HumanX 2026 shows AI agents and in-house applications are reshaping enterprise software budgets

Overview

This report summarizes the observations from UBS's software team after two days of field research at the San Francisco HumanX 2026 AI conference, focusing on Anthropic Mythos, OpenAI, enterprise AI deployment, and the impact of AI on the software industry. The report argues that attendees are generally more forward-looking than ordinary large enterprises, but customer interviews still show that enterprise software purchasing logic is changing: more budgets are flowing to model vendors, AI agents, and internally built tools, putting traditional SaaS/application software companies under a tougher growth narrative.

Core views

The core view is that improvements in AI model capability and agentic applications are lowering the barrier to building software in-house, especially hitting workflow-oriented, point, and adjacent application software. Customers still rely on third-party software for core system-of-record, compliance, and regulatory functions, but the willingness to self-build is clearly increasing in process automation, data visualization, sales support, and some security tools layered on top of those systems. By contrast, the rising importance of enterprise data is providing stronger support for data and knowledge-graph-related platforms such as Snowflake, Palantir, and Databricks.

Analysis framework

The report combines on-site conference research, independent customer interviews, feedback from partners and AI-native companies, and observations from public fireside chats. UBS maps customer feedback on build-vs-buy, model usage, Copilot adoption, data integration, system-of-record resilience, and AI agent deployment to the potential revenue growth, competitive risk, and valuation narratives of public software companies.

Methodology notes

• channel researchHumanX 2026 conference customer interviews

  Assess changes in enterprise software procurement and self-build behavior by speaking with AI-forward enterprise customers, partners, and AI-native companies.

  The report notes that HumanX attendees are more aggressive than the average Fortune 500 company, so the sample is not fully representative of mainstream enterprises, but it does reflect shifts in software industry direction.

• industry analysisBuild-vs-buy framework

  Compare the opportunity cost, data control, compliance requirements, and product maturity of building AI applications in-house versus buying third-party SaaS.

  When AI compresses development cycles from months to weeks or even less, enterprises are more likely to self-build workflow and adjacent applications; however, core systems of record, heavily regulated processes, and highly mature platforms are still difficult to displace quickly.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• SNOW.US

  Enterprise data platform; the research suggests that the rising importance of enterprise data is relatively favorable for Snowflake.

  Strengths

  Customers said they are investing in data unification and internal knowledge graphs, and a Snowflake customer said it plans to continue increasing related spending.

  Weaknesses

  Data platform stocks are also being scrutinized by the market because of AI substitution risk, and valuations may be sensitive to changes in growth and AI narratives.

  Comparison

  Compared with workflow-oriented SaaS, Snowflake is closer to the data infrastructure needed by AI agents; the report did not hear customers say Claude/GPT would directly replace Snowflake.

  Risks

  If model vendors or cloud vendors provide a more closed integrated data layer, or if enterprise data openness is constrained, Snowflake's growth narrative could still come under pressure.

• PLTR.US

  Enterprise data, knowledge graph, and AI application deployment platform; the report argues that enterprise data advantages are becoming more important.

  Strengths

  The research feedback emphasized the importance of internal knowledge graphs, proprietary data, and reasoning reliability, which aligns with Palantir's data integration and AI application positioning.

  Weaknesses

  Palantir's share price may also be scrutinized because of AI risk and a high valuation.

  Comparison

  Compared with ordinary SaaS point applications, Palantir is more like the data and orchestration base for enterprise AI applications; the report did not hear customers say Claude/GPT pose a direct substitution risk to Palantir.

  Risks

  If enterprises choose to fully build their platforms in-house using open models and their own engineering teams, or if competitors offer lower-cost data orchestration solutions, Palantir's incremental upside could be affected.

• Microsoft

  A beneficiary tied to Copilot, enterprise governance, security, and permissions management, but customer feedback is mixed.

  Strengths

  Microsoft has advantages in enterprise trust, security, access control, and governance, and customers believe its catch-up ability should not be underestimated.

  Weaknesses

  Several customers said Copilot adoption is poor, data-source connectivity is difficult, or Microsoft is behind in AI harness capabilities.

  Comparison

  Compared with OpenAI, Anthropic, and Glean, Microsoft's model and product-experience feedback is weaker, but its enterprise distribution and governance foundation is stronger.

  Risks

  If Copilot continues to fail to demonstrate clear ROI, enterprises may shift to Glean, OpenAI, Anthropic, or internally built solutions.

• Salesforce

  A CRM and application software leader facing risks from data openness, self-built CRM, and Agentforce adoption.

  Strengths

  It has a large base of customer usage data, mature functionality, connectors, and system-of-record capabilities, which are still difficult to fully replace in the near term.

  Weaknesses

  Customers are unhappy with data closedness and ecosystem lock-in, and some enterprises are considering reducing Salesforce/Tableau seats or building internal replacement tools.

  Comparison

  Compared with data platforms, Salesforce is more exposed to substitution risk at the workflow and front-end application layer; however, compared with smaller point-SaaS vendors, it has deeper customer data and broader functionality.

  Risks

  Over a 3-5 year horizon, if AI significantly reduces the cost of building custom CRM/ERP systems, Salesforce's seat-based revenue and incremental growth could come under pressure.

Key data

• Report date2026-04-13UBS Global Research publication date.
• Conference locationSan FranciscoUBS's software team conducted a two-day research visit at the HumanX AI conference.
• OpenAI enterprise contract size$100m+The report says OpenAI's initial enterprise contract value could exceed $100 million.
• Software development efficiency improvement example20-30%Some customers said overall software development speed had improved by at least 20-30%.
• COBOL modernization example8 months to 8 daysCustomer feedback showed that the timeline in some refactoring scenarios was dramatically compressed.
• Customer churn cost example$40 million/yearOne interviewee said a customer lost $40 million per year because of onboarding process issues, which AI agents could help improve.

Impact & implications

For investors, the report reinforces a cautious view on traditional SaaS/application software companies: AI agents, self-built tools, and integrated model-vendor applications may depress future growth rates and weaken seat-based subscriptions and point-software value. At the same time, software platforms with key enterprise data, governance, permissions, system-of-record capabilities, and deep industry knowledge remain defensive; data platforms may benefit from enterprise data unification, knowledge graphs, and demand for AI agent access.

Risks

• HumanX attendees are at the AI frontier, so the sample may overstate the speed of enterprise self-building and the maturity of AI adoption overall.
• System-of-record, compliance, regulation, and HIPAA-constrained scenarios may make the software displacement process slower than the market fears.
• Product roadmaps at OpenAI, Anthropic, Microsoft, Google, Salesforce, and others could quickly reshape the competitive landscape.
• AI self-build solutions require data unification, permissions, governance, security, and skilled engineering capabilities, which not all enterprises have at sufficient scale or maturity.
• The report does not provide formal stock ratings, price targets, or quantitative earnings forecasts; the investment conclusion is driven more by qualitative research.

What to watch

• Whether OpenAI and Anthropic continue launching stronger one-stop applications and enterprise products.
• How Anthropic Mythos performs in security, code, application-layer, and enterprise deployment use cases.
• Enterprise adoption rates, renewal rates, and ROI evidence for Copilot, Glean, Claude, GPT, and Gemini.
• Whether SaaS companies open up data, support cross-stack agent integration, and move from seat-based pricing to outcome-based pricing.
• Whether data platforms such as Snowflake, Palantir, and Databricks can convert the rising importance of enterprise data into revenue growth.
• Whether workflow software, point applications, developer tools, visualization, and some cybersecurity software start to show budget cuts or seat attrition.