Open Bar at mid-year: housing, employment and immigration triage, by the numbers

Six thousand one hundred. That is the number of paralegals, lawyers and volunteers using Open Bar at the end of June, across 410 legal-aid clinics, public defender offices and pro-bono programmes in 31 countries. A year ago it was 4,200. The Lagos cohort that began last autumn is now the second-largest national group after the United States.

This is a mid-year field note, which means counts rather than conclusions. The counts come from the aggregated, non-content telemetry clinics opt into. Where a number is soft we say so.

What comes through the door

Between January and June, AI Interview handled roughly 48,000 intake conversations for Open Bar organisations. That is intake, not matters: many conversations end in a referral elsewhere, and some callers ring twice.

Three categories account for 86% of them. Housing is 38%: eviction notices, deposit disputes, disrepair, and a rising share of informal tenancies with no written agreement at all. Employment is 27%: unpaid wages, dismissal, and injury at work. Immigration is 21%: status applications, appeals, family reunification, and detention. Family matters are 8%, consumer debt 6%.

The mix differs sharply by place. In the Lagos cohort, housing and employment together exceed 80% and immigration barely registers. In the EU clinics, immigration is the largest single category. The US public defender offices use Interview mainly for initial client contact in criminal matters, which we count separately and which is not in the 48,000.

Twenty-nine of Interview’s 38 languages were used at least once. The top five by volume are English, Spanish, French, Arabic and Yoruba. Eleven languages account for 95% of calls.

Housing and employment: deadlines and missing documents

The feature housing clinics use most is not a review feature. It is the limitation-period maths, repurposed. An eviction notice carries a response deadline; a deposit claim carries a filing window; a disrepair complaint carries notice periods before escalation. Interview computes the relevant dates per jurisdiction and puts “days remaining” at the top of the brief. Clinic supervisors tell us it changed how they order the morning’s work.

Review is used on tenancy agreements rarely, perhaps one in twelve housing intakes, and when it is used the issue is almost always the same: a clause purporting to waive a protection the tenant cannot waive.

Recall is used on the clinic’s own letters and precedents. A clinic with three years of history in Recall answers “what did we send the last time this landlord did this” in a few seconds, which matters when the landlord is the same in forty cases.

Wage and dismissal intakes share a different problem: the caller rarely has the documents. Interview was adjusted in March to collect what the caller does have (messages, photographs of rotas, bank statements) and to list what is missing, with the limitation clock running against the date the caller gives, not the date the documents arrive.

Draft is used here more than in housing, for demand letters in the clinic’s own voice. Clinics describe these as the first thing a volunteer produces on their first day that goes out without rewriting.

Immigration: the hardest calls

Immigration intakes are the longest, at a median of 31 minutes, and the most often abandoned, at 18% against 9% overall. They are also the calls where the empathy layer, which we wrote about last month, does the most work. Abandonment has fallen from 24% in the second half of last year, and we attribute part of that to the layer and part to clinics learning when to switch a caller to a person.

Twelve languages carry most of the immigration volume, and four of them are among the languages in which the layer’s phrasing has not yet been reviewed by a native-speaking clinician. We are not comfortable with that and are working on it.

What clinics ignore

Trial & Arbitration Simulation is used by 2% of Open Bar organisations. Legal aid is not where moots happen, and we do not expect this to change.

The trace view in Recall is opened in fewer than one in twenty queries. Clinics trust their own history and do not need to audit it the way a commercial firm audits a precedent pool.

Draft for contracts is nearly unused. Draft for letters is used daily. We built one and clinics adopted the other.

Playbook tuning in Review is untouched by most clinics, which run the default public-interest playbook we maintain. That is reasonable; nobody at a clinic has time to tune severity tiers.

What we changed

Phone-first, because that is what Lagos taught us. An offline-readable brief, so a paralegal with no signal in a courthouse corridor can still read it. Yoruba and Hausa support in Interview’s intake prompts, reviewed by native speakers. A shorter intake path for repeat callers, which the per-client vaults in Interview’s June release now make possible.

We are testing summaries delivered by text message for callers who cannot receive email. It is not ready, and the privacy questions are not simple.

The usual caveats: these are our counts, from organisations that opted in, using category labels that vary from clinic to clinic. The overall picture is consistent enough that we are willing to publish it. The details, less so.