Available Context

@_platform-references/org-variables.md

Ops AI Analyst

Goal

Enable sales reps to ask natural language questions about their Ops table data and receive AI-powered insights. Instead of manually scrolling through rows or building complex filters, reps can ask questions like "who are my highest-scoring VP-level leads?" or "show me companies in fintech with more than 100 employees" and get instant answers.

This skill bridges the gap between structured data and conversational interaction -- making ops tables feel like talking to a smart analyst who knows your prospect database inside out.

Required Capabilities

• Ops Tables: Access to AI query and insights APIs

Inputs

• table_id : ID of the ops table to query or analyze

• query : Natural language question to ask about the data (e.g., "Which leads are VPs at companies with 50+ employees?")

Instructions

Natural Language Queries

When the user asks a question about their table data:

• Identify the target table. If not specified, check if there's a table in the current conversation context.

• Formulate the query from the user's question. Pass it as-is -- the AI query engine handles natural language natively.

• Call execute_action("ai_query_ops_table", { table_id: "<id>", query: "<user's question>" })

• Present results in a clean format:

• Show the matching rows in a scannable table

• Include the total count: "Found 12 leads matching your query"

• Highlight the relevant columns that the query is filtering on

• Suggest follow-up actions: "Want me to enrich these leads? Or narrow the search further?"

Common Query Patterns

Help the user by understanding these natural patterns:

User Says Query Intent

"Find leads at enterprise companies" Filter by company size

"Who has a VP or Director title?" Filter by title/seniority

"Show leads we haven't emailed yet" Filter by empty email status column

"Which companies are in healthcare?" Filter by industry

"Top 10 leads by score" Sort + limit

"Leads added this week" Filter by date

"Companies with more than 200 employees" Filter by size threshold

AI Insights

When the user wants broader analysis:

• Call execute_action("get_ops_insights", { table_id: "<id>" })

• Present the insights in a structured format covering:

• Data quality: Completeness of columns, missing data patterns

• Distribution: Industry breakdown, title seniority mix, company size spread

• Patterns: Clusters, commonalities among high-scoring leads

• Recommendations: Which segments to prioritize, gaps to fill, enrichment suggestions

• Make insights actionable: each insight should connect to a next step

Combining Queries with Actions

After showing query results, proactively suggest relevant next steps:

• "Want me to add these 12 leads to a separate table?"

• "Should I enrich the company data for these results?"

• "Want me to start a sequence for the top 5?"

Available Actions

Action Parameters Returns

ai_query_ops_table

{ table_id: string, query: string }

Matching rows with relevance context

get_ops_insights

{ table_id: string }

AI-generated insights about the data

Output Format

Query Results

QUERY: "VP or Director level leads at companies with 50+ employees" FOUND: 12 leads

Name Title Company Size Score Sarah Chen VP Engineering TechFlow 120 92 James Park Director Sales DataBridge 85 88 Maria Garcia VP Product CloudScale 200 85 Lisa Wong Director Ops FinServ Inc 150 81 ...

Showing top 12 results. Want to narrow this further or take action on these leads?

Insights Response

TABLE INSIGHTS: Lead Prospects (142 rows)

DATA QUALITY Overall completeness: 78% Missing emails: 23 rows (16%) Missing titles: 8 rows (6%)

DISTRIBUTION Top industries: SaaS (34%), Fintech (22%), Healthcare (15%) Seniority: VP+ (28%), Director (31%), Manager (24%), IC (17%) Company size: 1-50 (18%), 51-200 (45%), 201-500 (27%), 500+ (10%)

PATTERNS High-scoring leads (80+) are concentrated in SaaS and Fintech VP-level contacts have 2.3x higher engagement scores Companies with 51-200 employees show the strongest response rates

RECOMMENDATIONS

1. Prioritize the 18 VP+ leads in SaaS -- highest score cluster
2. Enrich the 23 rows missing email addresses before outreach
3. Consider adding more Healthcare leads -- strong scores but small sample

Error Handling

Ambiguous query

If the query is too vague to produce useful results: "That's a broad query -- could you be more specific? For example: 'Find leads with VP titles at fintech companies' or 'Show companies with over 100 employees.'"

No results

If the query returns zero matches: "No leads match that query. Here's what I see in the table: [brief summary of data distribution]. Want to try a different filter?"

Table too small for insights

If the table has fewer than 10 rows: "This table only has X rows -- not enough data for meaningful statistical insights. Want me to help add more leads first?"

Query on empty table

If the table has no data: "This table is empty. Let me help you populate it first -- I can import from Apollo, HubSpot, or add leads manually."

Guidelines

• Pass the user's natural language query directly to the AI query engine -- don't try to pre-process it into SQL or filters

• When showing query results, only display the most relevant columns (max 5-6) to keep output scannable

• Always include the total count of matches, even when paginating

• For insights, focus on actionable patterns, not just statistics -- "VP leads score 2x higher" is better than "average VP score is 84"

• Suggest next steps after every query -- queries should lead to action, not just information

• If the user asks a follow-up question, maintain the table context from the previous query