LLM
Call a large language model directly from a formula. Ask any question, transform any text, generate any output. The Swiss Army knife of AI in Excel — but expensive per-call so use strategically.
LLM Complete Guide →The newest generation of Excel functions. Native LLM calls, Python execution, and classification/extraction functions that turn Excel into an AI-first analytics tool.
📌 Rollout notice: AI functions are being rolled out gradually across Excel 365 channels through 2025-2026. Some are still labeled "Preview" — they work but syntax may evolve. Feature availability depends on your Microsoft 365 subscription tier and rollout channel.
AI functions represent the biggest shift in Excel formulas since dynamic arrays. Instead of you writing precise logic, you describe what you want in natural language and let the AI handle the interpretation. Classify text as positive/negative sentiment. Extract structured data from messy strings. Summarize long text into short takeaways. Run Python code with pandas or scikit-learn inside a formula.
Call a large language model directly from a formula. Ask any question, transform any text, generate any output. The Swiss Army knife of AI in Excel — but expensive per-call so use strategically.
LLM Complete Guide →Run Python code inside a formula. Full pandas, scikit-learn, matplotlib, seaborn available. Cloud-executed for security. The most powerful analytical addition to Excel ever.
PY Complete Guide →Purpose-built AI functions for common tasks. Classify text into buckets (sentiment, priority, category). Extract structured data (emails, phone numbers, entities) from unstructured text.
AI.CLASSIFY Guide →Condense long text into short takeaways. Translate cells between languages. Purpose-built for common transformations — faster and cheaper than a generic LLM call.
AI.SUMMARIZE Guide →These functions are just one of five ways to use AI in Excel. The AI in Excel hub covers Claude via our Add-in, Microsoft Copilot, ChatGPT, Python in Excel, and custom LAMBDA+API formulas — with honest comparisons of when to use each.
=AI.CLASSIFY(A1, "positive, negative, neutral") — categorize customer reviews in one column=AI.EXTRACT(A1, "email address") — pull email from messy contact text=AI.SUMMARIZE(A1, 20) — 20-word summary of long support tickets=AI.TRANSLATE(A1, "Spanish") — translate product descriptions=LLM("Categorize this transaction as one of: food, transport, entertainment, other. Transaction: " & A1)=PY("import scipy.stats as s; s.ttest_ind(xl('A2:A100'), xl('B2:B100'))") — real statistical tests inside a cellYes for some. LLM and AI.* functions require Microsoft 365 Copilot ($30/user/month). PY (Python in Excel) is included in most M365 plans but has a per-execution cloud compute limit. Free tier is usually just enough to try them out.
Our Add-in gives you Claude-powered formula generation and AI chat at $29 one-time — no per-formula charges. Microsoft's built-in AI functions charge per API call within your subscription. Different pricing models for different use cases. Read the full comparison.
Yes. AI functions call Microsoft's Azure OpenAI service. Data policies match your Microsoft 365 enterprise data protection settings. Check your organization's approved AI usage policy before using these functions on regulated data.
No. All AI functions require an internet connection. This includes PY (Python is executed in Microsoft's cloud, not locally).
Included in your subscription with per-user monthly limits. Heavy users may hit limits and need to upgrade tiers. Cost per call is not directly billed but subscription costs scale with expected usage. Check Microsoft's current pricing for exact figures.
Try Claude-powered formula generation and chat inside Excel. Free tier available. No enterprise subscription required.
Get the Add-in →