A vendor called Kushal Sharma to tell him his $18,000 contract was now a $28,000 contract. That is normally the part of the story where the ops leader begrudgingly signs. He didn't have to — and the reason why had nothing to do with negotiation. Kushal leads the internal AI organization at Circle, a function he got to build because he had already spent two years merging Circle's data and revenue operations teams under one mandate: turn the ops people into product owners, turn the analysts into engineers, and put real architecture underneath the go-to-market motion. Medallion layers. A dbt semantic layer. Orchestration he controlled. Lineage on everything. Before Circle, he spent eight and a half years at Hootsuite, where he led the company's data organization. Then AI actually arrived — skills and agents shipped — and he was one of the very few operators in the market with something for them to operate on. Because this one goes deeper than usual, LeanScale CTO Jake Toepel joins Anthony Enrico to go layer by layer: what a semantic layer actually is, what a context graph actually does, when a vector database is worth buying, and why almost none of this is an LLM. WHAT YOU'LL LEARN How owning your orchestration layer quietly removes a vendor's pricing leverage before they ever call What a context graph does, why RAG existed before it, and why an LLM is not a store of knowledge How Circle built an AI SDR that generated close to seven figures, and what actually made it work CHAPTERS 00:00 — Cold open + intro 02:11 — The vendor call: an $18K contract becomes $28K 05:07 — Owning the data layer means owning your leverage 06:50 — What a semantic layer actually is 12:48 — The litmus test — if a human can't read it, neither can your AI 19:47 — Medallion architecture: bronze, silver, gold 26:17 — Context graphs from first principles 35:19 — "None of this is an LLM" — the software underneath 40:12 — The maturity ladder: when to actually buy a vector database 49:09 — Stop throwing your best model at a simple problem 51:33 — What it unlocks: a Slack bot in 10 minutes, and an AI SDR 1:04:27 — Relevance or support center: the choice facing RevOps 1:07:09 — Wrap ABOUT THE GUEST Kushal Sharma leads the internal AI organization at Circle, the community platform used by more than 15,000 communities. He joined Circle in 2024 to run a combined data and revenue operations function and spent two years putting real architecture underneath the go-to-market motion: medallion layers, a dbt semantic layer, orchestration the team controls outright, and lineage on everything. That foundation is what let Circle stand up an AI function when agents and skills actually shipped. Before Circle, Kushal spent eight and a half years at Hootsuite, where he led the company's data organization. ABOUT THE LEANSCALE PODCAST The LeanScale Podcast is the show for GTM operators building the next generation of revenue infrastructure. Hosted by Anthony Enrico, co-founder of LeanScale, we go deep with the operators, founders, and executives shaping how modern go-to-market teams are designed, scaled, and transformed. ▶ Subscribe for new episodes ▶ Learn more about LeanScale: https://www.leanscale.team ▶ Follow Anthony on LinkedIn ▶ Follow Jake Toepel on LinkedIn ▶ Follow Kushal on LinkedIn KEYWORDS / TAGS RevOps, Revenue Operations, GTM, GTM Engineering, AI in RevOps, AI Agents, AI Infrastructure, Semantic Layer, Context Graph, Vector Database, RAG, Medallion Architecture, Data Lineage, Data Warehouse, dbt, Analytics Engineering, Data Orchestration, LLM, Model Selection, AI SDR, Sales Operations, Salesforce, Snowflake, Circle, Hootsuite, B2B SaaS, RevOps Leadership, Data Org, LeanScale, Anthony Enrico, Jake Toepel, Kushal Sharma HASHTAGS #RevOps #GTM #AI #AIAgents #RevenueOperations #DataEngineering #AnalyticsEngineering #SemanticLayer #ContextGraph #VectorDatabase #dbt #B2BSaaS #AITransformation #LeanScale #LeanScalePodcast