Legal-aid groups test AI video tools for language access and staff training, stress human review

Legal-aid practitioners and a vendor onstage at a February workshop showed how AI video tools can expand language access and streamline staff training, but said human review and trusted voices remain essential.

Zane, project manager for the eviction right-to-counsel team at the Legal Aid Society of the Middle Tennessee and the Cumberlands, opened the session and framed the group’s aim: use AI to supplement—not replace—legal personnel while widening reach. “We by no means want to replace any existing legal personnel,” he said, noting ethical concerns that include language access, client trust and environmental impacts.

Adam Nastofsky, founder and CEO of Briefly, urged attendees to consider trade-offs. “Are there any ethical issues with not using AI for this?” he asked, arguing that if AI can reach many more people it may be ethically preferable to use it in some contexts. Panelists said the right approach depends on risk tolerance and on building review steps for higher-risk material.

The session included demos. Adam used a dubbing workflow to translate an English interview into Spanish in under 10 minutes, then edited sentence-by-sentence with a bilingual reviewer. He described the tool as a fast human-in-the-loop option for language access, saying the output is “pretty good” but benefits from quick human corrections.

Jillian Beach, communications and community engagement manager at Legal Aid of the Bluegrass, described the KYJustice project (kyjustice.org), funded initially by a 2019 TIG grant and a later adoption/expansion grant. She said video was essential to help community partners—clerks, librarians and shelters—adopt and promote the site. For external-facing content they tested AI voiceovers but “made a deliberate choice to step back from AI” and used a human voice actor (a former legal-aid client) because the synthetic voice sounded less credible to audiences.

Panelists described internal uses where AI video can offer clear benefits: converting dense HR policies and onboarding materials into short, locked videos that staff must watch; embedding short explainer clips inside pro se forms so users get targeted help while completing multi-page applications; and preserving institutional knowledge in editable video formats. Zach (senior deputy director of client services at Legal Aid of Middle Tennessee) said they now produce many outreach materials in multiple languages and embed short videos at relevant form pages to reduce confusion.

On cost and production, presenters said many tools use credit models: a starter plan they demonstrated costs roughly $18 per month and provides a pool of minutes that treat translated versions as separate videos. Adam said some text-to-voice tools (he mentioned 11 Labs) can require significant manual tuning for pronunciation and emphasis and may be more time-consuming than hiring a voice actor for highly polished, public-facing pieces; but he acknowledged useful use cases where scale and rapid accessibility matter.

Speakers recommended a human-in-the-loop workflow: subject-matter experts create primary content, LLMs or editing tools simplify language and tone, and bilingual reviewers or staff check AI translations before publication. Zach described one concrete pro se example: embedding a two-minute affidavit-of-indigency explainer on page 7 of a 20-page restoration-of-rights form so users can decide whether to file to waive fees.

The panel closed with offers for follow-up and contact information. Presenters said organizations should test tools against their own risk thresholds, prioritize human review where errors could cause harm, and use AI where speed and scale can materially improve access to justice.