TeachU.aiAI INTELLIGENCECompare tools ↗

How Multi-Family Offices Should Evaluate AI Decision Systems

Every vendor fact on this page is provider-reported (taken from the vendor's own published materials) and has not been tested by TeachU.

Who this is for

Principals, COOs, CIOs, heads of operations and investment committees at multi-family offices (MFOs), single-family offices and wealth firms that serve several families, who are deciding whether and where AI should take part in investment, reporting, tax, estate or household-level decisions.

This is general buyer education. It is not investment, legal or tax advice.

The decision you're actually making

"Should we buy AI?" is the wrong question. The useful one is: which decisions, on which data, may a system help with, and under what controls?

Products marketed to family offices do very different jobs. In TeachU's evidence they fall into at least four groups:

  1. Systems of record with an AI layer. Accounting, aggregation and reporting platforms that add an assistant or document AI (for example Asset Vantage, Eton Solutions AtlasFive, FundCount, Masttro).
  2. Document and data capture. Tools that pull alternatives documents and statements into structured data (for example Canoe Intelligence).
  3. Advisor workflow and CRM assistants. Meeting notes, follow-ups and CRM updates (for example Wealthbox AI).
  4. Decision-support systems. Products that aim to reason across tax, estate, legal and investment views for one household. TeachU's published evidence for this group is INSUFFICIENT_EVIDENCE.

A data platform's AI can be strong at answering "what do we own?" and still say nothing useful about "what should this family do?". Decide which job you're buying for before you compare vendors.

Evaluation framework

Assess every candidate against the same eight written areas and record the evidence for each. Where a vendor's public material is silent, write INSUFFICIENT_EVIDENCE and ask, rather than assuming.

1. Data governance and permissions

  • Can access be scoped by family, entity and person? Vendors describe this in several ways: entity-scoped access for different family stakeholders [V8-MAS-D3].
  • Is your data used to train shared models? One vendor states customer documents are not used to train public models [CN-C06]. For most products in our evidence, this is INSUFFICIENT_EVIDENCE.
  • Where is data hosted, and who holds the keys? Vendors claim client-controlled encryption keys [MO-C04], Bring Your Own Key encryption [ET-C04] and Swiss hosting [AO-C05]. These are vendor claims that TeachU hasn't reviewed. Ask to see the supporting documents.

2. Auditability and citations

  • Does every AI conclusion link to its source document or data point? Does the system log inputs, outputs and approvals?
  • Examples of what vendors document: immutable audit logs and input guardrails in a beta AI feature [TM-C01].
  • TeachU's wealth-decisions rubric treats "conclusions carry sources or provenance" and "decisions and inputs are logged" as separate criteria.

3. Model risk

  • Who calculates the numbers, the model or a deterministic engine? TeachU's rubrics make this a must-have: scenario and tax math computed deterministically and documented, and returns computed by a documented deterministic engine, not the model.
  • Which model and provider sit underneath, and how are version changes communicated? For many products this is not disclosed: model provider is "not stated" or unknown for Wealthbox AI [WB-C08], FundCount [FC-C08], Black Diamond [BD-C07], and Envestnet Tamarac [TM-C05]. INSUFFICIENT_EVIDENCE across the field.
  • Does the system surface uncertainty and disagreement instead of a single confident answer? This is a rubric criterion.
  • Ask the vendor what model-risk documentation it can provide for its AI features, and how it documents model changes.

4. Integration with custodial, portfolio and CRM systems

  • Map the systems the AI must read from and write to: custodians, portfolio accounting, general ledger, CRM, planning, document storage.
  • Documented examples include aggregation from banks and custodians [AZ-C01]; a custodial data API covering 500+ custodians [LS-C01]; API delivery to downstream systems [V8-CAN-D3]; and MCP connectors for customer-chosen AI tools [BD-C03] [AO-C02] [AL-C04].
  • For connectors, ask whether access is read-only or can write back. For one MCP product this is not stated [AO-C08].

