Private SuperIntelligence by Mitosis Labs

A coverage map for family office AI: which tool reads which room

PJ (Prakshal Jain) — Wed, 21 Oct 2026 13:26:01 GMT

Disclosure: I co-founded Mitosis Labs, which runs Private SuperIntelligence. Adapted from the original guide.

Picture a family office as a floor plan. One room holds the investments. Others hold email, documents, the household and travel. Every AI tool a family office might buy lights up some of those rooms and leaves the rest dark.

That picture is more useful than a ranking, because "best" only makes sense once you know which rooms you need covered. Below is the map, room by room, built from each vendor's own page (and for Aurora, a Fortune report), accessed 2026-10-15.

The map

Tool group	Investments	Email and documents	Household and travel
Enterprise seats (ChatGPT Enterprise, Copilot, Claude Enterprise)	Not described	Through what the office connects (Copilot: Microsoft Graph; Claude: Gmail, Google Drive, Slack, Microsoft 365)	Not described
Platforms adding AI (Addepar, Masttro, Asora, Eton, Aleta)	Yes, the platform's own data	Fund documents the platform ingests; the inbox stays outside	Outside it
Private chat apps (Lumo, Tinfoil, Venice)	No link to office information	No link to office information	No link to office information
Concierge services (Knightsbridge Circle, Aurora)	Not described	Not described	Yes, lifestyle and travel

"Not described" means the vendor's page, as cited in the original, does not describe that coverage. It is not a claim the product cannot do it.

The investments room

The platforms are built for this room, because the portfolio data already lives in them.

Addepar. A March 2026 Addepar post says Addison "draws on Addepar's unified data model to answer questions, surface relevant guidance and direct users to the right workflows." It fits investment teams that already report from Addepar.
Masttro. DocAI ingests PDFs for liquid and illiquid assets and handles capital calls, distributions and valuation statements, pulling key data points onto the dashboard.
Eton Solutions. EtonAI runs inside AtlasFive or on its own, with document extraction across 250+ document types, natural language querying and a knowledge graph. Eton describes it as a private, single-tenant instance and lists ISO 42001, ISO 27001, ISO 27701, SOC 1 Type 2 and SOC 2 Type 2.
Aleta. Reads capital calls, K-1s and fund statements, and supports MCP so AI agents can query wealth data in plain language. Aleta lists SOC 2 Type II and encryption at rest and in transit.
Asora. The AI Connector exposes Asora wealth data to Claude, ChatGPT, Gemini, Grok or any MCP-capable assistant, read-only, scoped to existing permissions, with every query audit-logged. Asora says it never uses your prompts or data for training, and notes that once an answer reaches your assistant, that provider's terms apply.

The MCP pattern from Asora and Aleta is worth noticing as an architecture choice. The platform stays the source, and the conversation happens in whichever assistant you connect. Your data-flow diagram then needs a line for that assistant's terms too.

These tools fit the CIO, CFO and controller. Each assistant answers from what sits in its platform, so the principal's inbox, the household's calendar, deal email and travel stay outside it.

The email and documents rooms

Enterprise seats can reach here, if the office wires them in.

Microsoft says Copilot works across email, chats, documents and meetings in Microsoft Graph, and that prompts, responses and Graph data are not used to train foundation models. Copilot shows each user any content that user has at least view permission for. A trust folder shared with the whole company years ago can surface for anyone in the company, so audit sharing first.
Anthropic lists Claude Enterprise connectors to Gmail, Google Drive, Slack and Microsoft 365, plus SSO, SCIM provisioning, role-based access control, audit logs, data retention controls and customer-managed encryption keys. Anthropic states it does not train on commercial inputs or outputs by default, though a conversation rated thumbs up or down can be stored for up to five years and used for training. An organization owner can turn the rating feature off.
OpenAI says it does not train its models on business data by default, that you own your inputs and outputs where the law allows, and that ChatGPT Enterprise supports SAML SSO, admin-set retention and a Compliance API audit log. Eligible new Enterprise customers can store content at rest in regions including the United States, the United Kingdom and Europe.

In every case the office decides retention, which apps connect and who gets a seat. That is real work for whoever owns IT.

