The Agentic Survival Filter
ANDYLΞNZ@andy
+3.0%
since Oct 1
The idea
I believe agentic AI is becoming a survival filter: companies that build a shared data and agent core will structurally cut costs and pull away, while laggards get repriced as uninvestable. Since this only holds long positions, it owns both sides: the platforms selling the company brain and orchestration plumbing, and early adopters in structured, labor-heavy fields like insurance, banking, tax, legal, payments, travel and logistics where repetitive knowledge work meets recurring decisions.
What it owns
- The platforms and data cores that companies must adopt to run agents through their operations at scale. · MicrosoftIt sells the default enterprise brain to companies scrambling to adopt agents: Copilot, Azure AI, and the data stack underneath., Palantir TechnologiesIts ontology functions as a literal company brain, a shared data core that agents act through in real operations., ServiceNowIt already owns the structured workflows agents need to automate, and companies tend to buy their agent layers where their workflows live., SnowflakeIt supplies the shared data core many enterprises rely on, and agentic workloads pull more data and compute onto its platform., SalesforceIt is pushing agents into the largest installed base of customer-facing workflows in software, and early adopters tend to buy agents through tools they already use.30%
- Underwriting, banking and payments are structured, rules-heavy and labor-intensive work where agents cut costs first. · JPMorgan ChaseBanking and wealth management sit squarely in the filter: structured data, high labor costs, recurring decisions, and it invests more in AI than any other bank., ProgressiveIt has proven that data-driven decisions win in insurance, and it owns the structured data and recurring decisions where agents cut costs first., LemonadeIt was built from the start with AI handling underwriting and claims instead of people, making it the purest test of agent-run insurance operations., VisaIt builds the rails for agents that shop and pay autonomously, and most agent-driven commerce likely settles through its network.18%
- Tax, legal, compliance and business process work is repetitive knowledge work on structured data, the purest match for agent automation. · IntuitTax and accounting is repetitive knowledge work on structured data, and it is pushing AI agents through TurboTax and QuickBooks to automate that work., Thomson ReutersLegal, tax and compliance work runs on its data, which is exactly what agents in those fields need to be trusted and compliant., Verisk AnalyticsIt supplies the structured decision data that insurance agents will run on, so it gets paid regardless of which insurer adopts agents fastest., GenpactIt is paid for human knowledge work and now sells agents to do that work instead, testing whether it can disrupt its own business.16%
- High-volume recurring decisions in booking, freight and operations are where companies deploying agents quietly widen margins. · Booking HoldingsTravel booking is high-volume, structured and digital end to end, and it operates the platform where AI travel agents actually transact., HoneywellIt is putting AI and autonomy into its own factories and its customers' operations, and the market still prices it as a traditional industrial., C.H. RobinsonFreight brokerage is repetitive knowledge work with high labor costs, and it is using agents to price and book loads with fewer people., WorkdayRecruiting and HR are structured work on people data, and it is where large companies keep the data agents need to automate those decisions.14%
- The orchestration, memory, monitoring and compute that agents run on, needed regardless of which platforms ultimately win. · NvidiaEvery agent runs on compute, and it still supplies most of the hardware that agentic AI demands., DatadogWhen agents run large parts of a value chain, someone must watch what they are doing, and it builds the observability layer for AI systems in production., MongoDBAgents need memory and state, context and search over messy operational data, which is its core product., UiPathIt already automates repetitive knowledge work and is rebuilding around agent orchestration, converting its automation base into agent-native customers.14%
- Cash8%