Industries / Government, education and community / Nonprofits / Labor unions
Member services
Unions and the funds they run give members benefits beyond the contract, such as health care, pensions, unemployment pay, legal help, discounts and hardship aid.
5
Milestones achieved, from 1901 to 2022.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1901
Union unemployment funds
In 1901 the Belgian city of Ghent began subsidizing unemployment funds run by trade unions. By the 1920s this "Ghent system" was the more common model in Europe, but most countries replaced it with compulsory state insurance after the Second World War, and union-linked voluntary funds survived only in Denmark, Sweden and Finland, by Jochen Clasen and Elke Viebrock's account.
Credit
the city of Ghent
1914
A union health center
The ILGWU's Union Health Center in New York City gave members medical, dental and prescription services. Cornell's ILGWU timeline dates it to 1914; other records at Cornell's Kheel Center give 1913.
Credit
the ILGWU
1986
Benefits away from the job
The AFL-CIO founded Union Privilege, now Union Plus, in 1986 as a nonprofit that offers benefits only to union members, retirees and their families, from credit cards and insurance to strike and hardship aid.
Credit
the AFL-CIO
2007
Fees that cost members
In January 2007 Sweden's new centre-right government sharply raised the fees for union-run unemployment funds, and over the next two years unions lost 8% of their members and union density fell from 77% to 71%, by Anders Kjellberg's count. In Finland, Petri Böckerman and Roope Uusitalo found that union density fell by more than 10 percentage points in under ten years, mainly because an independent fund offered unemployment insurance without union membership.
LessonWhere membership is tied to a service, a change in that service's price or provider changes membership.
Credit
Anders Kjellberg of Lund University; Petri Böckerman and Roope Uusitalo
2022
A pension rescue
On 8 December 2022 the US Pension Benefit Guaranty Corporation (PBGC) approved about $35.8 billion in Special Financial Assistance for the Central States Pension Fund, a multiemployer plan (one pension fund shared by many employers under union contracts) with 357,056 participants, under a program created by the American Rescue Plan Act of 2021.
Credit
the PBGC; the US Congress
Watch for
Contract answers for members
CustomGPT.ai, a chatbot vendor, says an unnamed North American union runs 28 AI assistants, one per collective agreement, that have answered 8,707 questions on topics such as pay, scheduling and military leave, with a 95.7% "satisfied answer rate" in the latest quarter, figures taken from its own usage records.
Achieved whenan independent test checks the answers against the agreements.
Claimed by
CustomGPT.ai, September 2026
Open
01
Pension health in plain words
For every multiemployer pension plan, a one-page summary in plain language of its funding level, contributions and outlook, drafted from the plan's public annual filing and checked line by line against the numbers in it.
The briefCredit · unclaimed02
Union legal help in data
A dataset of published UK employment tribunal judgments recording who represented the claimant (a union representative, a lawyer instructed by a union, another lawyer, or no one) alongside the claim type and outcome, published only as totals and not by person.
The briefCredit · unclaimed
Open 01 · the brief
Pension health in plain words
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
For every multiemployer pension plan, a one-page summary in plain language of its funding level, contributions and outlook, drafted from the plan's public annual filing and checked line by line against the numbers in it.
Why it is still open
The US Government Accountability Office reported in 2013 that participants often find required pension disclosures "overwhelming and confusing", and its recommendation that the Labor Department set readability criteria was closed as not implemented. The filings themselves are public: the Labor Department publishes Form 5500 datasets, including Schedule MB actuarial files, for 2009 onward. This research found no plain-language summary covering all multiemployer plans.
Built with what exists
The Labor Department's Form 5500 datasets; an AI model to draft each summary and a second pass that checks each figure against the filing (see the Note "Seven more flagged"); a readability checker; a credentialed actuary to review a sample.
First step
Draft summaries for 10 plans from the latest Form 5500 data and have an actuary check them.
Done looks like
Summaries for all multiemployer defined benefit plans in the latest filing year; an actuary finds no material error in a random 50; average reading level at or below US grade 8; in a test with 100 plan participants, more answer five comprehension questions correctly with the summary than with the plan's official notice.
Open 02 · the brief
Union legal help in data
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A dataset of published UK employment tribunal judgments recording who represented the claimant (a union representative, a lawyer instructed by a union, another lawyer, or no one) alongside the claim type and outcome, published only as totals and not by person.
Why it is still open
Unions give members legal help with tribunal claims, and the largest research dataset of tribunal judgments, CLC-UKET with about 19,000 cases, annotates facts, claims, references, outcomes, reasons and jurisdiction codes, but not who represented the claimant. This research found no dataset that does.
Built with what exists
Employment tribunal decisions published on GOV.UK; an AI model to pull out representation, claim type and outcome; two reviewers.
First step
Code representation and outcome in 100 recent judgments by hand.
Done looks like
10,000 judgments coded; reviewers agree with the model on representation for 95% of a 300-case sample; outcome rates by representation and claim type published, with a plain statement that represented and unrepresented cases differ in kind, so the figures show association, not cause.