Family office AI, automation and technology glossary

Agentic AI describes AI systems that can plan and act toward a goal with a degree of autonomy.

An AI agent is usually an individual software component. Agentic AI describes the broader approach or behavior of the system.

Agentic AI can be useful for open-ended work. It also requires stronger controls because the path taken by the system may not be fully predetermined.

An AI agent is software that uses artificial intelligence to pursue a goal and decide what action to take next.

Unlike a fixed workflow, an AI agent can adapt its next step based on what it finds.

In a family office, AI agents can support work that cannot be fully defined in advance. Their permissions should remain controlled. Higher-risk actions should remain subject to appropriate review.

AI governance is the framework used to control how artificial intelligence is selected, used and monitored within an organization.

For a family office, AI governance establishes responsibility for AI use. It also defines where AI may operate and where human approval is required.

Strong AI governance helps make AI adoption more controlled and accountable.

AI guardrails are controls that restrict what an AI system is allowed to do.

A guardrail can limit access to information. Another guardrail can prevent an AI system from taking an action without approval.

Guardrails reduce risk. They do not guarantee that an AI system will always produce a correct result.

An AI hallucination is an AI-generated response that appears credible but is false or unsupported by the available information.

Hallucinations are especially important in family office workflows because AI may interpret information that later informs reporting or operational decisions.

Important outputs should therefore be checked against trusted source information.

An AI operating layer is an emerging term for a governed technology layer that coordinates AI and automation across existing business processes.

There is not yet one universally accepted industry definition.

SFO Logic uses the term to describe a controlled layer that coordinates work around the systems a family office already uses. Existing systems can remain authoritative while the AI operating layer manages how work moves around them.

The AI operating layer can determine when AI is appropriate. It can also preserve deterministic controls and human review where they are more suitable.

AI workflow automation is the use of artificial intelligence within a defined automated workflow.

The workflow provides the structure. AI handles steps that require interpretation.

For example, AI may interpret an incoming capital call notice. The workflow can then move the information through review before further actions take place.

The phrase AI workflow is also commonly used for this type of process.

API integration is the process of connecting software systems through their APIs.

A successful API integration allows information or approved actions to move between applications automatically.

For family offices, API integration can reduce manual transfers between systems. It is most useful where reliable and permissioned connectivity already exists.

The application layer is the part of a technology architecture where software performs practical tasks for users.

In an AI architecture, the application layer sits above the underlying AI models and infrastructure.

A family office may interact with an AI application without interacting directly with the underlying model.

The phrase AI application layer is also used when the architecture being discussed is specifically focused on AI.

An application programming interface, or API, is a defined way for one software system to communicate with another.

An API allows software to request information or trigger approved functionality without requiring manual data entry.

In family office operations, APIs can connect existing systems to automated workflows when suitable access is available.

An audit trail is a chronological record that makes it possible to reconstruct what happened during a process.

A strong audit trail shows when an action occurred. It also shows who or what performed the action.

In a family office, audit trails improve accountability across automated workflows and human approvals.

The back office covers the administrative and operational work that supports the day-to-day running of an organization.

In a family office, the back office often handles recurring processes that sit behind investment and family activities.

Back-office automation can reduce manual effort while preserving appropriate control.

A board pack is the set of materials prepared for a formal decision-making meeting.

In a family office, a board pack may support a board or investment committee. It may also support another governance body.

Preparing the pack often requires information to be collected and reviewed before it can be distributed.

Business process automation, or BPA, is the use of technology to automate recurring business processes.

BPA normally focuses on the complete process rather than one isolated task.

In a family office, BPA can help make recurring operational work more consistent and less dependent on manual coordination.

A capital call is a request from a private investment fund for an investor to contribute part of its committed capital.

The request is normally communicated through a formal notice.

For a family office, a capital call creates an operational obligation. It also affects liquidity planning and investment records.

A capital call workflow is the operational process used to manage a capital call from receipt through completion.

The workflow begins when the notice arrives. It ends when the required actions and records have been completed.

AI can help interpret the notice. Automation can coordinate the process. Human review can remain in place before sensitive actions occur.

The term capital call processing is also used for this type of workflow.

