Prospectr Turns Hours of Prospect Research into Minutes
Agentic AI on AWS | Case Study
Prospectr is a Euro-African sales and lead generation business. Its team sells by telephone against prospect lists it researches and maintains itself, and the quality of that research is a large part of what Prospectr offers its clients.
INDUSTRY
Sales and Lead Generation
CHALLENGE
Building a single account record took hours of manual research, and records went out of date within months. The work scaled only by adding people.
RESULTS
Research that took 2 to 3 hours per account now takes under 15 minutes, and 75% of records are completed without manual research.
KEY SERVICE
Agentic AI on AWS
Dudobi built two AI agents on AWS that cut Prospectr's account research from 2 to 3 hours to under 15 minutes, with 75% of records completed without manual research.
The Challenge
Building a single account record took hours of work across public sources: checking a company's technology, size and maturity, finding the right decision-makers, and judging whether it fitted the profile a client wanted. Records then went out of date within months, so calls were often made against information nobody fully trusted. Accounts were not scored consistently, and the reasoning behind a judgement was not kept, so it could not be checked or improved.
The work scaled only by adding people. That set a floor under what Prospectr could charge and a ceiling on how many campaigns it could run at once. Buying pre-enriched data would have removed the cost but also the advantage, since the research is part of what Prospectr sells. Without a faster way to do it in-house, growth meant hiring, and each campaign's margin would get thinner rather than better.
The Solution
Dudobi designed and built two AI agents for Prospectr, deployed in Prospectr's own AWS account in the Europe (London) Region. The Research Agent starts from a company name and a domain and plans its own research. It decides which sources to consult, such as web research, public technical signals and Companies House records, judges when it has enough evidence, and marks a record as unresolved rather than filling a gap with a guess. It returns a scored account record with the supporting evidence kept alongside it, and any record can be refreshed on demand.
The Sales Agent works with the sales team directly. Drawing on the Research Agent's intelligence and the conversation so far, it drafts sales plays and outreach for the team to review, working with HubSpot and Outlook. The agents never contact prospects themselves; every approach is still made by a person.
Both agents are built with the Strands Agents SDK and run in Amazon Bedrock AgentCore Runtime, using Anthropic Claude through Amazon Bedrock for reasoning and Amazon Titan for embeddings. Amazon S3 Vectors stores embeddings for semantic retrieval of past research, AgentCore Memory keeps context across sessions, and AgentCore Gateway gives the agents a single controlled path, over MCP, to tools such as Tavily web research, HubSpot and Outlook. Amazon Bedrock Guardrails apply content and sensitive-information policies. The team uses a web application hosted on Amazon S3 and Amazon CloudFront, with Amazon Cognito for sign-in, Amazon API Gateway and AWS Lambda for application logic, and Amazon DynamoDB for prospect records and workflow state. AgentCore Observability and Amazon CloudWatch trace every agent run, tool call and cost.
Because Prospectr works under various data privacy and security laws, personal data is held separately from company data, kept to what the task needs and given a defined retention period. Dudobi delivered the work from discovery and scoring design through build, testing with the Prospectr team and production launch, and continues to monitor the agents and refine scoring as the team feeds back corrections.
The Results
Research that used to take 2 to 3 hours per account now takes under 15 minutes, and 75% of records are completed without anyone doing manual research. Because records can be refreshed on demand, the team calls against current information rather than research that may be months old, and every score comes with the evidence behind it.
The result is that Prospectr's people spend more of their day selling and less of it researching, and the business can take on more campaigns without adding headcount in proportion.
About Dudobi
Dudobi is an AWS Advanced Tier Services Partner working with small and medium-sized businesses and nonprofits across the United Kingdom and South Africa. We design, build and run workloads on AWS, from managed infrastructure to agentic AI solutions built on Amazon Bedrock and Amazon Bedrock AgentCore. Dudobi has achieved Managed Service Provider accreditation and competencies including SMB and Nonprofit. Dudobi has been certified to ISO 27001 since 2014.