Share
Embed
Stop Analyzing, Start Asking. How AI Agents Turn Data into Conversations
Tom Winter
June 13, 2025
Transcript
38 slides
-
Tom Winter CGO & Co-Founder Tom Winter CGO & Co-Founder
Stop Analyzing, Start Asking: How AI Agents Turn Data into Conversations -
160+ calls 3 conferences 3 learnings about SEO data Real
Conversations. Real Problems. -
#1 The Illusion of Analysis Most SEO ‘analysis’ watching charts.
= Almost things change. nobody knows why -
#2 Fragmented Data, Fragmented Insight GSC is all slices and
snapshots. Never the full story. -
#3 Dashboards Can’t Save You When it matters, it’s still
spreadsheets and guesswork. -
Just the Tip of the Iceberg 1,000 rows per export
Snapshots, not history 16 months, then it’s gone Basic filters, no deep segmentation -
You’re Still in Snapshot Land Stuck Apps help, but only
for today’s data Still exporting, slicing, guessing No true history, still stuck in sheets Basic filters, no deep segmentation -
BigQuery: Data Freedom - With a Catch Break free
from GSC’s export limits Possibility to join Analytics with GSC data and more Unlimited storage, no more 16- month cutoff Own your SEO data - on your terms -
BigQuery: Data Freedom - With a Catch But: Data
starts from today No backfill. If you don’t speak SQL, you’re locked out -
Why This Still Sucks for Most SEOs Hours lost digging,
exporting, merging Still no clear answers guessing, Data scientists get insights. -
The Dashboard Delusion Data Studio, Looker, Tableau: More graphs, same
old pain LLMs as search bars: “Summarize my data” - but disconnected You still need to do the hard part yourself -
The SEO Data Workflow Hell GSC export BigQuery Write SQL
Export Visualize Interpret Maybe an answer Every step = another chance for confusion, error, wasted time -
What If You Could Just Ask? “Show me all pages
that lost clicks after the May update” “Find keyword clusters with growing impressions” “Which landing pages lost the most visibility” The system fetches, clusters, and explains. No code, no pain. -
The Agent Revolution How It Works Not a chatbot. Not
a static workflow. An agent is an orchestrator. Thinks, plans, adapts. -
Why Simple Workflows or SQL Aren’t Enough AI workflows =
one path, no thinking SQL queries = fixed instructions, no adaptation Agents , adapt on the fly, and choose the best tools make real decisions -
Agents Agents Make the Decisions ' Every question$ ' Different
tools$ ' Different order$ ' The agent decides, not the user. -
Real-World Questions Only Agents Can Handle “What products are seasonal
based on searches?” “Which keyword clusters lost visibility across my domain in Q1 this 2025?” “What content is trending up that I might be missing?” “What is my best performing topic cluster that shows my topical authority?” -
GSC + BigQuery + Agent Game Changer GSC data in
BigQuery, agent orchestrates the flowR Natural Language to SQL to DataFramE : Transform, filter, aggregatE Human-readable summary No more snapshot syndrome. See the entire story. g sql-agent: g manipulation-agent g explain-agent: -
Example How the Agent Gets You Answers Query: Agent’s steps&
Find keywords semantically similar to 'seo' and their total clicks in the last 30 days g` Converts question to SQR W` Uses BigQuery VECTOR SEARCH for semantic similaritw h` Pulls all matching data - no 1,000-row ceilinI c` Returns DataFramD X` Summarizes results in plain English -
Vector Search Semantic Clustering, Not Keyword Matching Embeddings turn keywords,
pages, clusters into vectors BigQuery VECTOR SEARCH groups by meaning, not just strings Surface new topics, intent, and real trends -
Live Agent Path From Messy Question to Clear Answer Example:
Agent Path6 Show me pages that lost traffic after the May update XW Identifies the May update windos SW Compares clicks before and aftef IW Filters for the biggest drop HW Aggregates, loops, refetches if needed Explains what happened and what to do next -
Use Cases That Actually Matter Diagnose Find declining keyword themes,
Uncover hidden clusters Predict Save time, algorithm hitU not just single keywordU with vector searcV trendU get answers -
Why Agents Beat Every Dashboard No more export and pray
No SQL, no Python, no guesswork Fast, flexible, asks follow-up questions You get answers, not just more data -
The Big Shift: From Analyst to Explorer Old way: Export,
crunch, curse, repeat New way: Ask, explore, iterate, understand Every SEO becomes a data scientist. Without the pain -
Roadmap Where This Is Going More tools: automatic charts, forecasting,
ML models Integrate Analytics, CRM, and more Ask, “What will my traffic look like next quarter?” and get a real answer -
Key Takeaways #1 Stop analyzing, start asking: AI agents let
you query your data, not just export it. -
Key Takeaways #1 Stop analyzing, start asking: #2 Agents don’t
replace experts AI agents let you query your data, not just export it. - they supercharge them: Your know-how still matters. -
Key Takeaways #1 Stop analyzing, start asking: #2 Agents don’t
replace experts #3 Forget manual hell: AI agents let you query your data, not just export it. - they supercharge them: Your know-how still matters. Agents handle the heavy lifting, 10x faster than spreadsheets or SQL -
Key Takeaways #1 Stop analyzing, start asking: #2 Agents don’t
replace experts #3 Forget manual hell: #4 Great setup = great results: AI agents let you query your data, not just export it. - they supercharge them: Your know-how still matters. Agents handle the heavy lifting, 10x faster than spreadsheets or SQL Good data, good questions, smart configuration. -
Key Takeaways #1 Stop analyzing, start asking: #2 Agents don’t
replace experts #3 Forget manual hell: #4 Great setup = great results: #5 This is the next era for SEO: AI agents let you query your data, not just export it. - they supercharge them: Your know-how still matters. Agents handle the heavy lifting, 10x faster than spreadsheets or SQL Good data, good questions, smart configuration. Better questions lead to better answers and real impact. -
Thanks & Let’s Connect Stay in touch: [email protected] Let’s connect
on LinkedIn: Tom Winter Stop analyzing, start asking