Personal chats

Private chat apps make strong privacy promises for one person's conversations. Proton says Lumo does not log chats, uses zero-access encryption so "no one (not even Proton)" can read them, and never trains on your data. Tinfoil encrypts data to a secure enclave Tinfoil itself cannot access, verifiable through hardware attestation. Venice keeps history on your device and relays requests without storing them. None connects to custodians, entity records or the shared drive on its own, so for an office they are do it yourself.

The household and travel rooms

Knightsbridge Circle describes itself as "a dedicated extension of your personal office," with a personal manager who looks after four members. Fortune reported on 2026-10-11 that Aurora is an invite-only app pairing AI with human experts for dining, travel, luxury goods and wellness, with connected calendars and wearables such as Whoop and Oura used to anticipate requests. Both fit lifestyle and travel requests.

Reading the map

Before you choose, answer five questions for your own office:

Which information should the AI reach: the portfolio, the principal's email, entity and trust documents, or all of it?
Who in your office will set it up, review permissions and keep it current?
Does the provider train on your data, and what do its feedback buttons do?
Where is the data stored, and who at the vendor can see it?
Do the answers carry citations your team can check?

Where Private SuperIntelligence fits

Private SuperIntelligence is an AI concierge for family offices. It works across the family's own email, documents and data, answers with citations, and acts within permissions the family sets. Each deployment is built for one client and works alongside the platforms above.

Private SuperIntelligence takes a limited number of families at a time so every one gets the same standard of service. Book a private call (ref BAF1-HN) to talk through your office with Alex Morris.

Prakshal Jain (PJ), Co-founder and CTO, Private SuperIntelligence (by Mitosis Labs)

Alexander Morris, Co-founder and CEO, Private SuperIntelligence (by Mitosis Labs)

Originally published at privatesuperintelligence.si. Vendor pages accessed 2026-10-15; full source list in the original.

One fee holiday, two systems: who answers what when AI meets your portfolio platform

PJ (Prakshal Jain) — Tue, 20 Oct 2026 13:26:17 GMT

Disclosure: I co-founded Mitosis Labs, which runs Private SuperIntelligence. Adapted from the original comparison.

Here is a question a family office principal might ask on a Tuesday: "What did we agree with the GP about the fee holiday, and does it show in Q3?"

It sounds like one question. In systems terms it is two queries against two stores. The agreement lives in an email thread. The Q3 effect lives in the capital account. One tool holds the email and the other holds the capital account.

That split is the whole design problem when a family office adds AI. Addepar and Masttro, two portfolio platforms built for this work, both now ship AI that answers questions about portfolio data. They do that well. The work for the person wiring things together is deciding which AI is allowed to answer which half of a question, under which permissions, with what proof.

I read both vendors' own pages on 2026-10-15. Everything below about them comes from those pages, linked at the end.

Store one: the portfolio system of record

Addepar and Masttro aggregate custodian and alternatives data, map ownership across entities, and produce consolidated reporting.

Addepar says it can "transform multi-custodial data, illiquid alternatives and complex legal entities into a unified, single source of truth," mapping ownership dynamically so you can "analyze look-through ownership and calculate total net worth across family branches." It covers private equity, venture capital, hedge funds, real estate and direct investments, and calls illiquid and alternative assets "one of Addepar's core strengths."

Masttro lists traditional assets, alternatives (private equity, hedge funds, venture capital, direct investments, private credit, crypto), real estate, collections such as art, jewelry and vehicles, and entity structures: companies, trusts, LLCs and foundations. It delivers "consolidated, brandable reports to trustees, the next generation, or an outside advisor."

Both are strong at the job they are hired for: one trusted set of numbers across custodians, managers and entities.

The AI layer on store one

	Addepar	Masttro
AI product	Addison, native AI experience launched March 3, 2026	Masttro Intelligence: Documents AI, Alternatives AI, Masttro Assistant, Reports Playground
What it works on	Addepar's unified data model	Data, documents and reports inside the Masttro platform
Stated controls	"permission-aware access and traceable outputs grounded in real portfolio context"	Swiss-based infrastructure, ISO/IEC 27001:2022, "no client data is ever used to train the AI modules"
Extra surface	"Addepar MCP beta" bringing Addepar intelligence into Claude, with a waitlist	Documents AI processes capital calls, distribution memos and K-1s

Addepar says Addison helps with "summarizing portfolio performance, investigating drivers of returns or allocation changes, reviewing exposures or preparing more effectively for client conversations." Masttro says its Documents AI "automatically ingests, processes, and catalogs any document into a centralized hub," its Alternatives AI "collects, categorizes, extracts transaction-level data," and Masttro Assistant lets you "interact conversationally with their wealth data by voice or text." Masttro also states that "client data never leaves the security of the Masttro platform."

