PLATFORM

ADI is the control plane for private markets data.

ADI connects your private markets data into one governed set of records, so the workflows built on top of it stop depending on manual entry. Ingest, integrate and report. Integrate, ingest, and report. One platform for all of it.

01 / CAPABILITY MAP

Ten capabilities, one platform.

Most vendors sell these as separate products. Here they share one data model, so a correction made in extraction shows up in reporting without a reload.

01

Ingestion & connectivity

Email, cloud drives, SFTP, document portals and data-warehouse connectors, all flowing into a versioned store.

EMAIL · PORTALS · SFTP · WAREHOUSES
02

Document intelligence

OCR, a classification taxonomy, schema-driven extraction into runtime-typed models, a self-service extraction wizard and zero-shot named-entity recognition with a feedback loop.

OCR · SCHEMA-DRIVEN · EXTRACTION WIZARD
03

Document repository

A versioned repository for every document: classified, searchable and archived with a complete audit trail. Every document version is tracked, time-stamped, and archived for full traceability across reporting periods.

VERSIONED · TIME-STAMPED · AUDIT TRAIL
04

Entity resolution & master data

Ensemble matching with adaptive confidence, graph clustering, and golden records tied to your source of truth, enriched from commercial reference universes like S&P Capital IQ, Preqin, Cherre and ICE, with public identifiers (LEI, FIGI) layered on where they exist.

GOLDEN RECORDS · SOURCE OF TRUTH
05

AI agents & analytics

Document extraction, data quality, reconciliation and reporting agents on one governed layer, plain-language questions compiled to SQL, dashboards on demand, and an MCP server for external AI clients.

EXTRACTION · QUALITY · RECON · REPORTING
06

Search & RAG

Per-tenant vector collections, dual chunking strategies and a semantic catalog, five collections per vertical across assets, documents, summaries, entities and parties.

PER-TENANT · SEMANTIC CATALOG
07

Workflow & human review

Maker-checker task queues with field-level edits, approve and reject, and a full audit trail, on an orchestrator with a scheduler, event router and autoscaler.

MAKER-CHECKER · AUDIT · ORCHESTRATION
08

Reporting & BI

NL to SQL, drag-drop dashboards, charting and PDF export, holdings and ownership graphs, valuation and a query lab, plus RAG-driven branded Word and PDF report synthesis.

NL→SQL · DASHBOARDS · HOLDINGS · BRANDED PDF
09

Data distribution

A distribution layer with governed reader accounts, table sharing and reporting-client provisioning, white-label data delivery on top of verified data.

GOVERNED SHARES · READER ACCOUNTS
10

Security & platform

Per-tenant isolation, role-based access by organization and department, secret management, configuration services and platform-wide observability.

TENANT ISOLATION · RBAC · OTEL
02 / PIPELINE

One pipeline, every alts document type.

[ IN ]

Ingest from anywhere

Email, GP portals, SFTP, cloud drives and custodian feeds, with background loops that pull the moment a document posts.

EMAIL · PORTAL · SFTP · FEED
[ EX ]

Extract into typed fields

Schema-driven extraction maps each statement into defined, validated data, not a wall of text.

SCHEMA · VALIDATED
[ ID ]

Master the entity

Any entity: clients, portfolios, assets and people, resolved to one master record and tied to your source of truth.

CLIENT · PORTFOLIO · ASSET
[ QA ]

Review what matters

A maker-checker queue surfaces only the fields the model is unsure of. Humans spend minutes, not hours.

MAKER · CHECKER
[ WB ]

Write back, audited

Validated transactions posted into the systems you already run. Every field traceable to source.

WRITE-BACK · AUDIT
[ DX ]

Report and distribute

Branded client reporting and governed data delivery, built on data you already trust.

REPORTING · DELIVERY
03 / ENTITY MASTERING

Entity resolution, and why it sits underneath the rest.

A NAV figure is unusable until you know which fund, vehicle and investor it belongs to. Aggregation tools hand you the figure and leave the attribution to you.

Three systems write the same fund three different ways. ADI resolves all of them to one record, tied to your source of truth.
A short name on the custodian feed. The full legal entity on the capital account statement. An abbreviation in the PortCo report. Ensemble matching with adaptive confidence bands, graph clustering and golden records resolve all three to the same master record, anchored to the system you treat as the source of truth, enriched from commercial reference universes like S&P Capital IQ, Preqin, Cherre and ICE, with public identifiers (LEI, FIGI) layered on where they exist. Mismatches route to a human; everything else is automatic.
entity_resolution · fund names redacted
alias_1Sample Fund IX
alias_2Sample Capital Fund IX, L.P.
alias_3SCF Fund 9
masterGR-0417 · your source of truth
identifiersLEI · FIGI resolved
confidence0.98 · auto
04 / DATA QUALITY & SYNC

Greater than 98% out of the box. 100% after review.

Four stages sit between an extracted field and a posted one: validation, cross-document checks, human review and writeback QA.

