Satish Vutukuru

🎯 𝗖𝗮𝘂𝘀𝗮𝘁𝗶𝗼𝗻 > 𝗖𝗼𝗿𝗿𝗲𝗹𝗮𝘁𝗶𝗼𝗻 (10 and final) 𝗧𝗵𝗲 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗠𝗶𝗿𝗮𝗴𝗲 Everyone is excited about “autonomous marketing.” LLMs writing copy. Agents orchestrating journeys. AI deciding what every customer sees. It feels like the future. Most of it is optimizing inside a hallucination. I call it The Optimization Mirage. 𝗧𝗵𝗲 𝗳𝗶𝗿𝘀𝘁 𝗺𝗶𝘀𝘁𝗮𝗸𝗲: 𝗰𝗼𝗻𝗳𝘂𝘀𝗶𝗻𝗴 𝗱𝗲𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻Churn scores. LTV models. Propensity rankings. These are descriptive models. They contain zero intelligence about what will happen if you intervene. In lifecycle marketing, their role must be narrow: • As surrogate outcomes for experimentation when true LTV is delayed • As context features inside a real decisioning engine They are inputs. They cannot be policies. Using descriptive models as decision engines is like trying to fly a plane by reading yesterday’s weather report. 𝗧𝗵𝗲 𝘀𝗲𝗰𝗼𝗻𝗱 𝗺𝗶𝘀𝘁𝗮𝗸𝗲: 𝗹𝗲𝘁𝘁𝗶𝗻𝗴 𝗟𝗟𝗠𝘀 “𝗱𝗲𝗰𝗶𝗱𝗲” The natural extension! An LLM can: • Generate treatments • Embed context • Summarize history It can tell you what usually happens together. But it cannot reason about why a treatment caused which outcome. Correlation does not translate to policy. 𝗧𝗵𝗲 𝘁𝗵𝗶𝗿𝗱 𝗺𝗶𝘀𝘁𝗮𝗸𝗲: 𝗧𝗵𝗲 𝗛𝗼𝗿𝗶𝘇𝗼𝗻 𝗧𝗿𝗮𝗽 Even the most state-of-the-art reinforcement learning systems must initialize somewhere. If you let an LLM initialize treatments based on correlational patterns in your warehouse, you fall into: The Horizon Trap. It looks like intelligence because it reflects past patterns. But those patterns are not causal. They are status-quo artifacts. The decisioning will converge. But to what? A local maximum defined by correlational guesses. You can dynamically allocate traffic across five bad ideas and still lose money. The algorithm didn’t fail. Your initialization horizon was wrong. 𝗪𝗵𝗮𝘁 𝗿𝗲𝗮𝗹 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻𝗶𝗻𝗴 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝘀 If you want to avoid: • Boomerang effects • Slow convergence • Local maxima traps You need causal priors, and decisioning belongs in constrained, auditable, outcome-driven systems. Have you seen LLM-driven “autonomous” systems produce real incremental lift — or just elegant automation?

19

3 Comments

Recruitonomics is on Substack! Check out Sam Kuhn's inaugural post (first link in comments) covering this morning's December jobs report. Going forward, all posts on the former Recruitionomics.com site will now live on Substack. Now, to the data: 50k jobs were added last month and the unemployment rate ticked down to 4.4%. But to abstract away from the effects of the government shutdown and the D.O.G.E.-related deferred resignations, private job growth is averaging just 29k over the last three months — less than half the pace seen in the Spring. Policy uncertainty (namely tariffs) and restricted immigration explain a lot of the decline in job growth. However, it doesn’t explain all of it. Bottom line: Hiring is really low outside of healthcare. For 2025 as a whole, about 600k jobs were added, the slowest pace (outside of recessions) since 2003. It's a "low-hire, low-fire" economy – not recessionary, and no major signs of imminent layoffs, but hiring is trending down, not up.