For an engineer, the useful reading of that table is the boundary. Each AI is scoped to its own platform. That is a feature: the numbers come from the place that owns them.

Store two: everything around the numbers

Neither company's pages describe working across the principal's inbox. Store two is usually email, shared drives and people's heads:

Correspondence: estate counsel on a trust amendment, the CPA's K-1 questions, a co-investor's side letter.
Entities beyond the ledger: board minutes, operating agreements, and why a structure was set up the way it was.
The household: staff, properties, travel, insurance renewals, school deadlines, the principal's own requests.
Deals in motion: memos, data rooms and diligence notes for investments that have not reached the portfolio yet.

Back to the fee holiday. Half the answer is in store two, half in store one. If only store one has an AI, the agreement half falls to whoever remembers the thread.

Three design rules for the seam

Rule 1: Numbers have one owner. Does the new tool duplicate the system of record, or sit beside it? Duplicated numbers drift. If a second tool keeps its own copy of positions, sooner or later the two will disagree and someone will have to reconcile them by hand. Let the portfolio platform stay the source of the numbers, and have the other tool point to it rather than restate it.

Rule 2: Permissions follow the data, not the tool. Permission-aware answers matter more when the AI reads email, because staff should not see the principal's private threads. A portfolio permission model is built around accounts and entities. Email needs one built around people and threads. Make sure whatever reads store two enforces its own.

Rule 3: Every answer carries its source. Can you trace every answer to a source document? Citations let your team check the answer before acting on it. For a two-part question like the fee holiday, that means each half should say where it came from.

Questions for every vendor, in writing

Before anything new touches the office's data, I would ask:

Which data does each AI feature read? Addison and Masttro Assistant work on the data inside their own platforms. Anything else needs its own tool.
Who can see what?
Where does the data go for inference, and is it used for training? Masttro answers this on its page. Ask every vendor in writing.
Can every answer be traced to a source document?
Does the tool duplicate the system of record, or sit beside it?

Where Private SuperIntelligence fits

Private SuperIntelligence from Mitosis Labs is an AI concierge for family offices, enterprises and private clients, built for store two. It reads the office's own email and documents, answers with citations back to the source, and acts within the permissions you set. The portfolio platform stays the system of record for the numbers. Each deployment is built for one client, from the hardware to the choice of model provider to where the data is stored.

You do not have to replace Addepar or Masttro to use it. It sits beside them.

Speak with the founders

Private SuperIntelligence takes a limited number of families at a time so every one gets the same standard of service. Book a private call (ref AM1-HN) to talk through your office with Alex Morris.

Prakshal Jain (PJ), Co-founder and CTO, Private SuperIntelligence (by Mitosis Labs)

Alexander Morris, Co-founder and CEO, Private SuperIntelligence (by Mitosis Labs)

Originally published at privatesuperintelligence.si. Vendor pages read 2026-10-15.

Sources

AI readiness check for family offices

PJ (Prakshal Jain) — Mon, 19 Oct 2026 20:57:01 GMT

Ten yes/no questions. Give your office one point for each yes you can back with a document or a test record. The total matters less than the open answers: they are the list of things to settle before AI touches live family email, documents or reporting.

Prakshal Jain (PJ), Co-founder and CTO, Private SuperIntelligence (by Mitosis Labs)

Alexander Morris, Co-founder and CEO, Private SuperIntelligence (by Mitosis Labs)

We wrote this for the operator who will be asked, sooner or later, "are we ready to use AI on the family's information?" by a principal who has already seen a colleague do it. Answer on paper or in your head. This page has no answer fields and no scoring script, so it cannot collect, store or send anything you decide.

Does this sound like your office?

