Live on production. Portfolio transfers in dev (Jason): Transfer In/Out NAV logic, paid-in, capital gains, DPI/TVPI formulas all being updated. Contactable groups in dev (James). DocuSign SMS auth shipped — all Schwab AODS requirements covered.
Signed. Troy + Gareth Lewis on-site workshops June 9–12 for Phase 1 (Core OS). Phase 2 is portfolio data migration; Neil is prepping the requirements document. Phase 3 integrations are currently unscoped.
James's DEV-4561 adds: inline user attachment at entity creation (Barclays portal use case), and inline sub-group creation from an advisory group (Norcliffe advisor portal). These are tactical fixes on top of the existing system.
Horizon front-end decommission is blocked on Barclays migration — Ege hasn't responded and Sarah has chased twice. Once Barclays migrates, ~4 weeks to fully decom front end. Risk: Sarah is leaving in ~3 months and holds this relationship.
Option A: Delio-branded distribution. Option B: Customer Apple account. Push notification trust model is the core design tension — whether push notifications arrive from Helm's identity or the client's. Tom Davies is managing the WinterPrivateMarkets iOS/Android release separately, currently blocked on Quest setting up their Apple/Google developer accounts.
HCP-7 covers the missing ![] operator for PDF templating. Plan mode, preview mode, and auto-conditions are all in dev. The agreed direction is DocuSign-native e-signing — no conditional PDF mutation will be built. Backfill of existing subscription documents is in scope.
Portfolio, deals, investors, and users agents are already deployed. These are domain-specific agents that let a GP admin query and act on platform data through a conversational interface.
Documents agent and subscriptions agent. Once these ship, a usability overhaul of the full agent surface is planned — consistent UX, discoverability, agent handoff.
AI-surfaced signals about how the platform is being used — operator preferences, configuration drift, usage patterns. Surfaces recommendations to the GP admin rather than requiring them to dig for it.
MCP-enabled one-to-one action: one subscription doc → one investor record, or one LPA → one vehicle config. This is distinct from Morph — single-document actions, not bulk ingestion.
The RAG knowledge base is stale — limits agent accuracy. Fix underway. Direction from Jonathan: don't wait for perfect — feature-flag and ship when ready.
Follow-ons are hardcoded to the same portfolio asset across all tranches. Document tagging per tranche is also missing — investors can't distinguish which subscription document belongs to which round. Needed for Series A/B co-investment use cases.
Certain TM workflow steps should be able to float to any position in the sequence rather than being fixed to a stage. Dev team is aware; tickets parked until pulled into sprint.
New send engine through 3–4 review rounds, close to shipping. GC built an email template visualiser — all templates visible in one place. Two outstanding actions before go-live:
Three things remaining, all Barclays-related: old onboarding, old groups/teams, and auth endpoints. Blocked on Barclays migration — Ege hasn't responded. ~4 weeks to fully decom once Barclays migrates. Risk: Sarah leaves in ~3 months and holds this relationship.
Callum's 4-type user model: end investor, managed investor, advisor, entity. Non-fungible audit logs are a primary motivation (current logs are technically editable — a compliance risk). Target: Horizon decommissioned as a running service by end of 2026.
Morph v1 requires someone to manually download a data file from the fund admin and upload it through the portal — every time data needs refreshing. This phase removes that step. Fund administrators (like State Street, Apex, or a client's internal data team) connect directly via SFTP or a scheduled API, and Morph picks up new files automatically on the agreed cadence.
Same data quality rules as Morph v1 apply — structured transactional data only. Target clients are those with regular monthly or quarterly data refreshes from institutional fund administrators, where the manual handoff is the main friction point.
Currently any error triggers a full reverse — safe but limits adoption to teams with deep data expertise. The next UX layer adds a human review step: failed or ambiguous rows can be inspected, corrected, and approved before committing. This is the prerequisite for scaling Morph beyond pilot clients to teams that don't have a dedicated data expert in the loop.
The problem: Most LP portal clients can only provide point-in-time snapshots — paid-in, NAV, distributions, cap gains per investor per quarter. Today the team manually converts these to transaction format, a slow, fragile process requiring deep expertise. Morph v1 does not solve this.
The data contract: GPs give us a tabular version of what goes on a quarterly statement, or the PDFs and we extract it. Result: a time-series of per-investor snapshot data that enables LP experience — NAV build over time, commitment-to-call change, TVPI trend — without requiring full transaction history.
Technical approach: Spiking the import layer first — can snapshots translate to single-transaction-per-period entries in the existing structure cleanly? If yes, proceed. If not, build a separate flat table. Phase 1 will likely mean a client is either transactional or snapshot mode — not mixed.
More entity types and actions accessible via MCP — investor management, TM step progression, vehicle configuration. Longer-term north star: LPA ingestion — Morph reads the LPA to auto-extract capital call rules, transfer mechanics, and fund specs, replacing manual vehicle config toggles.
Rather than fix per-client in isolation, we're doing a holistic discovery to establish a canonical model for Groups, entities, and permissions. Core problems being solved:
James's group visualiser (in development) — a tree view of all groups, users, roles, and access rights — is the management foundation everything else builds on.
WestCap hard requirements for migration (both confirmed in flight):
Two distinct use cases that need to be clearly separated:
Current state: documents are in a legacy attachments service being migrated to core. Tom Davies is the likely owner after current queue items complete.
Two integrated concepts confirmed in the DSS bi-weekly:
Individual investor KYC stays embedded in TM — no change. GP-facing language throughout — no MLRO, no AML jargon visible to clients. FPC (Future Planet Capital) contract is contingent on this feature shipping.
Delio's deal configuration wizard is a long create-then-edit flow. Verivend's step-by-step guided setup — NDA toggle, transaction type picker, embedded configuration — compares favourably. WestCap's demo this week confirmed this gap: the configuration steps were more complex than expected for a first-time user. Simplifying this is a key retention and activation unlock for new GPs.
Pre-TM flow: invite institutional LPs to see a deal (no full onboarding/suitability classification required for institutional counterparties), NDA gate, data room access, indication of interest. Simpler deal sharing: reduce friction between initial interest and formal transaction. Prototype work underway in Passage.
Today, configuring a new client platform takes hours: manual flag-enabling, role creation, form copying, even getting defaults right. As the team scales and CS ownership evolves, this becomes a bottleneck.
Agreed as a committed workstream. Callum creating the epic in Jira.
The stated long-term direction: transfer behaviour and fund rules should be configurable per vehicle without manual toggles, with the LPA as the source of truth. Morph extracts those rules at ingestion time. Short-term: manual vehicle config. Long-term: Morph replaces it. Also covers recurring snapshot ingestion with validation rules extracted from the LPA — the bridge between "as reported" mode and full automation.
Individual AI features are building — Morph, agent actions, as-reported extraction, platform insights. But there's no coherent story for how a GP or LP discovers, trusts, and activates them. This is the cross-platform design question: how do AI features feel native to Helm rather than bolted on? Discovery workstream, separate from any individual feature.