Claude Training for Family Offices
Dr. Leigh Coney
Founder, WorkWise Solutions
June 17, 2026
17 min read
TLDR: A family office is a small generalist team carrying a wide surface: consolidated reporting across managers and custodians, manager and fund diligence, direct-deal screening, entity and household admin, and market intelligence, with no technology bench and an unusually high privacy bar. That shape is exactly why Claude training pays off, and why privacy is the first lesson, not the last. Train the team on the workflow, not on Claude in the abstract. Build the Projects first so the assistant knows your entity map and house formats, teach the reporting and diligence work that scales worst by hand, add Cowork once the chat is trusted, and measure it by whether anyone would mind if you switched it off.
Table of Contents
1. A Lean Team With a Wide Surface
A family office is the strongest case for an assistant precisely because it is the smallest team carrying the widest job. Three or four generalists cover what a fund splits across a dozen specialists.
Statements from every manager and custodian pulled into one view, with the figures checked, not trusted.
Reports, side letters, and DDQs read for what actually changed since last quarter.
A forwarded CIM screened against the family's criteria and a first-cut memo drafted.
The research, the summaries, and the entity housekeeping that fill every gap in the week.
Look at the surface one person touches in a week. Consolidating statements from six managers and three custodians into one view. Reading a manager report and a side letter for what changed. Glancing at a direct deal a friend forwarded. Keeping the entity chart and the household admin straight. Watching the markets the principal cares about. None of that is a full-time role. All of it lands on the same desk.
That is why training pays off here faster than at a fund. There is no specialist to hand the assembly to, so the assistant is the bench you do not have. The point of this guide is to train the team on that actual work, not on Claude as a piece of software. For the wider context of where AI fits a family office, the family office AI guide sets the frame; this is the training playbook underneath it.
2. Privacy Is the First Lesson, Not the Last
At a family office, privacy is not a compliance footnote. It is the whole reason the office exists. So it is the first thing the team learns, before a single prompt.
The rule is simple and it is not optional. Family and investment data belongs on Claude Team or Claude Enterprise, on the office account, never on a personal login. On Team and Enterprise your business data is not used to train public models. That is the line that lets people put real statements and real names into the tool instead of awkward placeholders that make the work useless.
For the most sensitive material, Claude Cowork can run inside your own cloud or tenant, so the data stays inside your perimeter. Teach that boundary on day one and you remove the objection that otherwise kills adoption in a private office: the fear that the family's information is leaking somewhere. The full posture, written for a principal who has to be sure, is in data privacy and security for family offices.
3. Build the Projects First
Most training jumps straight to the chat box. For a family office, the order is backwards. You build the Projects first, then teach the work inside them.
A Claude Project is a shared workspace that holds custom instructions plus your own knowledge files. For a family office, that means the entity map, the house reporting format, the manager and fund list, the standing questions the principal asks. Load those once and every answer comes back in your structure, using your names, against your template, instead of generic output you have to translate every time.
This is the difference between a tool the team adopts and one they abandon. A blank assistant makes you re-explain the office in every session. A Project that already knows the office feels like a junior who read the files. Building those Projects well is most of the training, and the mechanics specific to an office are covered in Claude for family offices.
4. Consolidated Reporting Across Managers and Custodians
Start the real training on the job that scales worst by hand: consolidated reporting. Statements arrive in a dozen formats, on a dozen schedules, and someone stitches them into one view of the whole balance sheet.
Trained on a Project that holds the house format, Claude reads each statement, pulls the positions and the changes, and assembles the consolidated picture in your layout. The work that ate a slow week becomes a draft the team reviews. That is where the hours come back, because this is assembly, not judgment, and assembly is exactly what the assistant does well.
Teach one rule alongside it, hard. Claude is not a calculator. Every figure is checked against the source, not trusted because it looks right. The assistant drafts the report; a person ties out the numbers and signs it. Done that way, the speed is real and the trust is earned. The deeper workflow, with the controls, is in consolidated reporting for family offices.
5. Manager and Fund Due Diligence
The next workflow is reading. A family office lives in documents it did not write: DDQs, side letters, manager reports, fund quarterly letters. The job is to read each one against the questions the office actually cares about.
Put your diligence questions in the Project and the work changes shape. Hand Claude a DDQ and ask what changed from last year, what the side letter grants that the LPA does not, where the fees and the liquidity terms sit, what the manager is quiet about. You get a structured read in minutes, with the passages pulled so a person can check them. The reading still happens. It just starts from a draft instead of a blank page.
The discipline is the same as everywhere else in this playbook. The assistant flags and summarizes; the principal or the CIO concludes. A manager report does not become a decision because Claude read it. It becomes a faster, more thorough read that frees the human to spend their attention on the judgment that only they can make.
6. Screening Direct Deals
Family offices increasingly invest directly, and the deals arrive informally. A CIM from a friend, a deck from a co-investor, a forwarded teaser. Most are a polite no, and the cost is the hours spent figuring out which.
Train the team to use Claude for the first cut. Feed it the CIM and ask for a first-pass memo against the office's criteria: what the business does, how it makes money, the obvious risks, the questions to ask before spending another hour. A two-hour read becomes a twenty-minute review of a draft, and the deals that do not fit fall away early, cheaply.
The same limits hold. The first-cut memo is a screen, not a recommendation. Claude does not know today's market unless it is connected to it, and it does not replace the diligence on a deal that survives the screen. It gets the office through the inbound faster so the real work goes to the few deals that deserve it. More on this sits in AI for family office direct investments.
