Upload a PDF, ask a question, get an answer. That is what most document AI does, and for a single file it works. But almost no real question inside a company is about a single file. What did we promise this client, across the proposal, the two amendments, and the email where we agreed to the extra deliverable? Which of our carrier contracts include a fuel surcharge clause, and which of those expire this year? What was the approved budget for this project as of March, before the change orders?
Each of those needs several documents read together, with the entities connected and the dates kept straight. Ask a single-document tool and you get a reading assignment: open each file, ask each one, and stitch the answers yourself. That is the research phase again, with a chatbot in the loop.
Across documents
PlanRight reads the whole body of material a company has and extracts the facts and relationships into a knowledge graph: the people, companies, contracts, projects, and figures, and how they connect. So a question like which clients are on the old pricing schedule is answered from every contract at once, with each one cited. Multi-hop questions, where the answer to one part determines where to look for the next, are handled the way an analyst would handle them, by following the thread.
Across time
The graph is also temporal. It knows not just what a fact is but when it was true. That makes as-of questions possible: what was the scope of this project before the third change order, what did the rate card say when we quoted this job, what was the headcount on that account in January. A company's record is not a snapshot. It is a history, and the answer to most questions depends on which point in that history you are asking about.
Memory that compounds
The record also includes what the team learns while working. PlanRight remembers context, decisions, and preferences as they happen and recalls them when they matter, so the platform does not start from zero each conversation. The longer a company uses it, the more of its own working knowledge is in the graph, and the more useful every answer becomes.
What it feels like in practice
Instead of opening seven files, you ask one question and get one sourced answer that draws on all seven. Instead of wondering whether you remembered every amendment, you ask what changed. The record becomes something the team can interrogate, rather than something they have to get through.