AI Resistance Signals
A practical lens for judging whether a SaaS product can be replaced or bypassed by AI agents.
AI resistance is a product’s ability to avoid being replaced or bypassed by AI agents operating on user-owned compute, user-owned data, and user-owned context.
The core distinction is whether the product’s value comes from something the agent can reproduce on demand, or from assets the product has accumulated over time.
Low-resistance SaaS usually looks like:
- User data in, transformed output out.
- User provides goals, context, and workflow knowledge.
- Vendor provides prompts, orchestration, formatting, and a polished UI.
These products can still be useful, but their advantage is fragile. If the customer can hand the same inputs to an agent and get a comparable output, the product is exposed.
Higher-resistance SaaS accumulates assets that agents cannot easily recreate:
- Proprietary datasets
- Network participation
- Operational history
- Trust and verification evidence
- External integrations
- Infrastructure
- Distribution
- Canonical state
Examples of higher-resistance shapes:
| Resistance source | SaaS shape |
|---|---|
| Proprietary datasets | Benchmarking platforms, fraud intelligence systems, vertical data products, and market data tools that improve as they collect unique observations. |
| Network participation | Marketplaces, procurement networks, payment networks, partner ecosystems, and platforms where value depends on other participants being present. |
| Operational history | Systems of record for customer support, compliance, fleet operations, clinical workflows, construction projects, or field service. |
| Trust and verification evidence | Identity verification, audit trails, compliance evidence, security monitoring, and vendor risk systems. |
| External integrations | Vertical SaaS embedded into accounting, payroll, EHR, ERP, CRM, logistics, or payment workflows. |
| Infrastructure | Developer platforms, observability systems, cloud services, data pipelines, communications APIs, and deployment infrastructure. |
| Distribution | Products with strong channel access, embedded reseller relationships, app marketplaces, or default placement inside larger platforms. |
| Canonical state | Systems that hold the accepted source of truth for inventory, tickets, contracts, financial records, medical records, or customer accounts. |
The most useful questions are:
- How hard is the product to reproduce?
- How hard is the product to bypass?
- What does the product know, own, coordinate, or guarantee that the customer cannot simply provide to an agent?
The second question is often more important in an agent-native world. A product can be hard to clone and still be easy to route around if the customer can get the outcome directly from their own data, tools, and context.