This is a smart, scalable answer to a widespread but under-recognised problem. By helping organisations catch ableist language before it goes live, Innervation has the potential to improve inclusion across thousands of workplaces and communications at once.

Infographic titled “The Intelligent Debiaser” showing a three-step process: scanning text for bias, AI redrafting to remove bias while keeping tone, and producing a transparent output with risk levels, explanations, and alternative wording options.

 

Innervation’s Ableism Debiaser tackles a problem most organisations do not even realise they are creating. Everyday language in job adverts, policies, marketing and public communications can quietly exclude disabled people, even when no harm is intended. Existing AI tools are not built to catch this kind of bias. They can spot spelling mistakes or overtly toxic language, but they miss the structural assumptions that make content ableist. The Debiaser uses a multi-agent AI system to review text against disability-inclusive language standards and explain what needs to change. It preserves the writer’s original voice while flagging risk and suggesting better alternatives. The impact is already strong. It won the G7 GovAI Grand Challenge in 2026 from more than 100 submissions, cuts manual review time by around 90%, and is being integrated into an HR platform with 50,000-plus users. This gives organisations a practical way to reduce exclusion at scale and build more accessible communications into everyday work.

Innervation’s Ableism Debiaser can 

  1. use best-practice guides and standards to identify ableist bias in user input, documents, or in the output of large language models, 
  2. redraft where required to eliminate the identified instances of bias, and 
  3. on a per-identification basis 
    1. categorize the risk (high/medium/low), 
    2. provide an explanation of why the original language contained bias, 
    3. provide alternatives in case the user prefers a different option, and
    4. maintain the writing style and tone of the original text across all edits.

The Ableism Debiaser can also be configured to identify and remove other forms of bias (sexism, racism, ageism etc.) if a best-practice identification guide exists.

Check out Innervation’s Multi-Agent Ableism Debiaser demonstration video.

Vote now for this finalist to win the Tech4Good People’s Award and register to join the online ceremony on Friday 3 July to see if they win!