Everything your firm knows, in one system it can answer from.
Funds, deals, portfolio, investors, private credit and compliance in a single governed model — so performance has one answer, and you can always show where it came from.
Most firms do not have a system of record. They have several.
The firm's truth is spread across systems that were never meant to agree with one another.
Fund data sits with the administrator. The pipeline sits in a CRM designed for selling software. Portfolio performance arrives as workbooks attached to email. Commitments, capital calls and covenants live in spreadsheets that one person maintains and everyone depends on.
Each of those is a reasonable choice on its own. Together they are not a system of record, because nothing reconciles them but people.
The cost is not the effort. It is that nobody can be certain.
A figure restated in one place does not propagate to the other five, so the same metric can appear differently in an LP report and a board pack — and it is usually someone external who notices first. Quarter-end turns into reconciliation rather than analysis. Asked who changed a valuation and when, the honest answer is that it is somewhere in an inbox.
This is also the reason AI has not helped. A model reading six sources that disagree does not hedge; it produces a confident answer that happens to be wrong.
Because the shape of the work keeps changing, and most software cannot follow it.
A firm that adds a credit strategy, a continuation vehicle or a new reporting obligation needs its model to change with it. Software that fixes the meaning of fund, deal and investor in code cannot follow, so the difference is made up in spreadsheets — and the cycle begins again.
Which points at what a durable answer has to look like: the model itself has to be data.
The model is data, so it changes when the firm does.
Adding a credit strategy or a continuation vehicle should not require anyone to ship software. Every object type, field, pipeline stage, KPI strip, view and profile section is read from one schema at runtime — there is no fund-shaped code path and no credit-shaped one.
Add a field and it appears in the table, the entry form, the filter menu, the export and the API at once. That completeness is what stops the parallel spreadsheet from reappearing a month later.
- Field types19
- Formula, rollup & derived fieldsComputed
- Conditional visibility & editabilityPer field
- Access rulesPer field, per role
One set of records, however each desk needs to see it.
When a tool offers a single shape, everyone it does not suit exports to a spreadsheet — and from that moment the firm has two versions of the same records, drifting apart quietly.
Table, board, cards, calendar, map and pipeline are layouts over one dataset, not copies of it. Filtering, sorting and grouping run on the server, so a view over a hundred thousand records opens like a view over ten.
- Six view layoutsOne dataset
- Filter, sort & groupServer-side
- Saved views → dashboardsComposable
- Export any viewCSV · XLSX · JSON
The record keeps the history, so nobody has to remember it.
Context usually lives outside the system: the diligence thread in one inbox, the board pack on a drive, the reason a valuation moved in nobody's notes at all. When the person moves on, the context goes with them.
Fields, relations in both directions, comments, files, tasks, activities, emails and meetings sit on the record itself. The relationship map walks outward and applies permissions at every hop — so it tells you three records exist that you may not see, rather than quietly omitting them.
- Profile tabs & sectionsConfigurable
- Forward & inverse relationsBoth ways
- Comments, tasks, notes, filesOn the record
- Relationship mapACL'd per hop
Meridian Health
- Sector
- Healthcare services
- Enterprise value
- $182.0M
- Ownership
- 64.0%
- MOIC
- 2.41x
- Deal team
- T. Okafor, P. Raman
- Covenant
- 2.4x / 3.0x
- Next IC
- 12 May 2026
- Entry date
- 04 Aug 2021
- Region
- North America
- Board seats
- 2 of 5
- Revenue FY26
- $420.3M
- Last mark
- Q1 2026
One number, and one way it was derived.
Portfolio figures are usually re-entered on the way up — company workbook to fund model to LP report. Every hop is an opportunity for two versions of the same number to exist, and for the disagreement to surface in front of an investor.
Metrics are entered, calculated, or consolidated up the relation path you already modelled, so a fund total is the sum of its companies by construction rather than by a monthly copy-paste. Actual, Budget and Forecast sit side by side, reporting currency converts on daily rates, and the grid round-trips through Excel with a revision history.
- Actual · Budget · ForecastScenarios
- ConsolidationUp relation paths
- Multi-currencyDaily FX
- Excel round-tripWith revisions
AI is only worth as much as the record underneath it.
Pointed at sources that disagree, a model does not hedge — it answers confidently and wrongly, which is worse than not answering at all. A governed record is the precondition for any of this being useful, not a refinement of it.
Ask a question and the answer comes back composed from the app's own components, under exactly your permissions. Agents propose rather than write: anything a model wants to create or change queues for a person to approve, and the whole run is kept as history you can reopen.
- Composed answersReal components
- Proposals, not writesApproval-gated
- Generates objects, views, dashboards, workflowsBuild mode
- Every run keptAudited
The process runs whether or not anyone remembers it.
Stage gates, sign-offs and quarter-end sequences tend to live in habit and a checklist. They hold for as long as the people who wrote them are in the room, and they are the first thing to slip in a busy quarter.
Workflows run as a graph and checkpoint after every node, so a restart resumes rather than starts again. Any step can be an approval gate: the change is computed, held, and applied only once the named approvers have signed off.
- Durable graph executionCheckpointed
- Approval gatesMaker-checker
- Scheduled runsDST-correct
- VersionedWith rollback
When someone asks who changed it, there is an answer.
Access granted screen by screen leaks: a figure an analyst cannot see on the record still appears in an export, a chart, or an AI answer. And a valuation that changed with no attribution is a problem you discover at the least convenient moment.
Every mutation is audited with its before and after, attributable to a person and restorable. Access is enforced per field and per role — in exports and AI answers too, not only on screen. Compliance domains are yours to define, and privacy covers subject access, erasure, retention and legal hold.
- Field-level accessPer role
- Full audit trailWith restore
- Compliance domainsUser-defined
- DSAR, erasure, retentionGDPR · CCPA
A record that does not fill itself goes stale.
Any system of record that depends on somebody remembering to update it decays, and once it is a little out of date everyone quietly goes back to the source they trust instead.
Point it at a REST endpoint, a CSV, an XML feed, or a page that needs a real browser to render. Nothing lands silently: rows that fail validation wait in a review queue for a person to accept, fix or reject. Connect Gmail or Outlook and threads and meetings attach themselves to the right deal.
- API, CSV, XML & browser sourcesOne editor
- Review queueBefore commit
- Scheduled syncsEncrypted secrets
- Gmail & OutlookAuto-linked
Everything else, once.
The rest of the product, listed rather than sold — ten groups here, the full 113 a click away. If something here matters to your firm, it is worth half an hour to see it working on your own data.
Built for deal teams, investor relations, finance and compliance — the same records, read four different ways. What each desk gets →
One place the numbers agree.
A record that can account for itself, a quarter-end that is analysis rather than reconciliation, and AI worth trusting because of what sits underneath it.
Bring a fund, a deal and a reporting pack you actually use. The question worth half an hour is whether the model fits your firm — not whether the slides look good.