Portfolio company ESG data collection where each company confirms prefilled values and no outlier reaches an LP unexplained
Each company confirms prefilled values, and no outlier reaches an LP unexplained.
See one case, screen by screen ↓Portfolio Company ESG Data Collection is ESG data management software for a GP's annual portfolio company ESG data collection: it collects, checks and benchmarks the ESG metrics private portfolio companies report each year, such as the EDCI metric set. Seven AI agents prefill values from documents the company already shared, cited to the page, follow up in the company's language and find the likely cause of every outlier. Each company confirms its own values, an analyst prepares each fund's dataset and the ESG lead approves it before any LP sees it.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
Portfolio company ESG data collection starts every spring, when a GP's ESG team asks each portfolio company for the same industry metric set — the EDCI metrics of the ESG Data Convergence Initiative, which more than 9,000 private companies now report: Scope 1 and 2 emissions, electricity and renewable share, a decarbonization plan, women on the board and in the C-suite, recordable injuries and days lost, net new hires, turnover and the engagement survey. It goes out as a blank survey or a spreadsheet to a CFO who forwards it to facilities, HR and finance, and the answers come back in different units, on different boundaries and definitions from last year.
Then the chase begins: analysts email, deal teams are asked to nudge, and a campaign can take up to twelve weeks while the industry submission, the LP report and, for EU funds, the SFDR principal adverse impact statement all fall due between April and June. Cleaning happens by hand in spreadsheets with no audit trail. One site billed in kilowatt-hours and typed as megawatt-hours can make a company's emissions intensity jump 340%, and in private equity ESG reporting a wrong number in an LP report reads as a misleading ESG claim.
Estimated split for one fund's campaign, by hand and with the solution. The agents' reading, checking and drafting runs in hours; the days that remain are companies confirming, deal teams explaining and the ESG lead approving.
Seven agents prefill, check, chase and benchmark every company; the company confirms and the ESG lead approves.
Quillfield Capital's private markets arm is GP to six funds holding 120 portfolio companies that report in 20 languages from 33 countries. The 2027 campaign, for FY2026 data, launched on February 1; companies confirm by Friday, March 26. It is Tuesday, March 16, 9:40 a.m. in New York. Here is what happens next, screen by screen, in the working solution.
Claire's home shows 71 of 120 companies confirmed, 86% of values prefilled from documents (1,263 of 1,464), 23 outliers to explain and 0 of 6 fund datasets approved. Top of Needs you: Fennick Logistics, emissions intensity up 340%, the last open item in Quillfield Capital Partners V (QCP V). She types “Why is Fennick up 340%?” and the Outlier agent answers with the bill and the tracker cited: the Laredo cross-dock's 9 bills total 102,350 kWh, the tracker lists 102,350 MWh; corrected, intensity is 25.9, down 3.5%.
“The agents found kWh entered as MWh. The deal team explained; the correction is with Fennick’s CFO. It is the last open item in QCP V.”
The confirmation grid puts companies down the side and the 13 metrics across the top: blue cells prefilled by agents and waiting for the company, green confirmed, gray missing, red outliers. In the European Mid-Market III fund, Atelier Dumesnil's electricity reads 3,391 MWh, prefilled at 92% confidence from page 2 of its French electricity report. The side panel shows the original sentence — “La consommation d'électricité s'élève à 3 391 MWh” — against last year's 3,332 MWh, a change of 2%.
Prefill agent: below 90% confidence it leaves the value for the company to answer; it leaves a value empty rather than guess.
Companies never see the grid. This is what Brandauer's CFO sees in Quillfield's company portal: the form in German, each value already filled, each with its source — Stromverbrauch 27,640 MWh from the Nachhaltigkeitsbericht, Seite 14 — and two buttons, Bestätigen or Korrigieren. Six of 13 are confirmed. The renewable share of 61%, up from 18% because both plants buy guarantees of origin from January 2026, is marked as under review by Quillfield.
“Only the company confirms its values.”
Fennick's facilities manager, Rob Haldane, uploaded an energy tracker on March 12: 124,728 MWh of electricity. The Carbon agent computed Scope 2 at 46,274 t against 8,165 t last year. The Outlier agent compared every value with the confirmed FY2025 dataset and 86 US logistics peers, flagged emissions intensity up 340% — 117.9 t per $M of revenue against a peer median of 31.2 — and routed it to the QCP V deal team.
Likely cause: “Units: one site's bills are in kWh but were entered as MWh (Laredo cross-dock).”
The outlier page shows the jump against the peer band, then where it came from: Fennick's tracker lists the Laredo cross-dock at 102,350.00 MWh, while the nine Rio Bravo Electric Cooperative bills for the site total 102,350 kWh. The other 13 sites are billed in MWh by Gulfline Energy Services and total 22,377.65 MWh. 22,377.65 + 102.35 = 22,480 MWh. Laredo opened in April 2026, on a co-op that bills in kilowatt-hours.
