---
title: "Connected Intelligence Advisors — FAQ"
sidebar_label: "Advisors FAQ"
description: "What Sensei Connected Intelligence Advisors are, how the Sensei-hosted AI service handles your data, and exactly what is sent when an advisor runs."
last_updated: 2026-06-22
---

import BrowserOnly from '@docusaurus/BrowserOnly';

# Connected Intelligence Advisors — FAQ

:::caution
This article describes **pre-release functionality and features**. Delivery details and the in-product connection steps are subject to change before general availability.
:::

Connected Intelligence is how Sensei IQ brings AI to the moment of decision — inside the blocks and forms you already work in, grounded in your own project data. This page answers the questions teams ask before turning it on: what it is, what it does with your data, and exactly what gets sent when an advisor runs.

If you only read one thing: **a Sensei Advisor is opt-in, runs only when you ask, and sends only a defined, named set of fields — never your whole project, your documents, or anything Anthropic is allowed to train on.**

---

## The short version

- **It's an advisor, not an agent.** The AI offers a recommendation you can accept, edit, or reject. It never acts on its own — you make the call.
- **It's opt-in and off by default.** Each advisor is enabled per environment by your administrator. Until then, nothing runs and nothing is sent.
- **It only runs when you ask.** You click an **AI** button. There is no background processing of your data.
- **Only scoped data is sent.** Each advisor sends a specific, named list of fields — defined in a configuration setting you control — not free-form access to your tenant.
- **Your data is not used to train AI.** Anthropic is contractually prohibited from training on your content, and API data is deleted within 30 days by default.
- **You stay model-agnostic, anchored on Claude.** Sensei IQ is designed to work with any leading AI model; today the hosted service runs on Anthropic's Claude.

---

## About the advisors

### What is a Sensei Advisor?

A Sensei Advisor is an AI-powered assist that appears in context — inside the block or form where the relevant decision happens. Instead of a chat window where you have to know what to ask and how to phrase it, the advisor produces a structured, in-context suggestion at the click of a button.

The naming is deliberate. An **agent** implies autonomy — acting, deciding, running unsupervised. A Sensei **Advisor** does the opposite: it informs *your* decision. The AI has a point of view; the human makes the call.

### What kinds of advisors are there?

Sensei Advisors come in three families, distinguished by where they live and what they help you do:

| Family | Where it appears | What it does | Example |
|---|---|---|---|
| **Block Advisor** | Its own block on your dashboard | Reads across your work and tells you what to look at first, then points you to the right place. It orients and routes — it never changes records itself. | **AI Insights** — "what should I focus on today?" |
| **Instrument Advisor** | Inside a block you already use | Helps you complete the task in front of you — drafting a field, suggesting language — using only that block's data. | **Status advisor** — drafts a status narrative on the Status Update block |
| **Generative Advisor** | On a "create new" form | Drafts a starting point for a new record, then steps out of the way the moment you start typing. | **Deliverables advisor** — proposes deliverables from your schedule |

Every advisor follows the same delivery pattern and the same data-handling rules described below. Adding a new advisor means a new instruction template and a new button — not a new way your data is handled.

### Do I have to use it? Is it on by default?

No, and no. Connected Intelligence is an **optional service that is off by default**. Your administrator enables each advisor for your environment; the on/off setting also controls whether the advisor's button is even visible. If you never enable it, Sensei IQ behaves exactly as it does today and nothing leaves your tenant.

### How is it delivered? Is there anything for us to deploy?

For the current offering — the **Sensei-managed (Turnkey)** option — there is nothing for you to deploy. Sensei hosts and operates the service, manages the AI model and the API key, and maintains the curated instructions. You turn the advisor on and use it. (Other delivery options are on the roadmap — see [Delivery options and roadmap](#delivery-options-and-roadmap).)

---

## How it works — the round trip

Every Sensei-hosted advisor follows the same six-step round trip. Your request leaves your environment only when you click, carries only the data the advisor needs, and comes straight back to where you were working.