Some staff are trying AI before anyone owns the decision. Citi's 2025 Global Family Office Report found 22% of family offices use AI for operational tasks or investment analysis, up from 13% in 2024. [1]
Nobody in-house feels qualified to judge the tools. In the same Citi survey, 57% of family offices named lack of internal expertise as the biggest barrier to adoption. [1]
There is no written AI policy yet. IBM's 2025 breach study found 63% of breached organizations had no AI governance policy or were still developing one. [2]
Someone may be using a tool the office never approved. IBM found one in five organizations reported a breach involving shadow AI, and only 37% had policies to manage AI or detect shadow AI. [2]
Reporting is the first job people want to hand over. Citi reports 16% of family offices use AI for investment performance reporting, more than double the year before. [1]

If two or more of these fit, the check below is worth ten minutes.

How do I score it?

One point for each yes you can support with a current document or test record. No and unsure both score zero. The total runs from 0 to 10, and every question points the same way: yes means the evidence exists.

The ten questions

1. Have we written down one task and the person accountable for it?

Yes if a named person can show the task, the records it may use and what the output should look like. Citi finds the AI uses that succeed in family offices are everyday ones: summarizing documents, transcribing meetings, managing email and automating reports. [1] Starting with one of those, with one owner, keeps the rest of this check concrete.

2. Do staff have a written rule for which records may go into the chosen AI tool?

Yes if the rule names permitted and excluded records and someone owns exceptions. Citi calls data privacy non-negotiable for family offices. [1] Among the breached organizations IBM studied, 63% had no AI policy or were still writing one. [2] Our one-page family office AI policy is a starting template.

3. Can someone explain the training terms for the exact plan the office uses?

Yes if you have the provider's terms for that plan and know any opt-in setting. Terms differ by plan. OpenAI, for example, states that data sent to its API is not used to train its models unless the customer opts in, which says nothing on its own about a personal chat account. [3] Our guide to whether your office's data is being exposed through AI sets out each major provider's terms by plan.

4. Have we separated what is kept from what is trained on?

Yes if your evidence lists every place the proposed setup keeps data. On its API, OpenAI does not train on your data by default, yet it keeps abuse monitoring logs, which can include prompts and responses, for up to 30 days. Excluding content from those logs currently needs OpenAI's prior approval. [3] "Not used for training" and "not kept" are different answers.

5. Do we know every service that receives the task's data?

Yes if you can name every company that sees the data: the assistant, the model provider and anyone else they pass it to. A brand name on a contract does not tell you that. Every extra recipient is another set of terms to check, so question 3 applies to each one.

6. Have we tested restricted access with harmless records?

Yes if a test with invented records shows the expected result for each role. OWASP recommends minimum permissions, acting in the user's own context and enforcing authorization in the downstream system rather than trusting the model. [4] A polite refusal in the chat window is not that test.

7. Have we tested that sending needs the intended approval?

Yes if test records show nothing went out unapproved. OWASP's own example is a mailbox assistant that only needs to read mail, so it should not be able to send or delete, and it recommends a human approve high-impact actions. [4]

8. Can a reviewer trace a reported figure back to its accepted source?

Yes if a sample report keeps the source, the period and who accepted it, or if reporting is outside the pilot. Addepar describes automated verification checks that flag data issues for its team to investigate. [5] Any AI that drafts reporting should go further and show where each figure came from.

9. Is a qualified person assigned to settle missing or conflicting evidence?

Yes if open items have an owner and cannot quietly become accepted answers. Citi notes that family offices keep final decisions, especially on investments, with experienced people. [1]

10. Do we have an agreed response when deletion cannot be checked directly?

Yes if the office records what it can inspect, what rests on contract or attestation, and who decides whether that gap blocks use. OpenAI's own page shows data can sit in more than one store. [3] A search that returns nothing does not prove erasure.

What does my score mean?

These bands are our editorial guidance, not research findings.

0 to 3: start with a written policy and one task. Name the owner, the permitted records and who approves. Keep live family information out until that exists.
4 to 7: close the open answers. Give every no or unsure an owner and a date. Test with invented records while terms, permissions and retention are checked.
8 to 10: review the evidence, not the total. Have the accountable operator, whoever handles security, and counsel read the open limits before any live pilot. Reporting work also needs its reporting owner or accountant.

At any score, a single unresolved answer on access, unapproved sending or data handling can be a reason to stop. A yes elsewhere does not offset it. Run the check again when the provider, the plan, the task or the data changes.

