Friday night, 9 p.m., a Texas high school football field. Head referee Mark is waiting on the next down. His earpiece crackles: "Heads up — UIL changed the holding-replay protocol this season." Mark winces. He works three levels this fall — varsity high school, JUCO, and semi-pro — each with its own rulebook, its own video review system, its own state association portal, and he has to remember the differences in real time. According to the U.S. Bureau of Labor Statistics (BLS) Occupational Outlook Handbook, Mark is one of 19,300 umpires, referees, and other sports officials in the United States. And in June 2026, a specification called Agentic Resource Discovery (ARD) hit the Hacker News front page — implemented by services like GitHub Agent Finder and Hugging Face Discover, it solves exactly the "tool fragmentation, can't find the rules" problem Mark hits every weekend.
This article wires together official BLS occupational data, the public ARD specification, and a concrete onboarding path that a state-association training director could evaluate tonight — giving these 19,300 jobs a real AI use case that augments rather than replaces.
1. The Pain: 3 Real Pain Points the BLS Data Surfaces for Sports Officials
According to the U.S. Bureau of Labor Statistics Occupational Outlook Handbook entry on Umpires, Referees, and Other Sports Officials, updated August 28, 2025, there were 19,300 sports officials in the U.S. in 2024 (SOC code 27-2023), with a median annual wage of $38,820 (the lowest 10% earned under $25,070, the top 10% over $93,180). BLS projects 6% employment growth from 2024 to 2034 — labeled "faster than average" against the 3% all-occupations baseline — with 4,600 openings each year. Eighteen percent are self-employed, 17% work in amusement/gambling/recreation, 17% in performing arts and spectator sports, 8% in education.
BLS spells it out in Work Environment: "the job is stressful because officials often must make split-second rulings." From that single line, three pain points fall out:
Pain point 1: Rule updates scattered across 50+ state and sport-specific associations. BLS writes plainly in How to Become One: "Requirements for umpires, referees, and other sports officials typically vary by state and local sports association." Every state, every sport (football, baseball, basketball, soccer, hockey) maintains its own rulebook, exam, and registration portal. A cross-sport official like Mark has to comb 5–8 different websites pre-season just to track rule diffs — entirely by hand.
Pain point 2: Video replay and AI review tools are deeply siloed. Research shows the sports world already runs dozens of niche AI tools — MLB's Hawk-Eye electronic strike zone, the Premier League's SAOT semi-automated offside, NCAA's video review portal, Hudl Assist for high school footage, the NFL Replay Center's proprietary stack. Data shows a cross-sport veteran official typically logs into 6–10 separate systems to complete a post-game review.
Pain point 3: 18% self-employment and seasonal moonlighting amplify information asymmetry. BLS data shows 18% of officials are self-employed, and "Many umpires, referees, and other sports officials are employed primarily in other occupations and supplement their income by officiating part time." Those moonlighting officials lack league-internal resource libraries — they have to stitch the toolkit together themselves. The 4,600 newcomers entering each year especially get stuck on "which AI tool do I use, which rulebook do I read?"
2. What the AI Tech Is: The Agentic Resource Discovery (ARD) Specification
According to the official ARD specification, Agentic Resource Discovery is an open spec that lets AI clients (Claude, ChatGPT, GitHub Copilot, and more) ask one unified question — "what agentic resource can help with this task?" — and search across MCP servers, Skills, APIs, workflows, and agents distributed across the public web, enterprise estates, and third-party vendors. The spec only handles discovery; invocation still flows through whatever protocol the target resource natively speaks (MCP, REST, agent framework, etc.).
ARD has three load-bearing design choices. First, discovery is separated from invocation — the client finds the resource with ARD, then runs it with the resource's native protocol. Second, multiple services implement the same spec — GitHub's Agent Finder and Hugging Face's Discover are both ARD implementations, so a resource published once is discoverable by many clients. Third, resource descriptions carry what/who/where/how metadata so AI clients can judge fit automatically.
