prompt to ensure consistent data extraction across different document types and languages. Sources: Claygent: AI Web Scraper [Automated Outbound] - Clay University Claygent: AI Web Scraper [CRM Enrichment] - Clay University Prompt Engineering Crash Course [Clay 101] - Clay University
Mike B.
·
·
Can we create the prompt in english and then you translate to find relevant documents linked to sustainability, or do we need local language keywords for this to work?
Channeled
APP
·
·
You can create your prompt in English and use Clay's DeepL Translate integration to handle multilingual document search effectively. Here's how it works: 1. Create your sustainability-related prompt in English with your keywords and search criteria. 2. Use Clay's categorization feature to set up your sustainability keywords and categories in English. 3. The system can identify sustainability-related content across different document types, as demonstrated in the example where it successfully found sustainability pages and reports even with varying naming conventions.DeepL's neural machine translation capabilities will help ensure accurate translation and identification of relevant documents across different languages, so you don't need to manually create keyword lists in local languages. Sources: DeepL Translate with DeepL Check a company's investment in sustainability | Clay Categorize Items Based On Keywords with Clay
Mike B.
·
·
What are the limits on the amount of data we can captured as unstructured data related to the information and insights we care about. Currently the clay summary is far to high level. How can we make sure its super detailed, and then we can do a further summary in clay to pull out high level summaries in another column
Channeled
APP
·
·
Here's how to get both detailed and high-level summaries from unstructured data in Clay: 1. Set up your AI integration with multiple output columns: • Use the "Define column outputs" feature to create separate columns for different levels of detail • Specify the data type as "Text" for your summary columns 2. Structure your data extraction: • Create one column for the detailed, comprehensive analysis • Set up additional columns to split components of the output for higher-level summaries • This improved data structuring allows you to segment information effectively 3. Use child columns for organization: • Click on your main enrichment cell to open the Cell details panel • Use "Add as column" for specific endpoints you want to extract • Create new columns for different summary levels by mapping specific data pointsThis approach lets you maintain both detailed analysis and concise summaries in separate, organized columns for better data management. Sources: Use AI integration overview | Documentation | Clay University Introducing Claygent Neon Table columns overview | Documentation | Clay University
Mark L.
·
·
Hey Mike, thanks for getting in touch. I see you've been talking about AI research. I saw it was recommended earlier, but would definitely emphasize the usefulness of the "Help me" feature when adding an AI column. It can help you polish up your prompts. I'll also add that when you add a column there's an option for "Browse templates", too. We have a few examples in there that might give you a headstart, or illustrate what a robust, tightly-instructed prompt can look like. The 10-k one is a good example. https://downloads.intercomcdn.com/i/o/w28k1kwz/1359857968/43259f4e1905e91d719f85ad638a/20250131-help-me-templates.png?expires=1738340100&signature=a417bcdad19d2d0c421006f092eb8ebc4c7d178a992efee2f9712e1e248ce469&req=dSMiH8F7mohZUfMW1HO4zepaMOG9mgJ0yup%2BZAVCU9ENjiZdPbpPk%2FyJV4m7%0AGdfk%0A Cells in Clay will only hold so much data that comes back from enrichments (~200kb), but that's decently large amount for plain text. And you can always break the tasks apart across multiple columns. Let us know of other questions as you explore and experiment.
Mike B.
·
·
Can we create a prompt where the data is pulled from the website at scale by Clay and then summarised into the columns, but that clay then pulls the downloaded PDFs e.g. annual reports, sustainability reports, innovation reprots for us to review elsewhere? or the links to those PDFs for download? Conscious that doing this at scale in our customer data platform in Airtable with Chat GPT may be the most efficient way to review these lengthy documents and pull relevant insights to then enhance what clay has pulled already, then can push as structured content into our Aurtable CRM fields for sales and marketing? LMK Thoughts
Mike B.
·
·
Thanks for helping Mark!
Mark L.
·
·
Sure, definitely, you could ask the AI in Clay to focusing on findin the docs, and generating a list of relevant URLs for the various PDFs that it finds. And then you could pipe the list of URLs to another service for download or archiving, instead of/in addition to doing text analysis in Clay. We do have an Airtable integration, so you could create/upsert records into a database there – great way to maintain an archive/source of truth for the long run. Otherwise, using "HTTP API" columns is the way to integrate with other services where we don't already have a native integration with. https://www.clay.com/university/lesson/http-api-clay-101 https://www.clay.com/university/guide/http-api-integration-overview
Channeled
APP
·
·
Hi Mike B.! This thread was recently closed by our Support team. If you have a moment, please share your feedback:
Channeled
APP
·
·
This thread was picked up by our in-app web widget and will no longer sync to Slack. If you are the original poster, you can continue this conversation by logging into https://app.clay.com and clicking "Support" in the sidebar. If you're not the original poster and require help from support, please post in 02 Support.