For the vendor conversation itself, use our five questions to ask any AI vendor before it reads family email. For reporting, read consolidated reporting across custodians.

Where does Private SuperIntelligence fit?

Private SuperIntelligence is private AI set up across the family's own email, entities and documents, deployed in an environment built for that family alone. It drafts freely while sending and payments need a person, and consolidated reporting cites every figure, so questions 7 and 8 become settings you can test rather than promises. [6] It does not replace questions 1, 2 and 9: those decisions stay with the office.

Speak with the founders

Private SuperIntelligence takes a limited number of families at a time so every one gets the same standard of service. Book a private call to talk through your office with Alex Morris.

Disclosure: Prakshal Jain and Alexander Morris co-founded Mitosis Labs, which runs Private SuperIntelligence.

Frequently asked questions

Does this page calculate or save my score?

No. There are no answer fields, no automatic scoring and nothing to submit. You count your own yes answers.

Does a score of 10 mean our AI setup is safe?

No. The check is a planning aid, not a validated test or a certification. Qualified people in the office still decide whether a proposed setup meets its requirements.

How is this different from the AI data exposure check?

The AI data exposure check looks at where information goes under a provider's published terms for the tools your staff already use. This readiness check looks at the office's own evidence before it connects any AI to live records.

Who in a family office should run this check?

The person who will own the AI decision, often the COO, chief of staff or general counsel, with the principal seeing the result.

Sources

[1] Citi Institute. AI in the Family Office: Privacy, Efficiency and Institutional Rigor. May 11, 2026, citing Citi Wealth's 2025 Global Family Office Report. Accessed 2026-10-16.

[2] IBM Newsroom. IBM Report: 13% of organizations reported breaches of AI models or applications. July 30, 2025. Accessed 2026-10-16. Figures describe breached organizations studied by IBM, not family offices specifically.

[3] OpenAI. Data controls in the OpenAI platform. Accessed 2026-10-16. API terms only, not every OpenAI product.

[4] OWASP Gen AI Security Project. LLM06:2025 Excessive Agency. Accessed 2026-10-16.

[5] Addepar. Why Addepar. Accessed 2026-10-16. Vendor's own description, not an independent assessment.

[6] Private SuperIntelligence. AI for family offices. Accessed 2026-10-16.

Originally published at privatesuperintelligence.si.

A private-AI questionnaire for family offices, built from the providers' own policies

PJ (Prakshal Jain) — Fri, 16 Oct 2026 09:37:58 GMT

Disclosure: we co-founded Mitosis Labs, which runs Private SuperIntelligence. Adapted from our original guide.

Most AI due diligence at a family office stops at one clause: "We do not train on your data." It is a good clause, and the business plans of the four providers below all offer it. It is also the least interesting answer on the page.

We spent 16 October 2026 reading what OpenAI, Anthropic, Google and Microsoft publish about training, retention, subprocessors and connectors. This piece turns that reading into a questionnaire you can send to any vendor, including us.

What the published pages already settle

Training. OpenAI does not train on ChatGPT Business, Enterprise, Edu or API content by default; personal Free, Plus and Pro accounts train unless the user opts out. [1] Anthropic's commercial products do not train unless you send feedback or opt in. [3] Google Workspace with Gemini does not use your content without permission. [5] Microsoft 365 Copilot does not train foundation models on your tenant's data. [6]

Retention. Deleted ChatGPT and Claude chats are removed within 30 days, with legal and safety exceptions. [2][4] Google keeps Gemini chats that human reviewers saw for up to three years, even after you delete your activity. [13]

Subprocessors. OpenAI lists Microsoft, CoreWeave, Oracle and Amazon Web Services as cloud hosts for ChatGPT Business and Enterprise, and TaskUs for support and for content moderation on the API and ChatGPT Business. [7] Microsoft says it "enables Anthropic models on by default for most customers in commercial cloud" for Microsoft 365 Copilot, outside Europe and the UK. [8]

So the clause you were going to ask about is already answered. The useful questions start with the ones every hosted service answers the same way for everyone: questions 1 and 2 below.

The questionnaire

Send these five, and ask for the answers in writing.

1. Where does it run? On whose hardware, in which location, is our information processed? Name every cloud host.

2. Who runs it? Who operates the system day to day, which of their staff or contractors can reach our content, and who do they answer to?

