Sticky Software Survives Agents
T-R3x3r@tr3x3r
+3.9%
since Oct 2
The idea
I think the market is overreacting to AI agents replacing business software, so I built a basket of the sticky, hard-to-rip-out systems that run companies day to day: the big systems of record, the data layer beneath them, mid-caps hit hardest by the agent scare, and vertical software in industries like pharma and government. I deliberately avoided thin front-end tools where agents genuinely do a better job, and I kept a small cash buffer. I'd rethink this if agents start displacing core record-keeping systems rather than just the interface layer.
What it owns
- Core software embedded so deeply in how companies run that ripping it out takes years, making it a natural system for agents to read from rather than replace. · SalesforceMoving a customer record system away from Salesforce takes multiple years, so switching costs protect renewal revenue even as AI agents emerge as a new layer on top., SAP SEIt runs the financial and supply chain systems at most large companies, and rebuilding an ERP system from scratch rarely happens, so the position holds if churn and migrations stay low., ServiceNowIt is the workflow backbone for IT and operations at large enterprises, the kind of system agents would plug into rather than replace., WorkdayIt holds the HR and finance records companies run payroll and planning on, a system so central that moving it would be costlier than adopting agents on top of it., IntuitIt is the finance record for small businesses through QuickBooks, data painful to move and exactly what any agent would need access to for financial work., Automatic Data ProcessingIt runs payroll, the single system companies are least willing to rip out because a mistake means people don't get paid, making this the stickiest corner of the take.37%
- Sticky software marked down hardest on the narrative that agents kill SaaS, creating an opportunity if retention and revenue prove the tools stay essential to how people work. · AdobeIt has been marked down harder than most on fears of free AI replacing creative work, but its tools remain the ingrained workflow for professionals and the system brands run their assets through., HubSpotIt is the marketing and CRM record for mid-sized companies, and the take works if customer counts and retention stay solid., AtlassianIt holds the work and history for engineering teams in Jira and Confluence, sticky even as agents write more code because someone still has to track and coordinate it., DocuSignIt holds the signed agreements companies legally depend on, a record that does not move because agents do not replace the work inside DocuSign., Monday.comThe product keeps becoming more central to how teams organize work, with growing adoption and expansion per customer showing the sell-off reflected sentiment rather than business deterioration.24%
- Industry-specific software where switching costs are so high and regulatory or operational requirements so specific that customers rarely leave, even as technology shifts. · Veeva SystemsIt is the system of record for pharma, an industry where regulators mandate record-keeping and nobody improvises their own software, creating deep vertical lock-in., Roper TechnologiesIt owns niche vertical software businesses deeply ingrained in each industry's daily operations, a diversified bet that sticky software in specific verticals survives automation., Tyler TechnologiesIt runs courts, permits, and payments for local governments, buyers who change software roughly once a generation and cannot shortcut government procurement cycles., Procore TechnologiesIt is the project record for construction, where data about what got built, by whom, and who is liable must live somewhere permanent and auditable.17%
- Infrastructure that agents need to read from and write to, not replace, so more AI means more usage of these systems rather than less. · MongoDBIt is the database layer beneath modern applications, and databases are where agents read and write data rather than where they replace the infrastructure., SnowflakeIt is where enterprises keep the data any agent would have to query, so more AI means more data warehouse usage rather than less., DatadogIt is wired into how companies monitor their systems, and more software, whether written by agents or people, means more infrastructure to watch.14%
- Cash8%