Will generative AI replace human market and competitive intelligence researchers? Absolutely not! Humans (“knowledge workers”) apply judgment to a body of information to make decisions on behalf of the business. No machine does that.
But generative AI capabilities can provide a valuable first step for a researcher investigating a market and competitive intelligence question. That’s because a well-implemented generative AI program can distill down a large body of information to accurately answer a specific question – assuming the program is deployed against only reliable source material.
Which is precisely what Northern Light does with its generative AI “question answering” in SinglePoint™ strategic research portals.
How Northern Light utilizes generative AI for market research and competitive intelligence
- A user asks a market research or competitive intelligence question – for example, “What is Pfizer’s strategy in oncology?”
- Northern Light generates a search result using our search engine optimized for market and competitive intelligence applied exclusively against a client’s content collections. These may include primary and licensed secondary market research, authoritative web-based business news sources, respected industry journals and thought leader consultants, corporate financial reports, and relevant government databases.
- Northern Light gathers the document text from the 20 most relevant documents and sends it off to OpenAI or Microsoft Azure (client’s choice) with a prompt to briefly answer the user’s questions from each document. We specify that the large language model that responds to us is GPT 3.5-Turbo. (We do not send the document text or ask for answers from the public version of ChatGPT.)
- The answers generated are presented to the user, segmented by document, with the document cited and hot linked so the user can drill into the documents that appear to have the best answers.
- The answers are gathered and sent off to OpenAI or Microsoft Azure with a prompt to write an executive summary. The prompt specifies that every observation in the executive summary must cite the documents that contributed to the summary. The citations are also hot linked to support click through.
- The generative AI is instructed to only use the text from the documents we send to answer the question and provide the executive summary. No sources other than what we send in that specific prompt may contribute (e.g., no internet data). This ensures only high-quality, curated content generates the answers and the executive summaries.
One significant note from an enterprise IT perspective: communications with OpenAI (or Microsoft) are secure and private between Northern Light and OpenAI. They do not retain the content or user question once they have responded to our API call or use the content or user question to train models or anything else. Northern Light’s client is not identified as the source of the user question and the content publisher for the included text is not identified to OpenAI. (That knowledge is retained in the Northern Light portal application and need not be sent to OpenAI.) Northern Light does not retain the answers or the executive summaries and makes no use of them other than to help the user find the best documents to read. Nothing persists after the round-trip transaction (text to OpenAI, answers and summaries back to Northern Light) is complete. There is no exchange of prompts, content, answers, or summaries between Northern Light clients.
Generative AI will empower, not replace, knowledge workers
So the bottom line is knowledge workers should feel comfortable taking advantage of this exciting new generative AI technology in support of their market and competitive intelligence research activities. It’s an aid to, not a replacement for, what they uniquely can do.