---
title: AI that does commercial work
seo_title: AI for CPG commercial teams | Intelligence and Autonomous | Advise
description: Intelligence answers your questions with verified numbers. Autonomous hands the work to AI agents. Governed, in-region, never trained on your data.
slug: /ai
template: standard
cta_primary: { label: Book an AI demo, href: "/demo?topic=ai" }
cta_secondary: { label: Join Autonomous early access, href: "/autonomous/early-access" }
schema: WebPage + FAQPage
related: [intelligence]
---

# AI that does commercial work

Your business has an AI budget. Advise is a place to spend it that pays back in the commercial team, every week.

The AI in Advise comes in two layers:

| | What the AI does | Available |
|---|---|---|
| **[Intelligence](/intelligence)** | Reads your dashboards and answers questions in plain English with verified numbers. Research writes structured briefs you can export to PowerPoint in your own template | Now |
| **[Autonomous](/autonomous)** | A team of AI agents that investigates overnight, prepares recommendations, emails and decks, and stops at your decision | Early 2027 · early access open |

## What makes it different

Most AI tools for data are a language model pointed at a database. They guess, and they sound sure when they're wrong. Advise was built the other way round.

- **The AI chooses, the engine computes.** Intelligence and Autonomous pick analyses from a governed Function Registry. Core computes every number. The AI never runs a query and never invents a number.
- **It knows CPG.** The analyses are built for category, revenue growth and account work: distribution gaps, rate of sale, promo ROI, price gaps, switching and more.
- **It shows its work.** Every figure is traceable to its analysis and source. In Autonomous, each point is labelled as data, business context or judgement, and every recommendation comes with the case against it.
- **A person decides.** Nothing is sent automatically.

## Governance {#governance}

- **Model:** Microsoft Azure OpenAI, served within your data region
- **Your data:** never used to train any model
- **Access:** the AI works inside Advise's role-based access. It sees only what the person using it may see.
- **Airlock:** the AI reaches data only through the Function Registry, never through raw queries
- **Audit:** every answer can be traced to the analyses it used
- **Where it runs:** In the cloud, in a dedicated environment in your chosen region

[Visit the Trust Centre](/trust)

## EU AI Act

We've assessed Advise and its AI layers against the EU AI Act (Regulation 2024/1689).

- **No prohibited practices.** No social scoring, manipulation, biometrics or emotion recognition of any kind.
- **Not high-risk.** Advise is commercial decision support for CPG and Retail, outside every Annex III high-risk use.
- **Transparent.** The AI in Advise identifies itself as AI, and AI-generated content is clearly marked as such.
- **Clear roles.** Advise is the provider of the AI system in Advise (Intelligence and Autonomous), and your organisation is the deployer. We support your deployer obligations with documentation, training and our [Responsible AI Policy](/trust#security-pack).
- **Optional.** AI can be switched off at deployment level. Core delivers its full value without it.

## Put your AI budget to work {#budget}

A simple case your finance team will recognise:

1. **What it replaces:** hours of manual reporting and deck-building every week across category, insights and account teams.
2. **What it adds:** answers in seconds, finished briefs with Research, and (with Autonomous) work that arrives finished.
3. **What it costs:** about the cost of one analyst a year for a typical mid-sized brand owner. [See pricing](/pricing)
4. **How you measure it:** hours returned, decisions made before buyer meetings, and promotions evaluated. We set the baseline with you in the adoption programme.

Intelligence and Autonomous can be bought as their own AI line item on top of Core.

## Questions

**Which model do you use?** A Microsoft Azure OpenAI model, served within your data region. We choose the model for accuracy, security and cost, and the governance around it stays the same.

**Can the AI see data I'm not allowed to see?** No. It works within your role-based access.

**Will our data train the model?** No.

**Is Autonomous available now?** It launches in early 2027. Early access is open. [Apply](/autonomous/early-access)