5. Human review

  • Which actions can the system take on its own, and which are held for a person? This is a must-have in TeachU's rubric.
  • Vendor positions vary widely. Examples: drafted client communications are reviewed before sending [WB-C02]. One vendor describes customer agents automating capital-call workflows without a human in the loop [AL-C05]. Ask every vendor which actions can run without a person approving them, and decide in advance which you'd accept.

6. Security and compliance attestations

  • Ask for the report or certificate itself, not a logo. Every attestation below is vendor-claimed; TeachU has not reviewed the report or certificate:
  • SOC 2 Type II: Canoe Intelligence [CN-C05], FundCount [FC-C06], Wealthbox [WB-C04], Aleta [AL-C06], Holistiplan [HP-C06].
  • SOC 2 Type 2 compliance: Asset Vantage [AS-C05].
  • ISO/IEC 42001 (AI management): Eton Solutions [ET-C04].
  • ISO 27001: Asora [AZ-C06], Landytech [LS-C05].
  • Not stated on the pages fetched: Masttro [MO-C08], Altoo [AO-C06]. INSUFFICIENT_EVIDENCE.
  • Ask which systems, controls and time period the report covers, and whether the AI features are included.

7. Vendor viability

  • Ask about ownership, funding runway, customer concentration, release cadence, and data export on exit.
  • Vendor-reported scale figures (assets on platform, client counts) appear on several sites. TeachU doesn't use them as quality signals, and company size and popularity are never inputs to TeachU assessments.
  • Watch for renames and product changes. One AI layer was renamed (EtonGPT to EtonAI) [ET-C03]; one vendor lists general-ledger accounting as "coming soon" [LS-C04]; AI agents may be early access depending on plan [V8-WB-L1]; one approved-actions feature is "available soon" [BD-C02].

8. Total cost

  • Public pricing is rare. Examples where vendors publish figures: FundCount single-family-office plans from $35,899/year plus digital transformation and hosting fees, as annual starting points [FC-C01] [FC-C02]; Asora: Price not re-verified (source unavailable 2026-11-03) [AZ-C02] [AZ-C03] [AZ-C04] [AZ-C05]; Holistiplan Basic $749/yr and Premium $1,499/yr for 30 households [HP-C04]. One vendor states fixed, non-AUM pricing without a figure [MO-C05].
  • Not stated publicly: Asset Vantage [AS-C08], Altoo [AO-C07], Eton [ET-C08], Landytech [LS-C06], Masttro [MO-C06]. INSUFFICIENT_EVIDENCE.
  • Prices were last checked on 2026-10-27 and change often. Confirm current pricing with each vendor.
  • Budget beyond licence: implementation, data migration, reconciliation of history, connector fees, internal review time and exit costs.

Buyer checklist

  • [ ] We've written down the specific decisions or workflows the system will touch, and which it must not.
  • [ ] Permissions can be scoped by family, entity and person, and we've tested that a user can't see another family's data.
  • [ ] Every AI conclusion cites its source, and we can export the audit log.
  • [ ] Numbers (performance, tax, scenarios) come from a documented deterministic engine, not the language model.
  • [ ] Model provider and version are disclosed, with notice of changes.
  • [ ] Read and write scopes for every connector are documented.
  • [ ] Nothing reaches a client, a custodian or a system of record without a named human approver.
  • [ ] We've read the actual SOC 2 / ISO report and confirmed the AI features are in scope.
  • [ ] We've asked the vendor how the contract handles data-training restrictions, data residency, breach notice and exit or export.
  • [ ] We've run a pilot on our own reconciled data with known answers before relying on outputs.
  • [ ] Total cost includes implementation, connectors and review time.
  • [ ] We've asked our own advisers which internal reviews this purchase needs.

Red flags

  • Outcome statistics with no method or artifact. TeachU excludes these vendor claims from its assessments, such as unqualified time-saved, accuracy or growth percentages [CN-C07] [CN-C08] [AL-C09] [HP-C07].
  • Superlatives such as "only fully integrated" [ET-C07].
  • AI that can act on client accounts or systems of record with no approval step.
  • Attestations shown as badges, with no report available under NDA.
  • No answer to "which model, which version, where does our data go?"
  • "Agent-facing" pages that make pricing or competitor claims that differ from human-facing pages. TeachU's researchers found and ignored such pages on two vendor sites [MO-INJ] [AL-C10].