3. What reports home? What does the deployment send back to you: prompt content, logs, usage telemetry, error reports, or nothing?

4. Who sees what? Can we set which staff, advisers and family members see which records? Which actions (sending email, moving money) always wait for a person? What can connected apps read on a user's behalf?

5. Where do answers come from? Can every answer be traced to the document behind it?

How to read the answers

For any hosted service, questions 1 and 2 have the same answer: the provider's data centers and the provider's staff and contractors. That is not a flaw; it is what hosting means. Hosted plans can still answer 3 to 5 well, and Microsoft and OpenAI document permissions in detail. [6][9]

Question 4 is where we would look first. Microsoft 365 Copilot "only surfaces organizational data to which individual users have at least view permissions." [6] That is sound access control, and it means a folder shared too widely years ago is now one natural-language question away. Fix sharing before connecting anything.

Where Private SuperIntelligence fits

We built Private SuperIntelligence so a family decides the answers to questions 1 and 3, and the people answering question 2 report to the family. Each deployment is built for one client, from the hardware to the choice of model provider to where the data is stored. A client deployment captures none of the client's data and sends no telemetry back to Mitosis Labs. It can run air-gapped, with the client holding the encryption keys. [10][11] It answers with citations, drafts within permissions the office sets, and leaves sending and payments to a person, and Mitosis Labs runs it with a named contact. [12] We take a limited number of families at a time. [11]

Before you send the questionnaire anywhere, take the AI data exposure check. It is five questions and takes under a minute.

Prakshal Jain (PJ), Co-founder and CTO, Private SuperIntelligence (by Mitosis Labs)

Alexander Morris, Co-founder and CEO, Private SuperIntelligence (by Mitosis Labs)

Originally published at privatesuperintelligence.si. Sources checked 16 October 2026.

Sources

Family-office AI: start with the capital-call handoff, not the chatbot

PJ (Prakshal Jain) — Fri, 16 Oct 2026 04:01:48 GMT

ChatGPT Enterprise gives a team a capable hosted AI workspace. A family office looking for a private system built around its records, permissions and daily work needs someone to build and operate that system. Private SuperIntelligence (PSI) is designed for that job. A capital-call notice is a useful place to see the difference: the value is in connecting the notice to the correct entity, source records and human approver. 7

Disclosure: I co-founded Mitosis Labs, which runs Private SuperIntelligence.

Does this sound like your office?

Reports, correspondence and document summaries consume the same hours every week. 6
The office has no spare internal expertise to implement and maintain AI. 6
Information needed for an answer is spread across custodians, accountants, fund administrators and email. 5
Family members, staff and external advisers need different access to records. 5
The office wants a clear processing location before connecting the trusts or deal room. 5

For an office facing these problems, the buying decision includes implementation and ongoing operation, not just access to a model.

What should a capital-call workflow actually produce?

It should produce a draft that a person can check against the records before approving any action.

Start the demonstration with a notice the office has authorized for the exercise. Follow the records used to match it to an entity, commitment and signatory. Then inspect the proposed response and its citations. A person remains responsible for sending and paying. That is the handoff described in PSI's family-office service, rather than an invitation to automate money movement. 7

This exercise makes the implementation conversation concrete. Ask which records the operator may see, which remain restricted to the principal, and how a signatory change reaches the system. The answers should describe the office's real responsibilities and access rules.

The same connected context supports other daily work. PSI describes consolidated reporting with figures cited to source statements, tracking deadlines and signatories across entities, drafting correspondence in the principal's voice, and answering private succession and scenario questions. The office retains judgment and approval while Mitosis Labs handles the technology. 7 8

Are you buying a workspace or an operated service?

The difference is who builds the environment and keeps it useful as the office changes.

OpenAI supplies ChatGPT Enterprise as a hosted workspace with documented controls and integrations. Your administrators configure access and govern its use. PSI builds a deployment for one family, connecting its own information and running the service with a named contact. 1 3 4 7

ChatGPT Enterprise	Private SuperIntelligence
Implementation: administrators configure the workspace, invite people and manage integrations 3 4	Implementation: Mitosis Labs builds around the family's entities, advisers, email and documents 7
Ongoing operation: administrators review access and monitor use 4	Ongoing operation: Mitosis Labs runs the service with a named contact 7
Processing: OpenAI's service, hosted with major cloud providers 1	Processing: a client-specific environment, from isolated cloud to air-gapped 7
Training: business content is not used for training by default 2	Training: the provider keeps no copy of the work and trains nothing on it 8
Work: the workspace and connected tools 3	Work: reporting, entities, correspondence, memos and private questions across the office 7

Illustration. Sources: OpenAI hosting, integrations, administration and PSI service, checked 15 October 2026.

