What challenges did Bancolombia face before AI implementation?

Marketing teams faced long, costly iterative processes. They had difficulty maintaining context and consistency across creative variations. They had limited control over AI-generated outputs when creating campaign visuals.

What was the impact of the AI solution?

The implementation streamlined the creative workflow. It reduced the time spent generating high-quality images. It improved brand consistency and gave users greater precision and control over outputs.

Source

Bancolombia marketing image generator using generative AI with Amazon Bedrock, Amazon OpenSearch Serverless, Amazon Titan Multimodal Embeddings, and Amazon Nova

Use case typeContent generationUpdated Feb 4, 2026

Bancolombia built a marketing campaign image generation system that uses historical campaign assets to improve new creative outputs. The workflow ingests reference images, generates descriptions and embeddings, stores them in Amazon OpenSearch Serverless, and searches for similar assets to guide prompt generation and image creation.

Organization: Bancolombia
Industry: Professional Services
Location: Colombia
Published: February 2026

Reported outcomes

Strategic outcomes

Cost efficiencyStreamlined creative workflowOther strategic outcomeReduced image generation timeCustomer experience & trustImproved brand consistencyRisk & complianceGreater control over AI outputs

Primary read

Use case focus

Showing 3 of 3

1Content generation
2Search modernization
3Workflow automation

Marketing teams faced long, costly iterative processes.
They had difficulty maintaining context and consistency across creative variations.
They had limited control over AI-generated outputs when creating campaign visuals.

The solution uses Amazon S3, AWS Step Functions, AWS Lambda, Amazon Nova Pro, Amazon Titan Multimodal Embeddings, Amazon OpenSearch Serverless, Amazon Bedrock, and Amazon Nova Canvas.
It indexes historical campaign assets, performs similarity search filtered by campaign objective and target audience, and enriches prompts with reference-image descriptions before generating new marketing images.

Technologies

Amazon Bedrock Amazon OpenSearch Serverless Amazon Titan Multimodal Embeddings Amazon Nova Pro Amazon Nova Canvas AWS Step Functions AWS Lambda Amazon S3 Amazon API Gateway

The implementation streamlined the creative workflow.
It reduced the time spent generating high-quality images.
It improved brand consistency and gave users greater precision and control over outputs.

Architecture

A web UI triggers an API Gateway and AWS Step Functions workflow. Lambda functions use Amazon Nova Pro to describe reference images, Amazon Titan Multimodal Embeddings to create image embeddings, and Amazon OpenSearch Serverless for vector similarity search over historical campaign assets. Selected reference-image descriptions are injected into an enhanced prompt, and Amazon Nova Canvas generates the final marketing image.

Sources & evidence1

Groundedness: 4/5Type: Blog PostPublished: Feb 4, 2026Publisher: AWSEvidence: VendorConfidence: Medium

Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence