Built on Amazon Bedrock: How BlueOceanAI Is Redefining Marketing with Agents

Marketers typically rely on surveys and market research to understand brand perception, measure campaign effectiveness, and assess competitive positioning. These inputs help marketers position and craft advertising campaigns that deliver enhanced consumer experiences through highly personalized interactions for target audiences. For example, an enterprise brand might commission consumer surveys to test messaging concepts, hire agencies to conduct focus groups, and purchase research reports to track market share—which can cost $20,000–$400,000 per data source per year. When combined with additional costs for data processing, modeling, and strategic consulting that can range from $250,000–$1.2 million per year, traditional brand intelligence becomes prohibitively expensive. Moreover, a marketing campaign cycle can take 42 weeks, during which market conditions may have already shifted, making any insights less actionable by the time they’re implemented.

BlueOcean is transforming this process with fast, cost-efficient, and data-driven brand management and competitive analysis. It built a data acquisition pipeline on AWS that combines publicly available marketing data and competitive intelligence such as customer reviews, social media, market research, and industry reports. This data foundation powers Spark, BlueOcean’s framework of domain-specific AI agents that transforms how marketers interact with brand intelligence. Instead of waiting weeks and spending $50,000–$250,000 on traditional one-time reports, marketers can ask Spark direct questions such as “What themes make our brand most distinct in this category?” and get immediate recommendations for positioning and content strategy as well as competitive analysis.

But increasingly complex marketing workflows, such as building a 90-day, channel-specific campaign strategy around the strengths of a brand, were stressing the open-source large language models (LLMs) that underpinned BlueOcean’s prototypes, causing latency, unpredictable uptime, and data throttling. “We ran into a major roadblock as we tried to accommodate our customer growth rates and their LLM usage,” says Adam Carr, BlueOcean’s chief technology officer. “The Amazon Bedrock team worked with us to figure out how to move forward with Spark. Amazon Bedrock helped with capacity, but the key was exploring new ideas together in an undefined space.”

BlueOcean revamped its prompt engineering strategy to optimize performance across different models and use cases, facilitating faster processing and more accurate responses. Spark uses open-source multi-agent frameworks as well as Amazon Bedrock to connect to data sources and tools for agentic AI. Using Anthropic’s Claude in Amazon Bedrock, the company runs different FMs in parallel, reducing data processing time from 5–6 days to hours. It primarily uses Anthropic’s Claude 3.5 Sonnet and Anthropic’s Claude 3 Haiku to annotate historical data, while relying on Anthropic’s Claude 3.7 Sonnet hybrid reasoning model to power Spark on demand.

Spark provides pre-seeded prompts to help users get role-specific, relevant responses. A marketer can ask, “How can I outperform my competitors?” or “What are the pressing issues we need to address this week?” and other similar questions in natural language. “We’ve trained our AI agents using role-specific prompt libraries and proprietary brand data, making it possible for them to generate tailored outputs for different roles—such as distinct actionable recommendations for product marketers versus brand marketers,” says Liza Nebel, BlueOcean’s cofounder and president. “It’s about helping curious customers ask the right questions and drive them as efficiently as possible to value.”

Spark acts as an always-on strategist that coordinates with a team of domain-specific AI agents, each with specialties such as measuring brand performance, generating content, or interpreting perception. When marketers ask complex questions, Spark, powered by multi-agent frameworks as well as Amazon Bedrock, breaks them down into sub-questions and hands them off to different specialized agents that have access to specific datasets. These agents then collaborate to provide strategic recommendations, analysis that previously required weeks of manual research and multiple vendor reports.

BlueOcean’s AI agents scaled to process about 1.2 billion tokens in 1 month after Spark’s launch, which translates to answering over 10,000 marketing-specific questions using BlueOcean’s proprietary data. BlueOcean’s customers have seen 97 percent operational improvement, spending 2 hours to complete tasks that traditionally took about 5 days. “What’s unique is that we’re one of the first companies to bring a multi-agent framework into production at enterprise scale using AWS,” says Grant McDougall, BlueOcean’s cofounder and CEO.

Using Amazon Bedrock, BlueOcean has lowered its operating expenses by 21 percent. Its customers have also reduced their own analytics costs by 66–96 percent and removed data source costs, which BlueOcean includes in its offerings. Companies that previously could not afford analog methods of brand analysis can now use data as a central aspect of decision-making. And BlueOcean’s customers are seeing faster returns on their marketing investments, with some reporting payback periods (the time it takes for a marketing investment to recover its original costs) that are four times faster than traditional approaches.

“Anybody can use BlueOceanAI because the learning curve is as simple as knowing how to ask a question,” says Nebel. “Individuals feel more confident in their decisions and how they present ideas and strategic direction to stakeholders.”

Continuing to work alongside AWS, BlueOcean intends to use its unique datasets, which are accessible through Amazon Bedrock, for content creation. In addition, the company is building a new product that will proactively provide a stream of insights and predictive best next actions to take to marketers.

“We had a 5-year vision for the new age of Marketing, which is the single most important lever for driving sustained and durable corporate value,” says McDougall. “On AWS, we’ve developed our product and put it in the hands of marketers in 6 months.”