24

1 Comment

Snowflake’s SnowWork positioning is a signal. The race isn’t just about data anymore...... but about owning the AI workflow layer. Because once you control the workflow, you control how decisions get made. But here’s the gap: Workflows amplify whatever you feed them. And in most enterprises, the weak point isn’t orchestration. It’s the inputs..... messy documents, missing context, no clear lineage. So you get faster automation, without defensibility. That’s the real tradeoff showing up at the exec level: speed vs. trust The way I’m thinking about it is in three layers: 1. Workflow (orchestration) 2. Models (reasoning) 3. Inputs (can I trust what goes in?) Because you don’t get reliable AI by improving workflows. You get it by controlling what enters them. As platforms push toward owning the AI control plane, where are you placing your bets? If you're exploring this idea of the “AI control plane,” here is a simple breakdown: https://lnkd.in/gc2_4r97? #EnterpriseAI #DataStrategy #AILeadership #Snowflake #DigitalTransformation #AIControlPlane

24

4 Comments

Interesting call out from GF Data on leverage usage between add ons and platform investments in the 25-50m enterprise value range of deals (dare I call it the middle middle market?). It’s a trend that fundamentally makes sense but still is interesting to see the data support it.

8

Most CMOs still think Databricks is "that thing my data team uses." That assumption is about to cost someone their job. Databricks just shipped a native Meta Conversions API integration on their Marketplace. Brands can now send first party data straight from their lakehouse to Meta for ad targeting and attribution. No middleware. No reverse ETL. No custom connectors. A data infrastructure company just built a direct bridge to the world's largest social advertising platform. Sit with that for a second. Eighteen months ago, activating warehouse data for paid media meant buying something in the middle. A reverse ETL tool. A CDP. Some category of software whose entire value proposition was: we move your data from A to B. Real contracts. Real budget. Now it's a free solution accelerator on the Databricks Marketplace. Here's the pattern, and it's not slowing down: the platforms on each end keep getting smarter and more directly connected. The middle keeps getting thinner. Databricks to Meta today. Braze already has native CDI ingestion from Databricks. The trend line is clear: if Google and TikTok aren't already in conversations with Databricks about native integrations, they should be. The edges are eating the middle. But here's what's actually interesting: the brands that get ahead of this don't just cut costs. They gain a performance edge. First-party data, clean and direct, no latency, no translation layer. Precision targeting built on data they actually own. Databricks isn't a data platform anymore. It's a marketing activation platform. The companies that figure this out first will outperform the ones still buying point solutions to do it for them. Not by a little. By a lot. The data platform IS part of the marketing platform. Not as a prediction. As a fact. So here's the only question that matters: is your marketing team in the room where your data strategy is being decided, or are they going to get handed a brief sheet when it's done? One of those CMOs is about to have a very good year.

71

5 Comments

Semantic layers are emerging as core infrastructure for agentic analytics. Why? Because agents need context for accurate analytics, and that context is the semantic layer: it tells agents how business metrics like revenue or MAU should be calculated. Without this, they fumble badly. The big vendor debate is: where should this layer live: in the BI tool (Looker’s LookML, PowerBI’s DAX), in middleware (dbt, cube), or in the database? I strongly believe in “semantic pushdown”: stick metadata and metric expressions in the database alongside table definitions. Make it easy to manage and access via SQL and let it inform optimizations like materializations. After all, what is a database if not a place to store and query data?

