# Using AI Enrichment in TexAu AI enrichment in TexAu lets you enrich your data without endless manual work. Whether you're generating personalized outreach messages, summarizing company information, classifying leads, or scoring data quality, TexAu's AI integration makes it simple. This guide covers everything you need to know about using AI to enrich your data. ## What Can You Do With AI Enrichment? AI enrichment opens up a world of possibilities. Here are some common use cases: **Personalized Messaging**: Generate custom email openers or LinkedIn connection requests tailored to each prospect. **Lead Scoring**: Ask AI to score leads on quality, fit, or readiness to purchase based on their profile data. **Company Research Summaries**: Turn raw company information into concise, actionable summaries for your team. **Data Classification**: Automatically categorize leads by industry, company size, or other attributes. **Insight Extraction**: Pull key information from messy text fields or company descriptions. **Sentiment Analysis**: Understand sentiment from customer reviews, feedback, or social posts. The possibilities are truly endless once you get comfortable with the workflow. ## How AI Enrichment Works in TexAu Here's the basic flow: 1. You create an AI column in your table 2. You write a prompt that tells AI what to do 3. You can reference other column values in your prompt (like {{firstName}} or {{companyName}}) 4. TexAu sends your data + prompt to the AI model you've chosen 5. The AI processes it and returns results 6. Those results appear in your table Once you set it up, it runs automatically on new rows or whenever you trigger it. ## Setting Up an AI Column Let's walk through creating your first AI enrichment column. ### Step 1: Create a New Column In your table, click the **+ Add Column** button. Choose **AI Enrichment** as the column type. ### Step 2: Choose Your AI Model TexAu supports three AI providers, each with different models: **OpenAI**: GPT-4, GPT-4 Turbo, and GPT-3.5 Turbo. GPT-4 is the most capable but also the most expensive. Great for complex reasoning, detailed writing, and nuanced analysis. **Anthropic (Claude)**: Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku. Claude models excel at nuanced writing, long-form content, and detailed analysis. Opus is the most capable; Haiku is the fastest and most affordable. **Google Gemini**: Gemini Pro is Google's multimodal model. Good all-around performance with competitive pricing. > **Tip**: If you're not sure which model to start with, Claude Sonnet or GPT-4 Turbo are excellent middle-ground choices. They balance quality, speed, and cost well. Your available models depend on your TexAu plan. Premium plans unlock access to the most advanced models. Standard plans get access to solid workhorse models like GPT-3.5 Turbo and Claude Haiku. ### Step 3: Write Your Prompt Your prompt tells the AI exactly what to do with each row of data. Give your AI column a clear name (e.g., "Personalized Email Opener" or "Lead Quality Score"). Then write your prompt in the prompt editor. Here's an example: ``` Write a personalized 2-sentence LinkedIn connection request for {{firstName}} who works at {{companyName}} as a {{jobTitle}}. Make it friendly and mention something specific about their company. ``` Notice how we're using {{columnName}} syntax to reference other columns? That's how you make your prompts dynamic and personal to each row. ### Step 4: Test Your Prompt Before running on all your data, test your prompt with a few sample rows. This helps you see if the results are what you're looking for. If not, refine your prompt and try again. ### Step 5: Run Your Enrichment Once you're happy with the results, you can run the enrichment on your entire table or just new rows going forward. TexAu will process each row through your AI model and populate the column with results. ## Choosing the Right AI Model With three different providers and multiple models, you might wonder which one to choose. Here's a breakdown: ### When to Use OpenAI (GPT-4 / GPT-3.5 Turbo) **Best for**: Structured data extraction, complex reasoning, technical writing. **Trade-off**: Can be pricier than alternatives. GPT-4 is slower but more capable. GPT-3.5 Turbo is faster and more affordable. Use GPT-4 when you need the absolute best results on complex tasks. Use GPT-3.5 Turbo when you want solid quality at a good price point. ### When to Use Anthropic Claude (Claude 3 Opus / Sonnet / Haiku) **Best