Outlier agent: never changes a value; always names a likely cause and its source.
Claire replays the check. Compared FY2026 with FY2025: intensity 26.8 to 117.9. Placed it against 86 US logistics peers: median 31.2, upper quartile 47.8. Ruled out growth: revenue up 5.9%, staff up 3.2%, sites 13 to 14. Re-read the bills behind electricity. Recomputed Scope 2: 22,480 MWh × 0.371 = 8,340 t. Routed to the QCP V deal team, Graham Holt and Mateo Ruiz.
Carbon agent: recomputes from activity data with a versioned emission factor set (v2025.1) and never overwrites a company figure.
Mateo Ruiz checked with Fennick's facilities team at 8:55 a.m. and agreed with the correction; at 8:57 the Follow-up agent put it in Dana's form with the bills attached. Signed in to the company portal, Dana sees only Fennick: “One value needs your confirmation” — electricity 124,728 MWh struck through, 22,480 MWh proposed, Scope 2 8,340 t. She confirms, ticks the attestation and signs. The outlier closes under rule R-07.
Rule R-07 closes an outlier on its own only when the agents propose, the deal team explains and the company confirms.
Tomás prepared the QCP V dataset at 9:12 a.m. — 24 companies, 312 values — with one open item, Fennick's electricity. With Fennick confirmed, all six checks pass — every company confirmed and signed, no open outliers, no missing values, every value with a source page or a signed company entry, no value changed after confirmation, units and boundaries checked. Claire approves with reason code DS-01. QCP V's Scope 1 + 2 is 148,589 t, up 3.6% on FY2025 — not 186,523 t, up 30%.
“Approval opens when every check passes. The preparer cannot approve.”
The Reporting agent drafts the QCP V section of the annual LP ESG report from the approved dataset — climate, data quality, intensity, people, governance — with every number linked to its company confirmation and source page, including a plain note that one company's electricity was corrected before reporting. Claire asks where 148,589 t comes from: 62,595 t Scope 1 and 85,994 t Scope 2 across 24 companies, with Fennick at 8,340 t after its CFO confirmed.
“I never send anything to an LP.” Investor relations edits and marks the draft ready; compliance releases it.
The audit trail holds the whole case: the Outlier agent's flag, the Prefill agent's re-read of nine bills, the Carbon agent's proposal with factor set v2025.1, Mateo's explanation, Dana's confirmation and attestation, the R-07 close, Tomás preparing as maker and Claire approving with DS-01 — each with who, which agent version, the evidence, the reason code and a hash. Internal audit can verify the chain and export an evidence pack.
The same campaign, seen by five of the people who carry it — what the spring looked like with blank surveys, and what it looks like now.
Seven agents prefill, check, chase and benchmark every company; the company confirms and the ESG lead approves.
Scopes each fund's campaign from the fund administrator's company list, last year's campaign and the CRM: companies new and exited, the metric set per fund, contacts, schedule and languages. It also proposes adding a company acquired mid-campaign, for the ESG lead to approve.
Fills each metric from documents the company already shared — utility bills, audited accounts, HR and safety reports, board minutes — in their own language, with the unit and the page, converting units, and re-reads documents when another agent asks.
Redoes the Scope 1 and Scope 2 emissions calculation from activity data with versioned emission factors, checks the company's figure and proposes a correction when the two differ.
Sends confirm-or-correct forms and reminders in each company's language, asks only for what is still open, predicts who will miss the deadline, asks the deal team to call and proposes escalation to an operating partner.
Compares every value with last year, sector peers and the company's own other values; above a 50% change, three times the peer median or an impossible value, it finds the likely cause in the documents and routes it to the deal team, the company or an analyst.
Places each company and fund against the licensed peer benchmark and Quillfield's own history, and drafts the line the LP report will use once the fund dataset is approved.
Builds the industry submission file, the LP report section, the SFDR principal adverse impact table and LP questionnaire answers, citing every number to its company confirmation and source page.
Confirm, explain, prepare, approve. The agents propose; a named person decides.
Ask about any company, value or fund in the campaign, or tell the agents what to do. Every answer cites the document it came from, and nothing changes or goes out without a person.
Why is Fennick up 340%?
One site's bills are in kWh but were entered as MWh. The Laredo cross-dock's 9 bills total 102,350 kWh, and Fennick's tracker lists 102,350 MWh. Corrected, electricity is 22,480 MWh, Scope 2 is 8,340 t and intensity is 25.9, down 3.5%. The deal team agreed; the CFO has the correction.
Which companies will miss the March 26 deadline?
8 companies are at risk. None has opened its form after reminders in its own language and a deal team call request. The Follow-up agent proposes escalating each to its operating partner; that needs your approval.
How much of our data is reported, not estimated?
1,263 of 1,464 values (86%) were prefilled from the companies' own documents, each with its page; the rest come from the company in the form. Nothing is estimated silently: a value without a document stays gray until the company answers.
