Helm
Platform Infrastructure

Morph

Bring whatever you have. The platform will make sense of it.

The Problem Morph Solves

Every customer has data trapped in documents and spreadsheets. Getting it onto the platform shouldn't require a data entry project.

Without Morph

  • 1. Receive fund admin statement (PDF/Excel)
  • 2. Open the platform's import template
  • 3. Manually re-key every field into the template
  • 4. Fix validation errors one by one
  • 5. Upload, pray, repeat next quarter

With Morph

  • 1. Data arrives — file drop, S3 event, or manual upload
  • 2. Morph reads the shape and proposes a mapping
  • 3. Review the result in a visual sandbox
  • 4. Approve and import — entities go live
  • 5. Next time? Same-shape file = automatic

Use Cases

Morph is horizontal — it serves any workflow where data is trapped in documents that should become live platform entities.

Prior Fundraise Migration

A stack of subscription documents from a previous raise. Morph materializes investors, contacts, commitments, accreditation status — everything the docs contain.

Investors Contacts Commitments

Cap Table Import

An Excel cap table from a fund admin or internal tracker. Investor-deal relationships with allocation amounts and ownership percentages.

Allocations Ownership % Relationships

Deal Setup from PPM

A private placement memorandum yields deal structure, terms, key dates, and fee schedules — ready to publish.

Deal Terms Fee schedule

Fund Admin Statements

Quarterly capital account summaries or NAV reports. Powers As-Reported mode in Portfolio for LP reporting without clean transaction data.

NAV Capital accounts Performance

CRM Migration

Exports from HubSpot, Salesforce, or a homegrown tracker. Contacts, organizations, and relationship metadata mapped onto the platform.

Contacts Organizations Relationships

Wire / Payment Matching

Wire confirmation PDFs or bank export CSVs become payment records matched to outstanding capital calls.

Payments Call matching

Data Room Ingestion

Connect to a third-party data room. Parse and classify documents, then route them logically into the platform — by deal, investor, document type.

Documents Classification Routing

Legacy System Migration

Bulk exports from a prior platform or internal system. Morph maps heterogeneous data shapes into platform entities to accelerate go-live.

Entities Relationships History

Two Modes

Same engine, different operating contexts. The mode determines how data arrives and how much human involvement the workflow needs.

InteractiveOne-offs and small batches

A user brings data to the platform directly. Drag-and-drop, file picker, or paste. Conversational — the agent walks through classification, mapping, and review in real time.

  • User-initiated — someone decides to bring data in now
  • Conversational mapping — the agent asks clarifying questions
  • Inline review — approve or correct results in the same session
  • Immediate — results land in minutes, not hours
Example A GP wants to set up five new investor records. They drag a small stack of investor profiles and subscription documents into the Investor Index and walk through the create flow.

PipelineContinuous and bulk ingestion

Data arrives programmatically — a fund admin drops files on an SFTP site, a client's system pushes to an S3 bucket, a scheduled job polls a remote data room. Batching, queuing, and monitoring become critical.

  • System-initiated — triggered by file arrival or schedule
  • Batched and queued — processes files in order, handles volume
  • Monitored — dashboards, alerts, SLA escalation for stalled files
  • Human-in-the-loop at validation — not at every step
Example A fund admin drops quarterly capital account statements for 200 investors onto a secure repository. Morph auto-detects the new files, applies a saved mapping profile, and stages results in a validation sandbox for review.

Where Morph Surfaces

Morph is an engine, not a screen. It can be invoked from the central agent or embedded directly into platform workflows — and either direction works.

Central Agent

General-purpose entry point

The Helm / Delio AI chat interface. Good for ad-hoc requests — "help me make sense of this file" — and as a router that identifies what the user is trying to do and delegates to the right specialized workflow when the task gets specific.

"I have a cap table to import" "Create investors from these sub docs" "What's in this PDF?"
See prototype: Delio Agent Chat
delegates / invokes

Specialized / Localized Workflows

Purpose-built UX

Morph capabilities embedded directly in platform screens. Drag-and-drop targets, structured review tables, provenance audit trails, approval flows — all tailored to the specific workflow context. The user may never see a chat interface.

Portfolio — drop statements on a fund Investor Index — bulk create from docs Data Room — classify and route imports Deal Setup — seed from PPM
See prototype: Portfolio LP Reporting
powers

Morph Engine

Core pipeline

The ingestion, classification, extraction, transformation, validation, and import pipeline. Runs the same whether invoked by the agent, a localized workflow, or an automated trigger. Stateless with respect to how it was called.

Ingest Classify Extract Transform Validate Import
Both directions work. Start in the central agent and it delegates to a specialized workflow when the task gets specific. Or start natively in a workflow — Portfolio, Investor Index, Deal Setup — and it invokes Morph at the juncture where data needs transforming. The engine doesn't care who called it.

How It Works

Six stages. Click any stage to learn more.

Step 1
Ingest
Data arrives
Step 2
Classify
What is this?
Step 3
Extract
Pull structured data
Step 4
Transform
Map to entities
Step 5
Validate
Sandbox & approve
Step 6
Import
Entities go live

What Morph Is — and Isn't

Morph is

  • A document-to-entity pipeline — files in, platform records out
  • Source-agnostic — reads the data shape, doesn't need pre-built connectors
  • Trigger-agnostic — works whether data arrives via upload, FTP, S3, API, or email
  • Human-in-the-loop where it counts — review and approve before anything lands
  • Self-learning — mapping profiles are saved and reused, eventually automatically
  • Deterministic where it counts — AI proposes rules, a pipeline executes them with no hallucination

Morph is not

  • A point-to-point integration (no "Citco connector" or "Apex connector")
  • A Salesforce/CRM sync — structured bidirectional integrations are a separate concern
  • A replacement for platform validation — the platform still decides what's valid
  • A reporting or analytics tool — Morph gets data in, not out
  • Limited to any single document type — any data the platform should know about is fair game

Key Properties

The design decisions that make Morph trustworthy at scale.

Deterministic

Same file + same rules = same output. Every time. The AI proposes mapping rules; a pipeline executes them row by row with zero model calls.

Reusable Profiles

First file from a new source costs one mapping session. Every subsequent file with the same shape reuses the profile — zero AI cost, one click.

Cost Scales with Schema

AI cost is bounded by column headers + sample rows. A 50,000-row file costs roughly the same to map as a 500-row file.

Resumable

Every stage transition is checkpointed. Browser crash, server restart — pick up exactly where you left off.

Self-Correcting

Validation errors route back to rule extraction. The AI amends the rules, the pipeline re-runs, and validation re-runs — no re-upload needed.

Fully Auditable

Every transformed value carries provenance back to the source row and the rule that produced it. Nothing is a black box.