for**: Long-form content, nuanced writing, personalized messaging, creative tasks. **Trade-off**: Slower than some alternatives. Claude Haiku is very affordable. Claude shines when you want thoughtful, well-written outputs. If you're generating outreach messages or detailed summaries, Claude often produces more natural-sounding results. Haiku is surprisingly capable for its price, making it great for bulk enrichment on a budget. ### When to Use Google Gemini **Best for**: General-purpose tasks, good balance of speed and quality, competitive pricing. **Trade-off**: Slightly less specialized than OpenAI or Claude for certain tasks. Gemini is an excellent all-rounder. If you're not sure which model to pick and want a solid, cost-effective option, Gemini is a great choice. > **Pro tip**: Try different models on small batches of data. You'll quickly develop a sense for which one produces results you like best. Your needs and preferences matter most. ## Writing Effective Prompts Good prompts are the key to great AI enrichment. See the [Writing Effective Prompts](./writing-effective-prompts.md) guide for a full walkthrough, but here are the essentials: **Be Specific**: Instead of "write a message," try "write a friendly, 3-sentence LinkedIn connection request that mentions their recent company news." **Provide Context**: The more your prompt tells AI about the situation, the better the results. **Use Column Variables**: Reference {{firstName}}, {{companyName}}, {{jobTitle}}, etc., to make your enrichment personal and relevant to each row. **Set Output Format Expectations**: If you want a number, say "respond with only a number from 1-10." If you want JSON, ask for it. Clear formatting = cleaner results. **Keep It Conversational**: Write prompts in plain English. You don't need fancy phrasing. "List three talking points" works great. ## Understanding AI Credit Costs Every time you run an AI enrichment, TexAu consumes AI credits from your plan's monthly allocation. The cost depends on a few factors: **Model Choice**: Advanced models (like GPT-4 or Claude Opus) cost more than budget models (like GPT-3.5 Turbo or Claude Haiku). **Prompt + Data Size**: Longer prompts and more input data = higher cost. Typically, enriching a single row costs anywhere from a fraction of a cent to a few cents, depending on your model. **Output Length**: If you ask AI to write a long essay, it costs more than asking for a one-word classification. When you create an AI column, TexAu estimates the cost per row and per 1,000 rows. You can see these estimates before you start, so there are no surprises. > **Budget tip**: If you're enriching thousands of rows, start with a budget model like Claude Haiku or GPT-3.5 Turbo. You'll stretch your credits much further. ## Tips for Getting Better AI Results Even with the same prompt, tweaks and iterations can significantly improve results. Here are some best practices: ### 1. Be Explicit About What You Want Instead of: "Summarize this company." Try: "Write a 2-3 sentence summary of what {{company}} does, focusing on their primary product and target customer." Specificity wins every time. ### 2. Use Examples in Your Prompt Showing AI what good output looks like helps it match your style: ``` Generate a lead quality score from 1-10 for {{firstName}}. Use this scale: - 1-3: Not a fit (wrong industry, too small, etc.) - 4-6: Possible fit (could be relevant) - 7-10: Strong fit (high priority to contact) Answer with only the number. ``` ### 3. Test Before Running at Scale Always test your prompt on 5-10 sample rows first. Spot-check the results. It takes 30 seconds and saves you credits and time. ### 4. Iterate Based on Results If results aren't quite right, adjust your prompt. Maybe be more specific, add examples, or ask for a different output format. Small tweaks often lead to big improvements. ### 5. Reference Relevant Columns If you're enriching with data from multiple columns (like {{firstName}}, {{company}}, {{lastInteraction}}), include all relevant context. More context = better, more personalized results. ### 6. Keep It Simple You don't need a novel-length prompt. Clear, concise instructions usually work best. If your prompt is taking forever to write, simplify it. ## What's Next? Ready to write your first AI enrichment prompt? Head over to the [Writing Effective Prompts](./writing-effective-prompts.md) guide for detailed use cases and prompt templates. Or, jump straight to setting up your first AI column in TexAu.