50

26 Comments

8 Claude Workflows for ABM in 2026 Claude changes how AI shows up in B2B GTM. Now it's engineering context into reusable Skills and autonomous Cowork workflows. Here are the 8 Claude workflows that will 10x your ABM 1. Account research at scale Cowork aggregates firmographics, tech stack, hiring, news, financials, and buying signals. Claude Cowork + Clay + Chrome + Notion 2. Tiered content personalization Tier 1: Bespoke, account-specific → Tier 2: Industry-level relevance → Tier 3: Programmatic nurture Claude Cowork + HubSpot + Notion 3. Self-updating stakeholder mapping Feed CRM + LinkedIn data via Clay. Cowork continuously updates buying roles, engagement levels, coverage gaps,and flags missing Buyers. Claude Cowork + HubSpot CRM + Clay 4. Account-specific messaging at scale The Claude Skill adapts your value prop using account context, recent news, proof points, and persona language. Stack: Claude Skills + Claude Cowork + HubSpot 5. Intent signal translation Cross-references with news, hiring, initiatives, and recommends next-best-action within 24 hours. 6sense/ Bombora export → Notion → Claude Cowork 6. Automated sales → CS handoff briefs Claude scans transcripts, extracts commitments, maps stakeholders, flags risks, generates a clean CS brief. Gong/ Fathom.ai (or integrate with HubSpot) transcript export → Google Drive → Claude + Cowork → HubSpot 7. Competitive positioning by account Claude Cowork generates account-specific positioning based on competitors in play, stated priorities, and champion intel. Claude Cowork + Notion + Chrome (competitor research) + Perplexity Pro 8. Dynamic account plans For Tier 1 accounts, Cowork builds and refreshes structured plans: goals, stakeholders, 90-day actions, resources. Claude Cowork + Notion + HubSpot CRM How the system works: Claude Skills --> Reusable knowledge (playbooks, messaging rules, research frameworks). Claude Projects --> Persistent account workspaces. Claude Cowork --> Autonomous execution. --- Want the full Claude ABM Playbook? (8 prompt templates, SKILL.md example, 4-week implementation roadmap) Comment "CLAUDE" and I'll send it over. #ABM

104

261 Comments

Before AI SDRs, “autonomous agents,” and the flood of lead gen agencies running automation, there was OutboundWorks. That’s where I met Ben Sardella, Justin Michael, and later brendan short Short, who joined the team after selling Hexa.ai to them. Back then, Justin ran operations. Long before he became one of the top sales trainers and community builders. And the problems were the same as today. Automation created activity. Not conversions. Most leads still needed human follow-up. Calls. Context. Judgment. The difference was Justin saw the real constraint early. You could not scale SDR and CX teams at $100k per head and win. So he looked south. LATAM talent had the skills, the work ethic, and the communication ability. What they lacked was access and structure. That’s where CloudTask came in. We built the team. He built the systems. Lower overhead. Higher quality. Real outcomes. Different tools today. Same truth. Automation does not replace people. It exposes whether your talent and systems are any good. Most companies are still learning that the hard way.

30

19 Comments

Founders: Ali Ghodsi, the CEO of Databricks explains the successful strategic partnership dynamic so well that I had to share it with you. In businesses development, there has to be a ‘give and get’. What does this mean? There has to be a very important value proposition on each side that it is very difficult to duplicate on your own, especially for the bigger company in the partnership. Great stuff!

1

Snowflake × OpenAI: why this deal feels less like a power move at the data layer: a multi-year $200M agreement to bring frontier models directly inside Snowflake's Cortex AI (Snowflake's Intelligence layer for enterprise agents) If you’re thinking in systems terms: CortexAI is Snowflake trying to make “AI on enterprise data” feel like a native database capability — not a separate product you bolt on later. The story is that agentic AI only becomes valuable in production when it can reason over governed, proprietary data (without your org turning into a compliance crime scene.) Yesterday, Snowflake and OpenAI announced a multi-year, $200m partnership to bring OpenAI models natively into Snowflake’s Cortex AI, available to Snowflake’s 12,600 customers across all three major clouds. A practical lens to read this deal --- We can 'think of modern AI' as a 'supply chain': raw material = enterprise data factory = compute + connectors foreman = evals + guardrails machines = models output = decisions + actions Most orgs are trying to buy “better machines” while the foreman is missing and the factory floor is chaotic. This partnership is Snowflake trying to own the factory floor, while OpenAI provides the best machines. Enterprises already park their most valuable asset (data + permissions + governance) inside Snowflake. If the AI runs inside that boundary, Snowflake becomes the default runtime for “AI on enterprise data” instead of being just storage/warehouse.

6

1 Comment