How Pixieset Achieved 35% AI Feature Adoption with Amazon Bedrock
Amazon Web Services
Anthropic
OpenAI
Meta
Mistral
Cohere
Pixieset, a photography business service, used Amazon Bedrock to build an AI alt-text generator, achieving 35% adoption among applicable users. The feature addresses a tedious task, launched in four months, and drove subscription upgrades. Key principles included focusing on must-have features and letting users build trust incrementally.
Pixieset, an all-in-one photography business service hosting over 8 billion photos, used Amazon Bedrock to create an AI-powered alt-text generator for photographer websites. The feature, launched in early 2025, achieved 35% adoption among applicable users and remains in use sixteen months later. It generates image descriptions one at a time, allowing photographers to accept, edit, or reject suggestions before enabling auto-apply. The integration added a single API call to their existing event-driven pipeline, requiring no new infrastructure or GPU provisioning. The feature has had zero downtime since launch, thanks to cross-Region inference and retry logic with a secondary model. Pixieset chose Anthropic Claude 3.5 Sonnet initially but can swap models via Amazon Bedrock's unified API. When asked about the response, photographers praised the feature for saving time and improving SEO, with some upgrading their subscriptions. Pixieset identified three principles: categorize AI features as moats or must-haves, solve non-creative tasks, and let users build trust at their own pace.
- Abbreviations
- AWS = Amazon Web Services — Amazon Web Services
- EC2 = Elastic Compute Cloud — Elastic Compute Cloud
- GPU = Graphics Processing Unit — графический процессор
- LLM = Large Language Model — большая языковая модель
- SQS = Simple Queue Service — Simple Queue Service
- API = Application Programming Interface — программный интерфейс приложения
Source: AWS ML blog —
original