Where does 148,589 t come from?
QCP V's Scope 1 + 2 is the sum of its 24 companies' confirmed values: 62,595 t Scope 1 and 85,994 t Scope 2. It includes Fennick at 8,340 t Scope 2 after its CFO confirmed the Laredo correction.
13 screens from the working solution, on its sample data. Pick one to see it large.
How every value is checked before a fund's dataset can be approved — set by the administrator, approved by the ESG lead, every change versioned and on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Outlier: change on last year | over 50% | Escalated to the deal team with the likely cause |
| Outlier: distance from peers | over 3 times the peer median | Held back from benchmarks until explained |
| Impossible values | shares over 100%, more women than seats, negative energy | Flagged at once |
| Prefill only when confident | 90% or more | Below it the company answers |
| Carbon cross-check | company figure and recomputed figure differ by over 10% | Flagged to an analyst |
| Rule R-07: auto-close | agents' correction + deal team explanation + company confirmation | Closes the outlier; logged |
| Deal team explains an outlier | within 5 business days | Then the ESG lead is told; an analyst has 2 business days to accept or correct |
| Retention | 7 years after the fund year | approved datasets and attestations kept for the life of the fund plus 7 years; SEC books-and-records rules require at least 5 |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (BCG with the ESG Data Convergence Initiative, 2025 — more than 9,000 portfolio companies reporting; a 2023 survey of about 100 GPs and portfolio companies by an ESG software provider — up to 12 weeks collecting ESG data; SEC enforcement orders, 2022–2024 — $4 million, $19 million and $17.5 million over ESG claims that did not match practice). Regulatory context: EU SFDR (2019/2088) and Delegated Regulation (EU) 2022/1288 on principal adverse impacts; SEC Rule 204-2 on books and records. Quillfield Capital and the people, companies, funds and documents named on this page are fictional — characters and sample data in the working solution.
Portfolio Company ESG Data Collection is ESG data management software for a GP's annual portfolio company ESG data collection: it collects, checks and benchmarks the ESG metrics private portfolio companies report each year, such as the EDCI metric set. Seven AI agents prefill values from documents the company already shared, cited to the page, follow up in the company's language and find the likely cause of every outlier. Each company confirms its own values, an analyst prepares each fund's dataset and the ESG lead approves it before any LP sees it.
Usually with a blank survey or an ESG data collection template in Excel. With this solution, ESG data collection in private markets starts differently: each company gets a confirm-only form in its own language, prefilled from its utility bills, audited accounts, HR and safety reports and board minutes, each with its source page. In the working solution, 86% of 1,464 values arrive prefilled. The company confirms each value, corrects it with a reason and, if it has one, a supporting document, or marks a metric not tracked, then signs an attestation.
Yes. It collects the common private-markets metric set — Scope 1 and 2 emissions, renewable electricity, decarbonization plans, board and C-suite diversity, injuries and days lost, net new hires, turnover and the employee survey, plus a cybersecurity incidents metric — and the Reporting agent builds each fund's industry submission file from its approved dataset only. An analyst generates the file and the ESG lead submits it.
The Outlier agent compares every value with last year, sector peers and the company's own other values, finds the likely cause in the documents — units, boundary change, definition, coverage, method, timing, a one-off — and routes it to whoever can resolve it. The Carbon agent redoes the Scope 2 emissions calculation from activity data with a versioned emission factor set, and checks Scope 1 the same way. In the demo, a site's kWh typed as MWh would have overstated one fund's Scope 2 by 37,934 t.
Yes, from approved data only. The Reporting agent drafts the table of SFDR principal adverse impacts — the PAI indicators — for an Article 8 fund, the annual LP ESG report section and LP questionnaire answers, with every number traced to the company's confirmation and source page. Investor relations edits and marks each draft ready, and compliance releases it, so ESG data collection and reporting run on the same approved numbers.
No. The agents never confirm, change or approve a value, and write only to contacts on file, inside local working hours. Escalation to an operating partner needs the ESG lead's approval, and an outlier closes on its own only when the agents propose, the deal team explains and the company confirms.
Yes. The company confirms, the deal team explains, an ESG analyst prepares each fund's dataset and the ESG lead approves it — never the person who prepared it — with a reason code. Every step, by people and agents, is on a hash-chained audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your metric sets per fund, last year's approved dataset, your company list and the systems that hold company documents — and it goes live once every quality gate has passed. We will walk you through it on your own portfolio companies first.
Issuer questions answered in minutes, every claim pinned to its source page.
1-min demoESG Data ExtractionHoldings move from estimated to reported ESG data, each figure cited to its page.
1-min demoESG Risk MonitoringControversies in holdings scored within hours, and no fund's screen breached unseen.
1-min demoWe'll run Portfolio Company ESG Data Collection on a sample of your own companies, their documents and last year's approved dataset.