The staffing question matters. Citi's May 2026 article, citing its 2025 Global Family Office Report, says 57% of family offices name lack of internal expertise as their biggest barrier to AI adoption. 6 The office still needs someone responsible for connecting information, maintaining permissions and supporting changes. For a family wanting that work handled for it, PSI is the direct fit.

What does private processing add to useful work?

It gives the office an explicit environment in which to connect sensitive information and set permissions.

OpenAI documents meaningful Enterprise protections: no training on business content by default, encryption in transit and at rest, and administrator controls. Content is processed in OpenAI's hosted service to answer requests. 1 2 A no-training rule and a processing location answer different questions; both belong in the deployment discussion.

PSI is designed for one client, including hardware, model-provider selection and storage location. The service description says the provider retains no copy of the work, does not train on it and cannot see outputs. A client deployment sends no telemetry back to Mitosis Labs. 8 That gives the family a basis for bringing together the information needed to answer operational questions.

Deloitte's 2024 report found that 43% of family offices had experienced a cyberattack in the previous 12 to 24 months. 9 RBC and Campden Wealth's 2026 report identifies cybersecurity and data breaches as the top near-term operational concern, cited by 59% of respondents. 10 These figures explain why processing, visibility and permissions deserve explicit answers during procurement.

Source: Deloitte Family Office Cybersecurity Report 2024, survey of 354 single-family offices from September to December 2023. Groups overlap; checked 15 October 2026.

How should the operator check an answer?

Open the source behind the claim and compare it with the draft before approval.

PSI answers with citations to the office's records. 7 8 For the capital-call exercise, the reviewer should be able to follow the entity and signatory information behind the proposed response. Connected context is useful when it makes that review easier, not when it hides the decision from the person responsible.

Mitosis Labs also publishes Cortex research. Its research page reports 96% correct answers, 0.0% incorrect answers and a 4.4-second answer time in the Agentic Memory Index's 272-task evaluation. 12 The source note below states the evaluation's scope.

Source: Mitosis Labs research, reporting the Verging Labs revision of 3 August 2026. Provider-reported evaluation, not a PSI client result.

When is ChatGPT Enterprise the right call?

An office that wants a general staff assistant, accepts hosted processing and has an internal owner for configuration and governance is well served by ChatGPT Enterprise. Its documented integrations, security protections and administration tools deserve a serious evaluation. 2 3 4

PSI is the clearer choice when the requirement is a private environment built for the family's information and operated for it. That is a service purchase, including the work of implementation and daily operation. 7 8

Five questions to ask any AI provider

Will any of our content be used to train or improve a model?
Where will the work be processed, and who operates that environment?
What can the provider see, retain or receive as telemetry?
How do staff, advisers and family members get different permissions, and which actions require human approval?
Who implements the system, maintains it and handles changes to the office?

Written answers should accompany the demonstration, so the operator can explain the service and its responsibilities to the principal.

Speak with the founders

Private SuperIntelligence takes a limited number of families at a time so every one gets the same standard of service. Book a private call to talk through your office with Alex Morris.

Prakshal Jain (PJ), Co-founder and CTO, Private SuperIntelligence (by Mitosis Labs)

Alexander Morris, Co-founder and CEO, Private SuperIntelligence (by Mitosis Labs)

Private SuperIntelligence is powered by Mitosis Labs.

Adaptation A1. Canonical source: ChatGPT Enterprise or Private SuperIntelligence: choosing AI for a family office.

Sources and context

Sources checked 15 October 2026. Cortex figures describe a memory-system evaluation, not a PSI client result or a ChatGPT Enterprise comparison. The selected model provider and any external processing must be identified in the deployment design. Deloitte describes historical survey groups; the RBC release covers 155 responding offices across the Americas. Workflow diagrams are illustrations, not measured client outcomes.