$ man clay-wiki/sculptor-guide

Most people think Sculptor is Clay's deduplication tool. It can dedupe, but that's like saying a smartphone is a calculator. Sculptor is an AI interface to query your table. You can ask it questions about your data in natural language and it returns answers. "Show me all contacts with VP titles at companies with more than 500 employees." "Which rows have emails from Google Workspace domains?" "Group these companies by industry and count them." It's the closest thing Clay has to a conversational data layer.

PATTERN

Sculptor is best for: (1) Quick data exploration — when you import a list and want to understand what you're working with before building enrichment columns. (2) Fuzzy deduplication — grouping "Microsoft," "Microsoft Corp," and "MSFT" without writing complex formulas. (3) Data quality checks — "How many rows have empty email fields?" or "Show me rows where the domain doesn't match the company name." (4) Ad-hoc analysis — when a partner asks "how many enterprise accounts are in this list?" and you need a fast answer without building a whole scoring system. It's low-effort. It's conversational. It didn't exist when I started using Clay, so I built everything with formulas and columns. But for beginners, Sculptor is a great way to start thinking about your data before you commit to column architecture.

PATTERN

Yes, Sculptor does dedupe well. Here's the pattern: (1) Import your list. (2) Open Sculptor and ask it to find duplicates by company name or domain. (3) It groups fuzzy matches — catches abbreviations, typos, parent/subsidiary relationships. (4) Review the groups and merge or delete. This is better than formula-based dedupe for messy datasets. Formulas need exact rules. Sculptor uses AI to catch what rules miss. For clean datasets with consistent formatting, formulas are faster and free. For messy imports from third-party sources, Sculptor saves hours.

ANTI-PATTERN

Sculptor isn't a replacement for proper column architecture. It's great for exploration and one-off queries, but it doesn't create persistent, reusable logic. If you need a classification that runs on every new row, build a formula column. If you need enrichment, use enrichment providers. If you need scoring, use the scoring integration. Sculptor is for answering questions about data you already have — not for transforming or enriching it. It also doesn't scale well to very large tables (50K+ rows). For big TAMs, use Supabase for storage and Clay for orchestration. Don't try to query a 100K-row table through Sculptor.

People confuse these. Sculptor queries data already in your table. Claygent browses the web and brings new data in. Sculptor: "Which of my existing contacts are at companies with Microsoft MX records?" Claygent: "Go to this company's website and summarize their positioning." Sculptor is free or low-credit. Claygent costs credits per row. Use Sculptor to understand what you have. Use Claygent to fill in what you don't.

PRO TIP

Sculptor is a good tool, not a great one. It's useful for beginners who don't yet think in columns and formulas. It's useful for quick questions when you don't want to build a whole enrichment flow. But I don't use it daily. My tables are architected with dedicated columns for every classification, score, and filter. By the time I'd open Sculptor, I already have the answer in a column. If you're just starting with Clay, use Sculptor to explore. Then graduate to formulas and column architecture. That's where the real power lives.