1. **You call for the advisor.** You click the **AI** button on a block — for example, the Status Update block — to ask for help. Nothing runs until you do.
2. **Sensei IQ assembles the request.** It combines Sensei's curated instructions with a *defined, scoped* set of fields for that advisor, read from a configuration setting.
3. **The request goes to Sensei's managed Azure Function** over an encrypted (TLS) connection. Sensei holds the AI key and the curated prompts; the function relays the request and does not store your data.
4. **The function calls Anthropic's Claude API**, which generates the response.
5. **The response returns** through the Sensei function back into Sensei IQ.
6. **You review, edit, or accept.** The suggestion lands where you were working. Editing clears the AI state — the AI proposes, you decide.

**Click any node** in the diagram below to see what happens — and what does and doesn't cross each boundary — at that step. The example shown is the Status advisor; every advisor takes the same path.

<BrowserOnly fallback={<div style={{padding: '40px', textAlign: 'center', opacity: 0.7, border: '1px solid var(--ifm-color-emphasis-300)', borderRadius: '16px'}}>Loading the Connected Intelligence round-trip diagram…</div>}>
  {() => {
    const ConnectedIntelligenceFlow = require('@site/src/components/ConnectedIntelligenceFlow').default;
    return <ConnectedIntelligenceFlow />;
  }}
</BrowserOnly>

---

## Your data

### What data is sent to the AI?

Only a **defined, named set of fields** for the advisor you invoked — nothing more. Each advisor's configuration setting lists exactly which fields are included. The advisor does not get free-form access to your tenant: it does not read your whole project, your other projects, your documents, or any field that isn't on the list.

As a concrete example, when the Status advisor drafts the **Risks** portion of a status narrative, the configuration sends only these named columns for the risks in scope:

```text
sensei_name, statuscode, statecode, ownerid, createdon, modifiedon,
sensei_id, sensei_assignedto, sensei_benefit, sensei_category,
sensei_consequence, sensei_consequenceindex, sensei_contingencyplan,
sensei_cost, sensei_description, sensei_duedate, sensei_likelihood,
sensei_likelihoodindex, sensei_mitigationplan, sensei_portfolio,
sensei_program, sensei_project, sensei_riskowner, sensei_riskscore,
sensei_trigger, sensei_triggerdescription
```

If a field isn't on the advisor's list, it isn't sent. That is what "scoping to specific pieces of data" means in practice.

### Can we control or see exactly what's sent?

Yes. The instructions and the field list are held in a configuration setting, so what the advisor sends is explicit and reviewable rather than hidden in code. Per advisor, the configuration covers:

- **On/off** — whether the advisor is enabled (and its button visible).
- **The endpoint** — the Sensei-hosted service the advisor calls.
- **The access key** — the credential used to reach that service.
- **The instructions and field scope** — the curated prompt and the exact list of fields each part of the request may include.

Your administrator sets these, and the field scope makes the data boundary auditable.

### Does our data leave our tenant?

For the base Sensei IQ product, no — your data stays in your Microsoft 365 tenant, as it always has. Connected Intelligence is the deliberate exception you opt into: when you run an advisor, the scoped request described above leaves your environment over an encrypted connection, is processed by Sensei's managed Azure Function, and is sent to Anthropic's Claude API to generate the response. It is processed to produce your result and returned — it is not stored to build a database of your data.

### Is our data used to train the AI?

No. Under Anthropic's commercial terms, **"Anthropic may not train models on Customer Content from Services,"** and by default Anthropic does not use commercial API inputs or outputs to train its models. You retain rights to your inputs and own the outputs.

### How long is it kept?

For the Anthropic Claude API, **inputs and outputs are automatically deleted within 30 days** by default. Anthropic also offers Zero Data Retention agreements for stricter requirements. Sensei's function processes your request in transit and does not retain it as a record of your project data.

### Is it encrypted?

Yes. The request is encrypted in transit (TLS 1.2 or higher) for the whole round trip — from Sensei IQ to the Sensei-managed function, and on to the Anthropic API.

### Who can see our data?

Sensei does not retain or build datasets from your project data, and Sensei staff do not access it in the course of running the service. Anthropic processes the request transiently to generate the response, under the commercial terms above. Nothing about Connected Intelligence changes the access model for the rest of your Sensei IQ data, which continues to live in your tenant under your control.