Questions to ask vendors

  1. Which decisions is the AI allowed to make, recommend or only summarize? Show us the configuration.
  2. Which calculations are deterministic, and where is that engine documented?
  3. Which model provider and model versions do you use today? How do you notify customers of changes?
  4. Is customer data used to train any model, yours or a third party's?
  5. Show us an AI answer with its citations, and the audit-log entry it creates.
  6. What are the read and write scopes of each connector, including any MCP server?
  7. Can we see your current SOC 2 Type II report or ISO certificate, and are the AI features in scope?
  8. How do you handle conflicting information across custodians, entities or advisors?
  9. What does a pilot on our own data look like, and what error rate have you measured with what method?
  10. What is the full first-year and run-rate cost, including implementation, connectors and hosting?
  11. How do we export all of our data and logs if we leave?

If you want an outside view on these questions for your own firm, our AI Readiness & Architecture Diagnostic is a fixed-fee review of where AI fits in your operations.

How TeachU assesses these products

TeachU keeps three measures separate and never blends them into one score: - Capability coverage: Documented capabilities, not tested. What vendor documentation says the product does, with links. - Architecture fit: Design assessment, not a test result. A published, field-specific rubric applied to the evidence. An AI system proposes criterion results with citations, and versioned rubric code computes the level. No person, including TeachU's founder, adjusts it. Fit is reported in words only: Strong fit, Moderate fit, Limited fit or Not enough evidence. - Measured performance: Not tested yet. No products in this field have been tested. Results will come only from pre-registered benchmarks.

Two rubrics apply here: wealth decisions (deterministic calculation, multi-entity intake, uncertainty handling, provenance, human approval boundaries, audit trail, confidentiality) and wealth portfolio reporting (deterministic performance calculation, multi-custodian reconciliation, corporate actions/fees/FX, AI audit trail, human validation, permissions). Both are early versions that may change, and no architecture-fit result has been published yet.

How we handle commercial relationships with products we cover is set out in our conflicts of interest policy.

Products in TeachU's evidence

Grouped by fit level and listed alphabetically. Because no fit assessment is published yet, every product is NOT_ASSESSED for architecture fit. Measured performance: Not tested yet for all. Capability notes are Documented capabilities, not tested.

Strong fit

None published.

Moderate fit

None published.

Limited fit

None published.

Not enough evidence

None published.

NOT_ASSESSED (architecture fit not yet published)

  • Aleta: consolidated multi-entity wealth view; AI reads private-markets documents; open MCP layer [AL-C01] [AL-C03] [AL-C04].
  • Altoo: consolidated total wealth view; MCP connection to an AI platform the user chooses [AO-C01] [AO-C02].
  • Asora: aggregation from banks and custodians; generative-AI features not stated [AZ-C01] [AZ-C08].
  • Asset Vantage: family-office general ledger and reporting; generative-AI features not stated [AS-C01] [AS-C07].
  • Black Diamond (SS&C): in-platform assistant; governed MCP server [BD-C02] [BD-C03].
  • Canoe Intelligence: alternatives document automation with rules that flag items for expert review [CN-C01] [CN-C02].
  • Eton Solutions AtlasFive: consolidated accounting, investment reporting and documents; EtonAI layer [ET-C01] [ET-C02].
  • Envestnet Tamarac: reporting with beta AI features [TM-C01].
  • FundCount: accounting-backed reporting; AI document intelligence with exception review [FC-C03] [FC-C05].
  • Holistiplan: tax-return OCR and planning scenarios [HP-C01] [HP-C03].
  • Landytech Sesame: custodial data API; aLi AI querying [LS-C01] [LS-C02].
  • Masttro: UHNW aggregation and reporting; AI for alternatives documents [MO-C01] [MO-C02].
  • Wealthbox AI: CRM assistant; drafts reviewed before sending [WB-C01] [WB-C02].

This list isn't a ranking and implies no order. Inclusion means only that TeachU holds a dossier.

