---
title: 'Be My Eyes: Conversational multimodal visual assistance through Be My AI'
slug: be-my-ai-visual-assistance
stable_id: d8b354365bb9e260
company: Be My Eyes
function_code: engineering
pattern_codes:
  - creator_to_judge
evidence_strength: verified
publication_tier: showcase
freshness: watch
reviewed_at: '2026-08-23'
updated_at: '2026-08-22'
source_quality_summary: 1 peer reviewed; publication outcomes are verified.
caveat_summary: >-
  Qualitative study; no causal time-saving or safety metric. Small purposive sample;
  limitations may change with model updates.
collections:
  - human-still-decides
  - negative-results
bundle_version: 1.0.0
bundle_fingerprint: sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93
canonical_url: https://brianletort.ai/transformations/be-my-ai-visual-assistance
---

# Be My Eyes: Conversational multimodal visual assistance through Be My AI

A blind or low-vision user can get immediate, private scene descriptions before choosing human assistance.

Function: Engineering design. Patterns: Creator to judge. Evidence: verified.

Freshness: watch. Reviewed: 2026-08-23. Updated: 2026-08-22.
Watch status: verify the cited source and deployment condition before reusing this case.

Source quality: 1 peer reviewed; publication outcomes are verified.

## Before

1. **Blind or low-vision user** — Contacts a sighted volunteer or trusted person and shares a camera view. (control: User explains intent and decides whether to act.)
2. **Human assistant** — Describes the visual scene and answers follow-up questions. (control: User judgment.)

## After

1. **User** — Captures an image and asks Be My AI for a description or follow-up detail. (control: User controls framing, prompting, and use.)
2. **User** — Evaluates the answer and either acts, reprompts, or escalates to a human for high-stakes or uncertain tasks. (control: Human escalation remains available.)

## Decision rights

The user decides whether to trust, reprompt, or escalate; humans retain action decisions.

## Exception path

Users compensate for hallucinations or misunderstood context and intent through spatial memory, auditory cues, personal judgment, reprompting, orientation and mobility skills, or human assistance.

## Outcomes

- **Demonstrated task scope in user study** (verified): Visual interpretation commonly required sighted assistance. → Fourteen users described Be My AI support across personal and social contexts.. Qualitative study; no causal time-saving or safety metric.
- **Observed failure modes** (verified): Human assistants can clarify intent and context interactively. → The study identified hallucinations, misinterpretations of social contexts, styles, and human identities, and failures to grasp or act on user intent.. Small purposive sample; limitations may change with model updates.

## Executive lesson

Accessibility value comes from giving the user control over when AI is sufficient and when a human is needed.

## Anti-pattern

Positioning probabilistic image description as safe navigation or identity verification.

## Questions for leaders

- Which tasks are explicitly out of scope?
- How quickly can users reach a human?
- How are hallucinations communicated accessibly?

## Sources

- [Beyond Visual Perception: Insights from Smartphone Interaction of Visually Impaired Users with Large Multimodal Models](https://doi.org/10.1145/3706598.3714210) — ACM CHI 2025