Commitment tracking is the process of monitoring capital committed to private investments over time.

It helps a family office understand what has already been funded and what remains outstanding.

The term commitment monitoring is also used for this activity.

Data extraction is the process of identifying useful information within a source and converting it into a form that can be used elsewhere.

In a family office, data extraction often turns information contained in documents into data that can enter a workflow.

It can also extract information from existing systems where appropriate.

Data governance is the framework used to control how organizational data is managed.

It establishes ownership and responsibility for data.

For a family office, data governance supports more reliable information and safer use of AI.

Data lineage shows where data came from and how it changed as it moved through a process.

Good lineage allows reported information to be traced back toward its source.

This is particularly important when family office information passes through several systems before reaching a final report.

Data reconciliation is the process of comparing information from different sources to identify and resolve inconsistencies.

Reconciliation helps determine whether information can be relied upon.

In a family office, automated reconciliation can reduce manual checking while routing unresolved differences for review.

Data residency refers to the geographic location where data is stored or processed.

Data residency can influence how a family office evaluates technology and AI providers.

The location of data does not by itself guarantee privacy or compliance.

Document automation uses technology to reduce manual work around document-based processes.

It can begin when a document is received. The workflow can then move the document through processing and review.

For family offices, document automation is particularly relevant because important operational information still arrives in document form.

Document intelligence uses technology to identify meaning and structure within documents.

It goes beyond recognizing individual characters on a page.

In a family office, document intelligence can help interpret financial statements and private-market notices before the information enters a workflow.

Exception handling is the process used when a workflow cannot continue through its normal path.

An exception may occur because information is missing. It may also occur because information fails a required check.

A controlled workflow should send the exception to the appropriate reviewer and preserve a record of what happened.

A family office is an organization established to manage the financial and broader affairs of a wealthy family.

The scope of a family office varies significantly.

Some family offices focus mainly on investments. Others have a broader remit.

The operating model normally reflects the specific needs of the family it serves.

Family office data aggregation is the process of bringing financial information from multiple sources into a consolidated view.

The information may come from different institutions. It may also come from different internal systems.

Aggregation alone does not guarantee reliable reporting. The resulting data may still need to be reconciled and reviewed.

The terms wealth aggregation and portfolio aggregation are also used for related processes.

Family office governance is the framework that defines how authority and accountability are organized within a family office.

Governance establishes how decisions should be made.

It also determines how those decisions are recorded and followed through.

Technology can support governance by making recurring processes more visible and traceable.

A family office operating model describes how a family office organizes itself to deliver its responsibilities.

It connects the family's objectives to the way the office actually functions.

The operating model determines how work is structured. It also influences how technology is used.

A family office operating model is an organizational concept. It should not be confused with an AI operating layer.

Family office operations are the recurring processes required to keep a family office functioning effectively.

The exact scope depends on the office.

Strong operations help information move through the organization reliably. They also reduce dependence on informal manual coordination.

Family office reporting is the process of producing consolidated information for family members and other authorized stakeholders.

Reporting can extend beyond investment performance.

The quality of a report depends on the quality of the process used to collect and verify the underlying information.

Automation can improve the reporting process before the final report is produced.

Generative AI is artificial intelligence designed to create new content in response to an instruction or input.

Large language models are commonly used to power generative AI applications.

In a family office, generative AI can support interpretation and drafting. Important outputs should still be verified before they are relied upon.

Human-in-the-loop, or HITL, is an operating pattern in which a person reviews or intervenes at a defined point in an automated process.

Human review is particularly useful where an AI output could affect a significant decision or action.

HITL strengthens control. It does not guarantee that every error will be identified.

Inbox extraction is the process of identifying useful information within incoming emails and attachments so that it can enter a structured workflow.

This is particularly relevant to family offices because important operational information still arrives through email.

Inbox extraction can provide an automation path where direct system connectivity is unavailable or unnecessary.

Intelligent document processing, or IDP, uses AI and related technologies to convert documents into structured information that can enter a business workflow.

IDP goes beyond basic text recognition.

In a family office, IDP can help process recurring documents while directing uncertain information to human review.