01
Greater than 98% out of the box
Schema-driven extraction gets 98% of fields right before anyone touches them.
02
Validations, reconciliations, ML
Accounting-logic checks, cross-document reconciliation and machine learning close the gap.
03
Human review where it counts
Low-confidence fields route to a maker-checker queue. Minutes of review, not hours of rekeying.
04
100%, written back
Verified data posted to your system of record, audit-stamped to source.
Verified once, synced everywhere.
ADI moves data between systems, from documents to the GL, from the GL to portfolio monitoring, and keeps your accounts, book of record and portfolio system continuously reconciled. Feeds are matched, breaks are explained with a root cause, and adjustments post with the same field-level audit trail as everything else. Nothing drifts quietly.
05 / THE AGENT LAYER

Agents for the work your team does by hand.

Every agent runs on the same rails: typed schemas, identifier grounding, a human review gate and writeback QA. Nothing reaches your system of record unreviewed.

[ 01 ]

Document Extraction Agents

Read capital calls, distributions, statements and K-1s into typed fields, with every value traceable to its source page.

TYPED SCHEMAS · SOURCE-LINKED
[ 02 ]

Data Quality Agents

Run accounting-logic validations and cross-document checks, then route low-confidence fields to human review.

VALIDATIONS · REVIEW GATE
[ 03 ]

Reconciliation Agents

Compare custodial and brokerage feeds against your book of record, surface breaks and run root-cause analysis.

FEEDS VS BOOK · ROOT CAUSE
[ 04 ]

Reporting Agents

Turn plain-language asks into SQL, dashboards and branded client reports, built on verified data.

NL→SQL · DASHBOARDS · BRANDED PDF
EVERY AGENT ACTION LANDS IN THE SAME MAKER-CHECKER QUEUE A HUMAN USES · NOTHING REACHES YOUR SYSTEM OF RECORD UNREVIEWED
06 / INTELLIGENCE

Asking questions of the data once it is verified.

Once every field is extracted, mastered and checked, two things come for free: a portfolio you can query in plain language, and reporting that builds itself from numbers you already trust. Questions compile to SQL and run against governed tables: the model writes the query, the database does the math, so answers are deterministic, auditable and cheap to run at volume.

ask · portfolio intelligenceLIVE
What's our total unfunded commitment across private equity?
ADI · ANSWERED FROM VERIFIED DATA
Across 14 active PE funds, total unfunded commitment is $127.4M, concentrated in three 2024-vintage funds.
FUNDUNFUNDED
Growth Equity IX$41.2M
Buyout Fund VIII$33.8M
Technology Fund XV$28.1M
SOURCE · 14 CAPITAL ACCOUNT STATEMENTS · Q2 2026 · EACH FIELD TRACEABLE · FUND NAMES REDACTED
Ask about positions, exposure, commitments…ASK →
report · exposure by strategy · auto-generated

Portfolio exposure by strategy

Private equity38%
Private credit26%
Real estate19%
Venture11%
Hedge funds6%
BUILT FROM VERIFIED POSITIONS · NO MANUAL ROLL-UP · REFRESHES ON EVERY NEW STATEMENT · EXPORT TO BRANDED PDF OR GOVERNED SHARE
07 / MODEL AGNOSTIC

Model choice is yours. The grounding underneath is ours.

Models change every quarter, so the platform stays portable across them and keeps the grounding, schemas and review gates in its own layer.

The model reads. ADI is the verification layer that makes the answer safe to act on.
Bring your own model, or use ours. The frontier vendors all ship strong extraction now, so raw capability is not the differentiator. Grounding every answer in typed schemas, mastered entities and a human-checked audit trail is. ADI selects the right model for each job, with a bias to open-source models where they meet the bar: lower cost, and documents that never leave your environment.
model routing · your choice or ours
Open source / self-hostedPREFERRED
Anthropic ClaudeCONNECTED
OpenAI GPTCONNECTED
Google GeminiCONNECTED
08 / HEAD TO HEAD

Compared with aggregation platforms.
Where the two overlap, and where they do not.

Aggregation platforms are strong at the top of the stack: portals, categorization, extraction. The differences start below that, in accuracy, entity mastering, where processing runs and what it costs to operate.

Capability
Aggregation platforms
ADI
Portal and inbox collection
Yes
Yes
Categorization and organization
Yes
Yes
Field extraction
Yes
98% out of the box
GL integration
Yes
Yes
Data quality validations
Partial
Built in
Feed and statement reconciliation
Partial
Built in
Maker-checker review in your control
Vendor-side only
Built in
Processing in your environment
No
Your cloud or on-prem
Document management and catalog
Partial
Built in
Agentic document automation
No
Built in
Search across documents and data
No
Built in
Conversational AI
No
Built in
Report builder
No
Built in

A session on your own stack. Your documents, your systems, our engineers.

A working session on your real documents and systems, across the capabilities your team needs most.