Corrections

If you think something here is wrong, tell us through our corrections page.

Claim table

All sources were retrieved by TeachU between 2026-10-21 and 2026-10-24; price claims were re-checked on 2026-10-27. Label key: PR = provider-reported. No claim on this page is independently reported or verified.

Claim ID Claim Source URL Label
AL-C01 Consolidated, reconciled multi-entity wealth view https://aleta.io/ PR
AL-C03 AI reads private-markets documents https://aleta.io/ai-ready-wealth-intelligence-platform PR
AL-C04 Open MCP layer for customer agents https://aleta.io/ai-ready-wealth-intelligence-platform PR
AL-C05 Customer agents automate capital-call workflows without a human in the loop https://aleta.io/ai-ready-wealth-intelligence-platform PR (sensitive)
AL-C06 SOC 2 Type II on Azure (report not reviewed) https://aleta.io/family-office-software-security PR (sensitive)
AL-C09 Time-saved % claim (excluded) https://aleta.io/ai-ready-wealth-intelligence-platform PR (excluded claim)
AL-C10 Pricing not on human-facing pages; agent-targeted page ignored https://aleta.io/ PR (gap)
AO-C01 Consolidated total wealth view https://altoo.io/ PR
AO-C02 MCP to a user-chosen AI platform https://altoo.io/wealth-api-mcp/ PR
AO-C05 Swiss developed and hosted https://altoo.io/ PR (sensitive)
AO-C06 SOC 2 / ISO not stated https://altoo.io/security/ PR (gap)
AO-C07 Pricing not stated https://altoo.io/ PR (gap)
AO-C08 MCP read/write scope not stated https://altoo.io/wealth-api-mcp/ PR (gap)
AS-C01 Family-office GL and reporting software https://www.assetvantage.com/ PR
AS-C05 SOC 2 Type 2 compliance https://www.assetvantage.com/about-av/ PR (sensitive)
AS-C07 Generative-AI features not stated https://www.assetvantage.com/ PR (gap)
AS-C08 Pricing not stated https://www.assetvantage.com/ PR (gap)
AZ-C01 Aggregation from banks and custodians https://www.asora.com/ PR
AZ-C02 Price not re-verified (source unavailable 2026-11-03) https://www.asora.com/pricing PR
AZ-C03 Price not re-verified (source unavailable 2026-11-03) https://www.asora.com/pricing PR
AZ-C04 Price not re-verified (source unavailable 2026-11-03) https://www.asora.com/pricing PR
AZ-C05 Price not re-verified (source unavailable 2026-11-03) https://www.asora.com/pricing PR
AZ-C06 ISO 27001 by BSI; GDPR compliant https://www.asora.com/security PR (sensitive)
AZ-C08 Generative-AI features not stated https://www.asora.com/ PR (gap)
BD-C02 In-platform assistant; approved actions "available soon" https://www.sscblackdiamond.com/solutions/black-diamond-ai/ PR
BD-C03 Governed MCP server https://www.sscblackdiamond.com/solutions/black-diamond-ai/ PR
BD-C07 Model provider unknown n/a (gap recorded in dossier) PR (gap)
CN-C01 Alternatives document automation https://canoeintelligence.com/solutions/ PR
CN-C02 Automated rules flag items for expert review https://canoeintelligence.com/solutions/ PR
CN-C05 SOC 2 Type II (report not reviewed) https://canoeintelligence.com/ai/ PR (sensitive)
CN-C06 Customer documents not used to train public models https://canoeintelligence.com/ai/ PR (sensitive)
CN-C07 Retrieval-time % claim (excluded) https://canoeintelligence.com/ PR (excluded claim)
CN-C08 "100% data accuracy" (excluded) https://canoeintelligence.com/ PR (excluded claim)
V8-CAN-D3 API delivery to downstream systems https://canoeintelligence.com/ PR
ET-C01 AtlasFive consolidates accounting, reporting, documents https://eton-solutions.com/solutions/ PR
ET-C02 EtonAI capabilities https://eton-solutions.com/ PR
