AI-powered screen reader with agentic functionality
Got what it takes to build the future of assistive technology?
What are we loooking for?
We are looking for a team – a startup, a group of independent experts, or a research team – ready to work on a genuinely exciting and meaningful AI use case. You’ll build a working prototype in three months, with direct access to one of Germany’s largest industrial corporations, full project management handled by boOst, and a real shot at follow-up projects if your solution lands well.
Application Process
We are running this project as part of our Technology Sprint concept, a focused, industry-connected delivery format. A major German industrial corporation has commissioned the development of an AI-powered screen reader prototype that goes beyond what classical screen readers can do today. It should enable semantic understanding, image interpretation, voice-based control and agentic task execution.
DURATION: 3 months (approx. 12 weeks) | 2- 3 days/week | Kickoff early September 2026
COMPENSATION: To be agreed based on your proposal and project scope
FORMAT: Remote-first with an onsite kickoff
TEAM SIZE: One team; consortia explicitly welcome
UPON POSITIVE EVALUATION: Follow-up projects with the industry partner are possible
Problem Statement
It’s 2026. AI can write code, generate images, and hold conversations. Yet for millions of blind and visually impaired people, the daily reality at work looks like this: navigate a document with hundreds of keyboard shortcuts, hope that someone remembered to add alt texts to the diagrams, and manually trigger every single action , one keystroke at a time. Screen readers – the assistive technology that millions of blind and visually impaired employees depend on every day – haven’t fundamentally changed in decades. They read out what’s on screen, linearly, structurally, without understanding. In a world of dynamic web apps, embedded visuals, and collaborative tools, that’s no longer good enough.
For companies, this means a real productivity gap for affected employees. For you, it means a wide -open space where multimodal AI, agentic systems, and smart UX design can have a genuine and meaningful impact.
Objective
The goal of this sprint is to develop a functional prototype of an AI-powered screen reader that demonstrably closes these gaps.
The core hypothesis: multimodal AI and agentic interaction patterns can transform the screen reader from a passive reading tool into a goal-oriented work assistant.
The prototype should outperform classical screen readers in four areas:
- Semantic understanding: summarising and contextualising content rather than reading it out linearly
- Image & diagram interpretation: describing visual content meaningfully, even without alt texts
- Voice-based control: natural language commands via push-to-talk, replacing complex keyboard shortcuts
- Agentic task execution: carrying out multi-step actions in web forms and standard office workflows on the user’s behalf
Expected Result
The prototype is not expected to replicate the full features of established screen readers. The focus is on demonstrating clear improvements over classical tools.
- Working prototype: demonstrated across browser, Office, PDF, Outlook
- Two end-to-end demos: document context + agentic web interaction
- Technical documentation: architecture, data flows, privacy & security, limits
- Short evaluation note: first user tests + observations
Conditions
boOst provides
- Single point of contact . We handle communication and coordination with the industry partner
- Project management: milestone tracking, blocker management, regular check-ins
- Access to internal experts at the industry partner for interviews and reviews
- Methodological support, sprint coaching, and quality assurance
- Moderation of all touchpoints: kickoff, 2 interim presentations, final presentation
Industry partner provides:
- Clear requirements, criteria, and KPIs
- Internal contacts and subject matter experts
- Access to relevant data, systems, and test environments
- Structured feedback
Application Process
Proposals are initially submitted to boOst; following review by boOst and the industry partner, an NDA will be signed and the final proposal addressed directly to the industry partner.
The following must be submitted by August 06, 2026:
- A brief introduction of your team and relevant experience
- An overview of which competencies you cover with supporting evidence where available (references, demos, publications, links)
- Your assessment: which part of the solution do you see yourself building, and how much of the overall scope could you cover?
- A rough solution concept or initial ideas on your approach
- A proposal outlining your offer
More Information
If you have any further questions, please contact:
