Build multi step n8n (or similiar) + LLM “research to outreach” system

I am looking for a skilled automation or AI developer to build a multi step, multi prompt, agent style workflow using n8n or a similar orchestration tool plus OpenAI and Claude. The system will take a list of target banks or contacts and automatically: • Research the target company, role, region • Search the latest news in the target’s relevant domain credit, wealth, AI, investment banking, etc • Search regional insights relevant to that domain • Research internal information about the bank public filings, annual reports, programs, training initiatives, strategic priorities, leadership changes, etc • Combine and summarise all findings • Generate a personalised 3 to 5 step outreach sequence using our business context plus real facts from research • Output everything clearly in text or JSON • Allow multiple passes repeat research anytime for updated info This is not a single prompt task. This must be a proper multi step agentic workflow. What I Will Provide To You You will receive: • A sample spreadsheet of target banks plus names plus roles • My business context who we are, what we offer, tone of voice • My prompt templates for research summarisation and outreach copy and fallback behaviours • Examples of the type of output I want • Any API keys required OpenAI, Claude, Search API, etc What You Will Build 1. Multi agent or multi prompt research pipeline The workflow must: • Call OpenAI with agent mode search tools • Call Claude separately with the same research request • Perform web search on the target institution, role or department, sector specific news credit, wealth, AI, etc, and regional insights GCC, MENA, MEA • Read and extract data from recent news, annual reports, press releases, public filings, published training programs, leadership announcements • Store the collected research temporarily inside n8n or send it to a Google Doc for transparency 2. Summarisation stage A second prompt should: • Clean the research • Summarise it • Remove irrelevant or duplicated info • Identify the top insights that matter for outreach 3. Outreach generation stage A third LLM call should generate: • A 3 to 5 step outreach sequence • Each message must reference actual insights from research • Tone must follow my business context • Clear, simple, professional • End goal is securing a meeting 4. Fallback logic If research sources return little or no data: • Use a safe fallback summary • Create a simpler outreach sequence 5. Output formatting Must produce a clearly structured summary, insights list, and message sequence. Format must be reusable and easy to export. 6. Documentation You must provide: • A clear explanation of how the workflow works • How to update the prompts • How to add new contacts • How to re run research • How to swap in new LLM models Skills You Must Have • Strong experience with n8n or Make or similar automation tools • Experience with agent style LLM workflows • Ability to integrate OpenAI, Claude, Search APIs • Experience with scraping and HTTP requests • Ability to design multi prompt pipelines, not single prompt shortcuts • Clear English communication • Ability to build repeatable and maintainable workflows To Apply, You Must Answer These Questions Please answer these in your proposal: Explain exactly how you would build the multi step workflow research then summarisation then outreach. List the tools or APIs you would use for search Google, SerpAPI, etc. Tell me what you need from me to start prompts, sample data, access, etc. Explain how you will run OpenAI and Claude and then combine the results. Share an example of similar work you have done AI agents, automation, research bots, multi step workflows. Explain how you would store or display the raw research Google Doc, JSON, n8n datastore. If a proposal does not answer all six questions, it will not be considered.