For sports officials, the killer benefit is that ARD lifts the "tool map" out of the official's head and into a machine-readable layer.
3. How to Use It: A 3-Step ARD Onboarding Path for Sports Officials
Step 1: State associations publish rule updates as ARD-discoverable resources. State associations (Texas UIL, California CIF) wrap their annual rule-diff documents, video-review specs, and exam question banks as MCP servers or APIs conforming to ARD's resource description, then register with any ARD discovery service (GitHub Agent Finder, Hugging Face Discover). One publication — discoverable by every ARD-aware AI client.
Step 2: Officials query through natural language in an AI client. Mark types into Claude or ChatGPT: "I'm officiating UIL 5A football in Texas next week — find me the latest rule diffs and approved video review tools." The client hits the ARD backend and returns: UIL 2026 preseason rule update doc, Hudl Assist UIL integration guide, Texas UIL video replay protocol MCP server.
Step 3: Invoke the matched agentic resource directly. ARD returns the what/who/where/how four-tuple; the AI client uses the how field to invoke each resource natively (MCP over stdio, API over REST). The entire chain is transparent to the official.
4. Case / Outcome: The Librarian Pilot as Reusable Template
According to the ARD spec docs, GitHub Agent Finder and Hugging Face Discover are already live as reference implementations. The librarian case study published the day before (2026-06-22 librarians-agentic-resource-discovery) validated that in any industry where tools are fragmented but rules are standardizable, ARD broadcasts a single resource publication to many clients and meaningfully drops integration cost.
For a 19,300-person workforce spread across 50+ associations with 4,600 new openings each year, the value isn't replacing officials — it's giving the moonlighter and the first-year official the same toolkit a league's full-time officials get, accessed through one interface in the 30 minutes before kickoff instead of by trawling 8 websites.
5. FAQ: 5 Top Questions
Q1: How does ARD relate to MCP?
A: ARD is the "discovery protocol," MCP is the "invocation protocol." ARD helps an AI client find a usable MCP server (or API, agent, skill); the call itself still goes over MCP. They're complementary. As the ARD docs state, ARD "only handles discovery — the client invokes the agentic resource through that resource's own mechanism."
Q2: How does a state association publish its rule resources to ARD?
A: Per the ARD publishing guide, the state association exposes resource metadata (what/who/where/how) per the spec and registers with any ARD discovery service (GitHub Agent Finder, Hugging Face Discover). Publish once, discoverable by many clients.
Q3: According to BLS data, what's the job outlook for sports officials?
A: BLS 2024 data shows 19,300 jobs, with 6% projected growth from 2024 to 2034 — well above the 3% all-occupations baseline and tagged "faster than average." About 4,600 openings open each year, with a median annual wage of $38,820.
Q4: Self-employed officials don't have IT teams — can they use ARD?
A: ARD is designed so the AI client does the discovery for the user. Mark just asks Claude or ChatGPT; the client hits the ARD backend automatically — no code required. That's exactly the spec's "discovery is the bottleneck, not invocation" position.
Q5: Will ARD let AI replace human officials?
A: BLS lists Important Qualities as "communication, decision-making, good vision, physical stamina, teamwork" — none of which AI can substitute. ARD's job is to drop the cost of finding rules, replay tools, and training resources to zero, so officials can keep their attention on the call and the conversation on the field.
Closing Note
19,300 officials, 50+ state association rulebooks, 6–10 niche AI tools — that's the fragmentation the BLS data exposes. The Agentic Resource Discovery spec isn't another forgettable standard; it's the tool-map infrastructure for the AI era. If you run training for a state association, you officiate part-time, or you're building a sports-officiating SaaS that wants to plug into the AI ecosystem, you can publish your first ARD-discoverable resource tonight from the ARD Get Started page.
The tool-discovery era starts here — beginning with high-school football refs like Mark.