[{"data":1,"prerenderedAt":199},["ShallowReactive",2],{"$f3nzgke2dhofmo":3},{"slug":4,"title":5,"date":6,"category":7,"kind":8,"description":9,"image":10,"imageAlt":11,"pinned":12,"downloads":13,"body":14},"handing-claude-a-data-room","We handed Claude a data room and our connector","2026-11-03","product-update","explanation","Claude can now start an impact valuation for you, from the documents on your computer to the brief and the results. We tried it the way a fund analyst would, with five public reports and five messages.","\u002Fblog-assets\u002Fclaude-data-room-card.svg","Five steps as rows, from choosing the documents to the committee summary, each marked with who does it, Claude, the analyst or ImpactAccounting.ai",false,[],{"frontmatter":15,"meta":16,"nodes":17},{},{},[18,22,25,30,33,36,111,114,118,121,124,127,137,140,147,151,154,157,161,164,167,170,174,177,183,186,190],[19,20,21],"p",{},"More of the people we work with spend their day in Claude. They draft memos there, read data rooms there, and would rather not open another tool to start an impact valuation. Until this week our Claude connector could only read published results. It can now also create a project, take its documents, start the analysis and settle its brief with you, all from the conversation.",[19,23,24],{},"To see how that feels, we played the part of a fund analyst. The company was Vestiaire Collective, the second-hand fashion marketplace, because it publishes its impact reports. The data room was a folder of five of its public PDFs. Claude, running in Claude Code, had our connector and that folder, and nothing else. We ran it against a test copy of the platform, so nothing here touched a client's data.",[26,27,29],"h2",{"id":28},"the-first-message","The first message",[19,31,32],{},"We wrote one paragraph: we are looking at Vestiaire Collective, this folder is their data room, please set up a project for 2025, pick the documents that matter, and ask before you start.",[19,34,35],{},"Claude began by reading the recipe the connector serves for this job, then opened each PDF to see what it was. It sorted the folder in a way an analyst would recognise:",[37,38,39,55],"table",{},[40,41,42],"thead",{},[43,44,45,49,52],"tr",{},[46,47,48],"th",{},"File",[46,50,51],{},"What Claude made of it",[46,53,54],{},"Kept",[56,57,58,70,80,90,101],"tbody",{},[43,59,60,64,67],{},[61,62,63],"td",{},"Impact Report, 2026 update",[61,65,66],{},"The only document with 2025 figures",[61,68,69],{},"Yes",[43,71,72,75,78],{},[61,73,74],{},"Impact Report 2025",[61,76,77],{},"Mostly 2024 data, but it holds the method for what resale avoids",[61,79,69],{},[43,81,82,85,88],{},[61,83,84],{},"Non-financial performance declaration",[61,86,87],{},"Scale, risks, social and governance data",[61,89,69],{},[43,91,92,95,98],{},[61,93,94],{},"Impact Report 2025, French edition",[61,96,97],{},"The same report in French",[61,99,100],{},"No",[43,102,103,106,109],{},[61,104,105],{},"Impact Report 2023",[61,107,108],{},"Replaced by the later reports",[61,110,100],{},[19,112,113],{},"Two of the reports were too heavy to send as they were, so it asked before making lighter copies.",[26,115,117],{"id":116},"the-second-message","The second message",[19,119,120],{},"We agreed to the lighter copies, added that the analysis should not count the carbon credits Vestiaire sells as an impact of their own, and told Claude to start once the files were in.",[19,122,123],{},"Claude compared the copies' text with the originals, uploaded the three files, started the analysis, and checked on it every few minutes.",[19,125,126],{},"About ten minutes later the analysis had read the documents, researched the company and drafted its brief. The brief is where the analysis asks you to confirm what it will assess before it models anything: its scope, the impacts on each group of people, the figures it knows, and where the company is heading. Claude presented it in that order, in plain words, with the two questions the analysis could not settle on its own:",[128,129,130,134],"ul",{},[131,132,133],"li",{},"Without Vestiaire, what would buyers buy instead? The recommended answer was mostly new fashion, because the company's own buyer survey says most orders replace a new purchase.",[131,135,136],{},"Without Vestiaire, what would sellers do with their pieces? The recommended answer was other resale channels, since luxury items already have them.",[19,138,139],{},"It also pointed out what was missing: no document gave the number of orders in 2025, and nearly every impact scales with it.",[19,141,142],{},[143,144],"img",{"src":145,"alt":146},"\u002Fblog-assets\u002Fclaude-data-room-brief.png","Claude presenting the brief: the lines of activity in scope, the seven effects with who each one changes, the figures the documents give and the ones they don't, where the company is heading, and the two questions",[26,148,150],{"id":149},"the-third-message","The third message",[19,152,153],{},"We left the brand partnership business out, kept the rest, accepted both recommendations and left the missing figures blank. Claude sent the brief with our instruction and stopped there, as asked.",[19,155,156],{},"The analysis then modelled the company for nearly an hour and a half and published its first result on its own, as a first analysis does.",[26,158,160],{"id":159},"the-committee-summary","The committee summary",[19,162,163],{},"When we came back and asked for a one-page summary for an investment committee, Claude read the published results, the pathways, the outlook and the recipes our analysts wrote for a board pack and a due diligence, and wrote the page.",[19,165,166],{},"This was a test run that no expert has reviewed, so we leave its figures out. The shape of the answer is the useful part. It opened with a verdict and a range, then showed where the value comes from and how concentrated it is. The largest share rested on two estimates about what buyers save, and on an order volume the model had to estimate. The summary said so plainly, and turned it into five requests for the company: completed orders and average order value, the buyer survey behind the replacement rate, a seller survey, the method behind its own avoided-emissions figures, and its headcount and taxes.",[19,168,169],{},"A short follow-up added the release's six ranked risks and its opportunities to the page.",[26,171,173],{"id":172},"what-it-took","What it took",[19,175,176],{},"Five messages from the analyst, about two dollars of Claude usage, and about an hour and forty minutes from the first message to the committee page, nearly all of it the analysis doing its work. The connector did what it is meant to do: the documents went in as documents, every figure on the page was read from the published model or worked out from it, and the decisions in the brief stayed with the person.",[19,178,179],{},[143,180],{"src":181,"alt":182},"\u002Fblog-assets\u002Fclaude-data-room-connector-calls.png","The connector calls behind the five messages: reading the recipe, creating the project, adding and confirming three files, starting the analysis and following it, answering the brief, then reading the published results",[19,184,185],{},"The image of the brief is a shortened excerpt of the trial's transcript; the trial ran in Claude Code, in a terminal.",[26,187,189],{"id":188},"try-it","Try it",[19,191,192,193,198],{},"Connecting takes a minute: ",[194,195,197],"a",{"href":196},"\u002Fblog\u002Fconnect-impactaccounting-to-claude","Connect ImpactAccounting.ai to Claude"," walks through it. Then give Claude the folder with your documents, in the Claude desktop app or in Claude Code, and ask: \"Value this company: pick the documents that matter and start the analysis.\" The connector tells Claude to ask before it creates anything, before it starts a run, and whenever a decision is yours.",1793806704605]