What happened
OpenAI published its first B2B Signals report, a benchmark built from aggregated enterprise usage of OpenAI products. The headline claim is simple: the companies getting the most out of AI are no longer just giving employees access to chat tools, they are using far more AI per worker and applying it to more complex tasks.
According to the report, so called frontier firms now use 3.5 times as much intelligence per worker as typical firms, up from 2 times a year ago. OpenAI argues that only part of that gap comes from sending more prompts. Most of it comes from deeper usage, richer context, and more substantive work being delegated to AI.
The strongest signal in the report is around agentic tooling. OpenAI says frontier firms send 16 times as many Codex messages per worker as typical firms, and they also over-index on tools like ChatGPT Agent, GPTs, Apps, and Deep Research. In plain English, the leaders are moving from asking questions to handing off multi-step work.
Why it matters
This matters because it gives enterprise buyers a clearer maturity model. The market spent the past two years talking about adoption in seat counts, pilots, and internal excitement. B2B Signals suggests the real dividing line is now workflow depth. The winners are the firms that connect AI to real tasks, real systems, and real operating constraints.
It also supports what many engineering teams already suspect: chat alone does not create durable advantage. If AI stays trapped in side-window productivity, the gains stay shallow. Once AI starts drafting code, handling claims intake, triaging support, or creating records in systems of record, the economics change because work actually moves.
There is a second angle that matters for Europe. Many organizations here are still cautious about governance, vendor lock-in, and data handling. That caution is rational, but it can also slow execution. Reports like this increase pressure on leadership teams to move beyond experimentation and build controlled production workflows before the gap with faster adopters widens.
Laava perspective
At Laava, the interesting part is not the benchmark itself. It is what the benchmark rewards. The firms pulling ahead are not the ones with the loudest AI messaging. They are the ones embedding AI inside constrained, repeatable business processes where the output matters and the handoff is clear.
That aligns with how we approach production AI agents. Start with a document-heavy or workflow-heavy process. Map the trigger, the business rules, the exception path, and the system action. Then build an agent that can read, reason, and execute inside existing ERP, CRM, email, or internal tooling, with auditability and human approval where needed.
This is also why we stay skeptical of demo theater. A polished chat interface can make an organization feel innovative without changing much underneath. B2B Signals reinforces the opposite lesson: enterprise value compounds when AI becomes operational infrastructure, not when it becomes another browser tab.
What you can do
If you lead operations, finance, customer service, or another back office function, this is a good moment to audit where AI is still stuck at the assistant layer. Look for repetitive workflows involving PDFs, email, forms, approvals, or data entry between systems. Those are often better candidates for production agents than another generic chatbot rollout.
Then measure depth, not hype. Track whether AI is reducing handling time, improving throughput, cutting exception work, or increasing the share of work completed inside the actual workflow. That is the difference between an AI program that looks busy and one that creates compounding operational value.