---

## For your security and procurement team

This section consolidates the details a vendor security review will ask for. It supplements — and is consistent with — the [Security Review FAQ](../../security/security-review-faq.md), [Data Protection](../../security/data-protection.md), and [AI Policy](../../policy/ai-policy.md) pages, which describe the base product where **no data leaves your tenant**. Connected Intelligence is the single, opt-in exception to that posture, and applies only when you enable it.

**Processing model.** When enabled and invoked, a scoped payload (above) is transmitted from your environment to a Sensei-operated Azure Function, which calls the Anthropic Claude API using a Sensei-managed key. In this flow, Sensei acts as a processor and Anthropic as a subprocessor, for the scoped data only. When the service is not enabled, there is no such flow and no subprocessor.

**Subprocessors.** The subprocessors below are used exclusively for the optional Connected Intelligence service and process data only when it is enabled and invoked. Because the service is opt-in (off by default), there is no opt-out deadline — if you never enable it, none of these subprocessors process your data.

| Subprocessor | Purpose | Location |
|---|---|---|
| Anthropic, PBC | AI model provider — generates advisor responses via the Claude API | United States |
| Microsoft Azure (Sensei-managed subscription) | Hosts the Connected Intelligence Function App that assembles and relays the scoped request | United States (Azure West US) |

**Data residency.** Sensei's Azure Function operates in the **Azure West US** region. If your Microsoft 365 environment is provisioned outside the United States, note that the scoped request is processed in the US for the duration of the call. Data residency for the rest of your Sensei IQ data is unchanged — it remains in the region you selected for your own environment.

**Key and credential management.** The AI API key is held by Sensei in Azure Key Vault and accessed via managed identity; it is not embedded in client configuration or application code. The endpoint requires an access key to call.

**Retention and training (Anthropic).** No training on customer content (contractual and default); API inputs and outputs deleted within 30 days by default; Zero Data Retention available. See Anthropic's [commercial terms](https://www.anthropic.com/legal/commercial-terms) and [data retention policy](https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-data).

**Anthropic certifications.** Anthropic maintains **SOC 2 Type II, ISO/IEC 27001:2022, and ISO/IEC 42001:2023** (AI management systems), with HIPAA support available. See [Anthropic's certifications](https://privacy.claude.com/en/articles/10015870-what-certifications-has-anthropic-obtained).

**Encryption in transit.** TLS 1.2 or higher across the full round trip, including the outbound call to Anthropic.

**Turning it off.** Disabling the advisor's configuration setting removes its button and stops all calls immediately. No data is sent while an advisor is disabled.

**Scope auditability.** Because the field list lives in configuration rather than code, your administrators can review exactly which fields any advisor is permitted to send.

---

## Delivery options and roadmap

The Connected Intelligence architecture is designed so the in-product experience is identical regardless of how the AI is delivered — only the endpoint the advisor calls changes. There are three delivery options across the roadmap:

| Option | What it is | Availability |
|---|---|---|
| **1 — Sensei-managed (Turnkey)** | Sensei hosts the function and manages the key, model, and prompts. Nothing for you to deploy. | The current offering (pre-release) |
| **2 — Self-hosted (Bring Your Own Model)** | You run the function in your own Azure tenant with your own AI key, so the data and credentials stay in your infrastructure. Sensei provides the code and templates. | Targeted for early 2027 |
| **3 — In-tenant (Microsoft Copilot)** | A fully in-tenant path via Microsoft Copilot, with no separate function required, once Microsoft Copilot Studio reaches general availability. | Future |

The hosted (Turnkey) option is where the pattern is proven first; the self-hosted and in-tenant options extend the same architecture for organizations with stricter data-sovereignty requirements.

---

## Related pages

- [Data Protection](../../security/data-protection.md) — encryption, residency, and how your tenant data is protected
- [Security Review FAQ](../../security/security-review-faq.md) — answers for vendor security and procurement reviews
- [AI Policy](../../policy/ai-policy.md) — how Sensei governs AI use, in development and in the product
- [Compliance](../../security/compliance.md) — frameworks and certifications
- [Data Residency](../platform/data-residency.md) — where your data lives

## Changelog
| Date | Change | Author |
|------|--------|--------|
| 2026-06-22 | Page created — Connected Intelligence Advisors FAQ (Sensei-hosted / Turnkey), with interactive round-trip diagram and data-handling detail | Claude Code |