Customer support · Production

Brainly builds Snap to Solve with Google Vision AI, achieving 70% satisfaction and 6x photo query engagement

The problem

Students found typing queries on smartphones cumbersome, and Brainly needed a multilingual AI solution supporting users across 35 countries — alternatives they evaluated focused mainly on English and a few additional languages.

First attempt

Other AI solutions evaluated by Brainly focused mainly on English and a few additional languages, failing to meet Brainly's multilingual requirements across 35 countries.

Workflow diagram · grounded in source
1
Student snaps photo of question
trigger
“with the Brainly app, users simply snap a photo of the question”
2
Vision AI OCR extracts text
ai_action
“Vision AI's optical character recognition (OCR) analyzes the image's content to extract words and sentences”
3
Instant answer or fallback search
routing
“If there's a perfect match for the query, Brainly delivers an instant answer within seconds, if not, the app falls back to a full text search, combing through Brainly's entire knowledge base to find a selection of relevant answers”
4
TensorFlow ML finds instant answers
ai_action
“Using TensorFlow's Multilingual Universal Sentence Encoder, the team developed an in-house solution to find instant answers for queries”
5
Answer delivered within seconds
output
“Brainly delivers an instant answer within seconds”
Reported outcome

Snap to Solve achieved 70% user satisfaction, drove 6x more engagement than typed queries, and boosted paid subscription numbers; improving machine learning algorithms with TensorFlow yielded 10x more instant answers.

Reported metrics
user satisfaction with Snap to Solve70%
Photo query engagement vs typed queries6x
instant answers increase via TensorFlow10x
Paid subscription numbersboosting paid subscription numbers
Reported stack
Vision AIOCRTensorFlowML KitAndroid
Source
https://cloud.google.com/customers/brainly
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Snap to Solve achieved 70% user satisfaction, drove 6x more engagement than typed queries, and boosted paid subscription numbers; improving machine learning algorithms with TensorFlow yielded 10x more instant answers.

What tools did this team use?

Vision AI, OCR, TensorFlow, ML Kit, Android.

What results were reported?

user satisfaction with Snap to Solve: 70%; Photo query engagement vs typed queries: 6x; instant answers increase via TensorFlow: 10x; Paid subscription numbers: boosting paid subscription numbers (source-reported, not independently verified).

What failed first in this deployment?

Other AI solutions evaluated by Brainly focused mainly on English and a few additional languages, failing to meet Brainly's multilingual requirements across 35 countries.

How is this customer support AI workflow structured?

Student snaps photo of question → Vision AI OCR extracts text → Instant answer or fallback search → TensorFlow ML finds instant answers → Answer delivered within seconds.

Related customer support cases