Superpower Daily
Figure Signs $3.5B Nscale Deal for Up to 100,000 GPUs for Human...Trump Order Opens Path to Restrict Foreign Power Gear for AI Da...Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have...Uber and Wayve Put 15 Supervised Robotaxis on London’s StreetsMostik Links AI Models Through Their Weights, Claiming a 20x Co...LlamaIndex and Kaggle Launch a Document Benchmark for Cost and...Figure Signs $3.5B Nscale Deal for Up to 100,000 GPUs for Human...Trump Order Opens Path to Restrict Foreign Power Gear for AI Da...Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have...Uber and Wayve Put 15 Supervised Robotaxis on London’s StreetsMostik Links AI Models Through Their Weights, Claiming a 20x Co...LlamaIndex and Kaggle Launch a Document Benchmark for Cost and...
Toolspublished

Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have No Coverage

The new ATE dataset offers a supply-side map of what developers have made available for AI agents, not proof of workplace adoption. Its central finding is less about how many tools exist than which job tasks they reach—and which they leave alone.

By 3 min read
Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have No Coverage
Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have No Coverage
Sign in to save and react

Listen to this story

The audio brief

About 1:49
0:001:49
Read transcript
Cohere Labs found that 419 of 923 U.S. occupations have no publicly listed AI-agent tool mapped to them at all. That is the headline finding from its new Agentic Task Ecosystem dataset, or ATE: a snapshot of what developers have made available, not evidence that companies or workers are actually using those tools. Cohere examined 696,291 tools on 123,069 public MCP servers, gathered from seven directories in May. MCP, the Model Context Protocol, is an open standard that lets an AI system connect to software and data—for example, querying a database, editing a file, or scheduling a meeting. The test was deliberately strict. A tool had to match an O*NET occupational task and complete it end to end, rather than simply provide information or handle one step in a larger human-coordinated process. By that measure, just 2.6 percent of tools qualified, covering 1,380 task statements—about 15 percent of the software-performable work in O*NET. The rest is not necessarily irrelevant. Cohere classified 693 categories as smaller substeps and 411 as multi-task workflows. Tool supply also correlated with theoretical AI exposure across 178 occupations, but not with workers’ preferred automation targets. Specialized healthcare and computing work saw coverage, while legal, production, and sales tools leaned toward routine tasks. The key constraint is visibility: private enterprise MCP servers are missing, and ATE says nothing about reliability, deployment, or long-horizon completion. The open question is how much of that unseen tooling changes the 2.6 percent floor.

Story brief

3 key points

Cohere Labs has released the Agentic Task Ecosystem dataset, a public snapshot of how MCP tooling maps to U.S. occupational work. Its strict end-to-end test identified only 1,380 O*NET tasks across 696,291 listed tools, leaving 419 occupations without a represented tool. The dataset is useful for measuring public tool supply and builder priorities, but not workplace adoption: it excludes private servers and says...

  1. 01

    The dataset covers 123,069 public MCP servers across seven directories, collected in May and deduplicated across listings.

  2. 02

    Only 2.6% of tools matched a complete O*NET task; 693 categories were smaller substeps and 411 were multi-task workflows.

  3. 03

    Tool availability correlated with theoretical AI exposure across 178 occupations, but not with workers’ preferred automation targets.

The public market for AI-agent tools is vast, but its direct footprint on recognized occupational work is still narrow. Cohere Labs’ new Agentic Task Ecosystem dataset finds that 419 of 923 O*NET occupations have no represented agentic-tool activity, while only 2.6% of the nearly 700,000 catalogued tools meet its strict test for completing a recorded work task end to end.

That result comes from a new public dataset, not an employment measure. Cohere collected 696,291 tools listed on 123,069 public Model Context Protocol, or MCP, servers across seven directories in May. MCP is an open standard that connects AI systems to external software and data, allowing an agent to call functions such as querying a database, editing a file, or scheduling a meeting.

A deliberately demanding definition of automation

Cohere matched each tool description to the closest task statement in the U.S. occupational database O*NET, then used a language model to assess whether the tool executes that task. The protocol excludes tools that merely inform a person or perform one step while a person still coordinates the wider process. Under that definition, 2.6% of tools matched a full O*NET task, covering 1,380 task statements—about 15% of the software-performable work O*NET records.

The unmatched tools are not simply irrelevant

The low match rate does not mean the rest of the catalog is useless. Many tools describe existing work at a different level of detail than O*NET: some break a recognized task into smaller steps, while others bundle several tasks into a workflow. The remaining groups include infrastructure for agents themselves and a smaller category of work that appears to exist because agents do.

How Cohere classifies work outside a full O*NET task match

  • 693 categories were “subatomic”: recognizable work divided into units smaller than an occupational task statement.
  • 411 categories were composite workflows spanning multiple recorded tasks rather than one task statement.
  • Only 35 categories were classified as having no plausible occupational counterpart; most concerned managing or evaluating agents.

Exposure can predict volume without revealing the target

Across 178 occupations, Cohere found a 0.54 correlation between theoretical AI-exposure scores and the number of MCP tools available. But exposure did not predict whether tools landed on routine tasks or specialized ones within a job. That distinction matters: tools reached toward specialized work in healthcare and computing, while legal, production, and sales tools clustered nearer routine work.

The pattern also follows what builders think is technically feasible, rather than what workers say they want automated. For 65 occupations covered by WORKBank data, available tools tracked expert capability judgments but showed no relationship to workers’ automation preferences.

Availability is not deployment

ATE measures the public supply of tools, not their use, reliability, or ability to complete long-horizon workplace assignments. It also excludes bespoke internal MCP servers, which may be concentrated in back-office processes. Cohere argues that omission makes the 2.6% figure a floor for publicly observable tooling, but the dataset cannot establish how much of the listed automation has reached real workplaces.

Sources

  1. cohere.comAutomation’s Early Footprint: The ATE Dataset

Related coverage

Continue reading