Key-person risk is the risk created when an organization depends heavily on one individual.

In a family office, the dependency may relate to operational knowledge. It may also relate to how a recurring process is completed.

Better workflow design can reduce this dependency without removing human accountability.

A large language model, or LLM, is an AI model trained to understand and generate language.

LLMs power many modern AI applications.

They can interpret complex text. They should not automatically be treated as authoritative sources of truth.

The middle office supports investment-related operations between investment decision-making and administrative processing.

In a family office, the middle office often supports reporting and investment oversight.

The exact scope varies from one family office to another.

A multi-family office, or MFO, provides family office services to more than one family.

Its operating model differs from that of a single-family office because infrastructure is shared across multiple client relationships.

Appropriate separation of information and responsibilities is therefore particularly important.

Open-weight AI refers to AI models whose trained model weights are made available for others to use under specified terms.

Open-weight models can often be deployed on infrastructure controlled by the organization using them.

For a family office, this can provide more flexibility over deployment and data handling.

Open-weight does not automatically mean open source. It also does not automatically make a model private or secure.

An operational workflow is the defined sequence through which a recurring piece of work is completed.

A workflow begins with a trigger. It ends when the required outcome has been reached.

In a family office, operational workflows can provide consistency across processes that would otherwise depend heavily on manual coordination.

Optical character recognition, or OCR, converts text contained in an image or scanned document into machine-readable text.

OCR identifies characters.

It does not necessarily understand the business meaning of the information it reads.

Modern document workflows therefore often combine OCR with more advanced document intelligence.

A portfolio management system, or PMS, is software used to manage investment portfolio information.

In a family office, a PMS can provide an important source of investment data.

It can remain a system of record while workflow automation coordinates work around it.

Private AI refers to an approach in which an organization retains greater control over how its data is used within an AI environment.

Private AI does not necessarily mean that an AI model runs on premises.

The actual level of protection depends on the architecture and provider arrangements.

For family offices, private AI is particularly relevant because confidentiality is often a central requirement.

A private investment office is a privately controlled organization focused primarily on managing investments for an individual or family.

The term does not have one universally accepted definition.

Some private investment offices resemble investment-focused single-family offices. Others operate with a narrower mandate.

Prompt injection is an attack that attempts to manipulate how an AI system behaves by introducing malicious instructions.

The instruction may come directly from a user.

It can also be hidden inside information that the AI system has been asked to process.

This is particularly relevant to family offices because external documents may enter AI-enabled workflows.

Role-based access control, or RBAC, assigns system permissions according to defined roles.

A user receives access because of the responsibilities associated with that role.

RBAC can also help control which information an AI agent is allowed to access.

A single-family office, or SFO, is an organization dedicated to managing the affairs of one family.

Its structure is normally tailored to that family's needs.

An SFO may manage activities internally. It may also coordinate work with external providers.

Structured data is information organized according to a predefined format.

A spreadsheet with consistent fields is one example.

Structured data is generally easier for software to process because the structure of the information is known in advance.

A system of record, or SoR, is the authoritative system for a defined category of information.

An organization can have more than one system of record.

In a family office, an AI operating layer can coordinate work around these systems without replacing them.

An unfunded commitment is the portion of committed capital that an investor has not yet been required to contribute.

It represents a future funding obligation.

For family offices investing in private markets, tracking unfunded commitments supports liquidity planning and reporting.

Unstructured data is information that does not follow a predefined database-style format.

An email is one example.

A PDF can also contain unstructured information.

AI can help interpret unstructured information before it enters a structured workflow.

A workflow is the defined sequence through which work moves from a starting point to completion.

A workflow can involve software. It can also involve people.

In a family office, clearly defined workflows make recurring operational work easier to manage and improve.

Workflow automation uses technology to execute or coordinate parts of a defined workflow automatically.

The workflow normally follows predetermined logic.

AI can be added where interpretation is required.

Other workflow steps can remain rule-based.

Workflow orchestration coordinates multiple tasks across an end-to-end process.

It focuses on how separate steps work together rather than on automating one isolated action.

For a family office, workflow orchestration can coordinate work across existing systems while preserving defined review points and controls.