Agentic AI / LLM Engineer for Dynamics 365 Sales Market Research Automation

I am conducting independent market research on how SDRs, BDRs, Account Executives, and other salespeople use Microsoft Dynamics 365 Sales. I am looking for an experienced AI/automation engineer to help me build a small Agentic AI market research prototype. The objective is not to create a large production application initially. I want to start with a focused prototype that demonstrates whether AI agents can reliably assist with research, verification, evidence collection, and interview analysis. I am particularly interested in someone who understands: •Agentic AI / AI agents •LLM workflow orchestration •LLM APIs •Microsoft Copilot Studio •Power Automate •Dynamics 365 / Dataverse •n8n or similar automation platforms •RAG, MCP, APIs, structured outputs, and human-in-the-loop workflows You do not need to use all of these technologies. I want you to recommend the simplest and most cost-effective architecture. Initial Prototype: I would like the first prototype to demonstrate approximately three workflows: 1. Salesperson / Company Research and Verification Input could be a salesperson's name, LinkedIn profile, company, or similar information. The workflow should help determine: •Current role and company •Whether the person is genuinely sales-facing •Whether they fit personas such as SDR, BDR, AE, Inside Sales, Account Manager, etc. •Whether there is credible evidence that the company or sales organization uses Microsoft Dynamics 365 Sales •Evidence classification such as: o Confirmed o Probable o Possible o Not Found •Sources supporting each important conclusion •Clear identification of uncertainty rather than inventing information The system should NOT assume that someone uses Dynamics 365 simply because their company uses Microsoft technology. 2. Interview / Research Transcript Analysis Given an interview transcript or research notes, automatically extract structured information such as: •Pain point •Job to be done •Frequency •Severity •Current workaround •Tools used outside Dynamics 365 •Time or business impact •Dynamics-specific problem vs general sales problem •Willingness to change/pay •Existing tools being used •Relevant verbatim quotes •AI/automation opportunity •Confidence in the finding •Follow-up questions 3. Cross-Interview Evidence Aggregation Across several interviews or research records, help identify: •Recurring pain points •Number of independent sources supporting a finding •Contradictory evidence •Weakly supported hypotheses •Dynamics-specific vs general sales problems •Important evidence gaps •Research questions that should be investigated next The system should help prevent premature conclusions from one interview or one online source. Important Requirements: Accuracy and traceability are more important than simply generating answers. I want workflows that: •Preserve source links/evidence •Clearly distinguish fact from inference •Flag uncertainty •Minimize hallucinations •Support human approval at important steps •Produce structured outputs that can later be stored in Dynamics 365/Dataverse or another database •Can eventually process larger research lists efficiently For the prototype, I am open to using a simple storage layer such as Google Sheets, Airtable, Excel, a database, or Dataverse. I do NOT require Dynamics 365 integration during the first prototype unless you believe there is a compelling reason to include it. What I Need From You: Before implementation, I would like you to recommend the architecture. Please explain: •Technologies you recommend •Why you chose them •What should be handled by AI agents vs deterministic automation •Where human approval should occur •How you would reduce hallucination and unsupported conclusions •How research evidence and sources would be retained •Estimated effort for the prototype •What you would build within the initial budget I prefer simple, maintainable solutions over unnecessary complexity. Initial Scope / Budget: This is initially a small paid proof-of-concept project. Target initial budget: Approximately $300–$500 I expect roughly 10–15 hours of focused work, depending on your rate and proposed architecture. If the prototype works well, this can become an ongoing project involving additional research agents, Dynamics 365 integration, outreach workflows, interview processing, dashboards, and other automation. Ideal Freelancer: You may be a good fit if you have hands-on experience building: •AI agents •Agentic workflows •OpenAI/LLM applications •Copilot Studio agents •n8n AI workflows •Power Automate workflows •RAG systems •MCP integrations •Structured LLM outputs •API integrations •Human-in-the-loop AI systems Experience with Dynamics 365 Sales, Dataverse, Power Platform, or Microsoft Copilot Studio is highly desirable. I value demonstrated implementation experience more than certifications. How to Apply: Please do not send a generic AI-generated proposal. Start your proposal with: "Research Agent" Then briefly answer these questions: 1. What similar AI-agent or LLM automation systems have you personally built? 2. Which architecture would you initially recommend for this project and why? 3. Would you use OpenAI, Copilot Studio, n8n, Power Automate, Python, MCP, Dataverse, or another approach? You do not need to use all of them. 4. How would you prevent the AI from presenting unsupported assumptions as verified research findings? 5. How would you retain sources/evidence so every important conclusion can be independently verified? 6. What could you realistically deliver within a $300–$500 prototype budget? 7. Please provide links or descriptions of one or two relevant projects you personally implemented. I am looking for someone who can think about the research problem and recommend the right solution—not simply implement whatever technology I mention.