Why The AI Token Economy Could Reshape In-House EHS Software Strategies

AI is lowering the barrier to EHS software innovation, but also raising the cost. The EHS software market is entering a new phase of experimentation, driven by the rapid commoditization of AI tools and the emergence of platforms that make development more accessible. The result is increasing uptake of in-house builds over purchased pre-built solutions. While this shift signals a new frontier of flexibility and control, it also introduces a less visible – but increasingly critical – constraint: the rising cost of AI.

The growing accessibility of AI development platforms is redefining the competitive landscape. Organizations can now assemble bespoke EHS capabilities more quickly and at lower upfront cost. In parallel, heritage EHS vendors can accelerate their own AI roadmaps to defend their position against AI-native start-ups, reducing dependence on purely organic development and allowing them to close capability gaps through faster iteration and integration.

However, early evidence suggests that the in-house trend will be self-limiting. Building an EHS platform is only the first hurdle; sustaining it is far more complex. Organizations must replicate not just functionality, but also the depth of regulatory content, data, auditability and workflow maturity embedded within established platforms. As a result, many firms are likely to encounter diminishing returns as internal systems scale, particularly in risk-sensitive environments where reliability and compliance are non-negotiable.

At the same time, the economics of AI are becoming a defining factor in long-term viability. While AI tools promise efficiency gains, the underlying cost structure is increasingly scrutinized. Token consumption, the unit cost associated with running large language models (LLMs), is emerging as a hidden pressure point across enterprise use cases. In EHS, where applications often involve large volumes of unstructured data (such as incident reports, safety observations and permit documentation), token usage can scale rapidly. This creates a direct trade-off between capability and cost, particularly for organizations pursuing continuous monitoring, real-time insights or autonomous workflows.

This challenge is amplified by broader market signals indicating that AI compute costs can exceed traditional labour costs, with some firms significantly overshooting initial token budgets. The recent OpenClaw token cost story provides a useful illustration: users of the open-source AI agent framework bypassed per-token API payments by using Claude subscriptions. This resulted in the running of heavy, ‘always-on’ agent workloads at flat subscription prices intended for conversational use. In early April 2026, Anthropic changed its policy to block third-party tools such as OpenClaw, OpenCode and Clawdbot from using Pro/Max Claude subscription credentials, pushing that usage onto pay-as-you-go API billing instead. For EHS teams, this introduces a fundamental tension. Leaders are under pressure to adopt AI-driven innovation, yet must do so within cost frameworks that are often incompatible with high-volume, compute-intensive use cases.

The solution is likely to be a hybrid approach. Rather than fully replacing vendor platforms, organizations will combine targeted in-house agents with established EHS systems, leveraging model context protocols (MCPs) and agent-to-agent (A2A) protocols to connect bespoke AI workflows with external solutions. Vendors, in turn, have an opportunity to differentiate by optimizing AI delivery through more efficient model architectures, better token management and clearer cost transparency.

To learn more about how AI, start-ups and heritage vendors are reshaping the EHS software market, look out for the upcoming Verdantix Market Insight: The Rise Of EHS Start-Ups And The Value Of Industrial Heritage.