7. Market Intelligence and the Inbox
A principal asks the same kinds of questions on a loop. What is happening in this sector. What did this manager just publish. Summarize these three reports. What should I be watching this week. The answers are scattered across an inbox and a stack of PDFs.
Train the team to turn those standing questions into standing prompts. A research request, a summary of a long report, a digest of what changed since last week, all framed by the Project so the output matches what the principal wants to see. This is the lightest workflow to learn and often the first one a principal feels, because it answers their own recurring questions in their own format.
One caveat to teach plainly. Claude does not know the live market unless you connect it to a source, so current research either comes through a connector or gets verified against one. Treat it as a fast first read on a question, not as a live data feed, and it earns a permanent place in the week.
8. Add Cowork Once the Team Trusts the Chat
Everything above is the chat: ask, answer, review. Once the team trusts that, there is a second gear. Claude Cowork is the agentic mode that takes a whole multi-step task end to end on your own files, with a plan you approve and steer as it works.
The order matters. A team that has not yet learned to trust the chat will not hand an agent a reporting cycle, and they are right not to. But once they have watched Claude read statements and draft a consolidated view a dozen times, handing Cowork the whole sequence (pull the statements, build the report, flag what moved) is a natural next step rather than a leap of faith.
The same human boundary holds at this gear, only it matters more. The agent drafts and flags; a person concludes and signs. Cowork does the multi-step assembly across the reporting and review work; the principal still owns the decision and the signature. How that runs in an office, step by step, is in Claude Cowork for family offices.
9. Who Learns What on a Small Team
On a lean team you do not train by role, because everyone is several roles. You train by depth. The generalists go deep; the principal goes short.
The one or two people who carry the reporting, the diligence, and the deal screening learn the tool properly. They build the Projects, write the prompts, learn where Claude is strong and where it must be checked. They become fluent because they live in it daily, and that fluency is what makes the office faster.
The principal does not need that. A principal needs a short, honest briefing: what the assistant does well, where it must not be trusted, and what to ask of the team using it. Enough to decide with it and to govern it, not to operate it. That split, a few deep operators and a well-briefed principal, is the right shape for an office and matches how the firm runs its Executive Briefing against its hands-on training.
10. Make It Stick: One Champion and an Honest Read
Training that sticks needs one champion. On a small team, name the generalist who is genuinely interested and give them the room to own it: build the Projects, fix the prompts, answer the questions, show the principal what changed.
Then read the result honestly, with no invented metric. The test is the off switch. If you turned Claude off next month, would the team mind. If the consolidated report quietly went back to a manual week and someone groaned, it was adopted. If nobody noticed, it was installed, and the training has not landed yet.
That honesty is the whole game in a private office, where there is no large team to hide a failed rollout inside. WorkWise approaches adoption as a behavioral problem before a technical one, because the tool only sticks when a real person depends on it for real work. One champion, real workflows, and the off-switch test are how you get there.
11. Where to Start
Pick one workflow and start there. For most offices it is consolidated reporting, because it costs the most hours and the saving is felt immediately. Build the Project, train the one or two people who own it, and run it on the next real cycle.
For a lean office, the format that fits is small. An Executive Briefing gives the principal the short, honest read they need to decide and govern. A hands-on session gets the generalists building the Projects and running the first workflow on real files, with the privacy boundary set correctly from the first minute.
If you want the first workflow scoped before you train on it, an AI Readiness Sprint picks the job worth winning first and sets the data posture around it. From there, AI consulting for family offices or an ongoing AI Operating Partner runs it with you, so the office that has no technology bench has one anyway.
"The trick to working with AI is to treat it like a person, even though it isn't one... Treat it like an infinitely patient new coworker who forgets everything you tell them each new conversation, one that comes highly recommended but whose actual abilities are not that clear."
Ethan Mollick, "Co-Intelligence" (2024)
- •A family office is the strongest case for an assistant: the smallest team carrying the widest job, with no specialist bench to hand the assembly to.
- •Privacy is the first lesson, not the last. Family and investment data belongs on Claude Team or Enterprise, on the office account, never a personal login.
- •On Team and Enterprise your business data is not used to train public models, and for the most sensitive work Cowork can run inside your own tenant.
- •Build the Projects first. A Project holding your entity map, house reporting format, and manager list makes every answer come back in your structure.
- •Train on the work that scales worst by hand: consolidated reporting, manager and fund diligence, and first-cut memos on direct deals.
- •Claude is not a calculator and does not know the live market. Figures are checked against the source; the agent drafts and flags, a person signs.
- •Make it stick with one champion and the off-switch test: if you turned Claude off next month and nobody minded, it was installed, not adopted.
Related Guides & Articles
AI for Family Offices: The Complete Guide
The wider frame this training sits inside: where AI fits a lean office, from reporting to diligence to direct deals.
Claude for Family Offices
The mechanics of running Claude in an office: building the Projects that hold your entity map, house format, and manager list.
Claude Cowork for Family Offices
The agentic gear: handing a whole multi-step reporting and review cycle to an agent, with a plan you approve and steer.
AI for Family Office Consolidated Reporting
The reporting workflow in depth: statements to one view, with the controls that keep figures checked, not trusted.
AI Data Privacy and Security for Family Offices
The privacy posture, written for a principal: Team and Enterprise, no training on your data, and the in-tenant option.
AI Consulting for Family Offices
Scope, build, and run the first workflow with you, so an office with no technology bench has one anyway.
Want a lean team trained on Claude the right way?
For a small office, an Executive Briefing gives the principal the short, honest read to decide and govern, and a hands-on session gets the generalists building Projects and running the first workflow on real files. An AI Readiness Sprint scopes the workflow worth winning first; AI consulting for family offices then runs it with you.
Book a Call