ET-C03 EtonAI formerly EtonGPT https://eton-solutions.com/solutions/atlasfive/ PR
ET-C04 BYOK encryption; ISO 42001 https://eton-solutions.com/ PR (sensitive)
ET-C07 "only fully integrated" superlative (excluded) https://eton-solutions.com/solutions/atlasfive/ PR (excluded claim)
ET-C08 Pricing not stated https://eton-solutions.com/ PR (gap)
FC-C01 SFO plan from $35,899/year plus fees https://fundcount.com/pricing PR
FC-C02 Prices are annual starting points https://fundcount.com/pricing PR
FC-C03 Document monitoring/extraction; exceptions reviewed https://fundcount.com/solutions/ai-document-intelligence PR
FC-C05 AI Assistant for reports and data https://fundcount.com/ PR
FC-C06 SOC 2 Type II, AICPA examined https://fundcount.com/pricing PR (sensitive)
FC-C08 Model provider not stated https://fundcount.com/solutions/ai-document-intelligence PR (gap)
HP-C01 OCR of tax returns https://info.holistiplan.com/holistiplan-tax-planning-financial-advisors PR
HP-C03 Roth conversion projection https://info.holistiplan.com/holistiplan-tax-planning-financial-advisors PR
HP-C04 $749/yr and $1,499/yr (30 HH) https://www.holistiplan.com/pricing/ PR
HP-C06 SOC 2 Type II and MFA https://www.holistiplan.com/pricing/ PR (sensitive)
HP-C07 Growth bps claim (excluded) https://info.holistiplan.com/holistiplan-tax-planning-financial-advisors PR (excluded claim)
LS-C01 Custodial data API, 500+ custodians https://www.landytech.com/ PR
LS-C02 aLi queries data in Sesame https://www.landytech.com/product/ali PR
LS-C04 GL accounting "coming soon" https://www.landytech.com/ PR
LS-C05 ISO 27001 https://www.landytech.com/security PR (sensitive)
LS-C06 Pricing not published https://www.landytech.com/ PR (gap)
MO-C01 Aggregation and real-time UHNW reporting https://masttro.com/ PR
MO-C02 AI for alternatives documents and capital calls https://masttro.com/ PR
MO-C04 Client-controlled keys; Swiss infrastructure https://masttro.com/ PR (sensitive)
MO-C05 Fixed, non-AUM pricing (no figure) https://masttro.com/platform/mobile-app PR
MO-C06 Price figure not stated https://masttro.com/pricing PR (gap)
MO-C08 SOC 2 / ISO not stated https://masttro.com/ PR (gap)
MO-INJ Agent-targeted page with pricing/competitor claims, not used https://masttro.com/ai-info-page TeachU research note (not a product claim)
V8-MAS-D3 Entity-scoped access for family stakeholders https://masttro.com/insights/ai-family-office-reports PR
TM-C01 Beta AI features; immutable audit logs; input guardrails https://help.tamaracinc.com/help/content/tamarac%5Freporting/enhancements/2026-08-21_-release_notes-_reporting.htm PR
TM-C05 Model provider unknown n/a (gap recorded in dossier) PR (gap)
V8-WB-L1 AI Agents may be early access by plan https://help.wealthbox.com/hc/en-us/articles/51252672927771-How-to-Use-AI-Assistant PR (analyst note)
WB-C01 AI Assistant answers questions on client record https://www.wealthbox.com/products/ai-assistant/ PR
WB-C02 Drafted communications reviewed before sending https://www.wealthbox.com/products/ai-assistant/ PR
WB-C04 SOC 2 Type II https://www.wealthbox.com/pricing/ PR (sensitive)
WB-C08 Model provider not stated https://www.wealthbox.com/products/ai-assistant/ PR (gap)

Counts: 71 claims. 70 are provider-reported (36 standard, 13 sensitive vendor claims TeachU hasn't reviewed, 5 vendor claims TeachU excludes from assessments, 15 gaps where something isn't stated, 1 TeachU analyst note), and 1 is a TeachU research note. Independently reported: 0. Verified: 0. Distinct source URLs: 35.

Browse the full software index, listed alphabetically by sector and task.