AI at Promote:Global
Last updated September 15, 2026
Know what the AI knows. Know what it does. Stay in control. Promote:Global uses AI to help you understand markets, prepare projects and work with your own records. It does not decide anything for you. It does not approve transactions. It does not replace an official source. This page describes what the product does today. The limits below are built into the system, not just written here.
AI Nutrition Facts
The short version, in one panel. Every row is something the product does, or refuses to do.
Can make mistakes: Yes is the row that matters most. Every AI assistant gets things wrong sometimes. This one does too.
Where we use AI
Meridian is one assistant you can reach from several places. It answers, drafts and suggests. It never acts on its own.
| Feature | What it does | Can act alone |
|---|---|---|
| Meridian | Answers questions about markets, products, projects and expansion | No |
| Project guidance | Suggests next steps and points at information a project is missing | No |
| Research | Reads trade, tariff and market information and explains what it may mean for your project | No |
| Drafting | Prepares messages, summaries and working notes for you to edit | No |
| Discovery | Finds relevant companies and explains why they may fit | No |
| Your own documents | Answers from files you have added, and says which one it used | No |
AI proposes. You decide.
Meridian may suggest a next step, draft some text or recommend an option. You decide whether to accept it, edit it or ignore it.
You ask. Nothing happens until you do. Opening a screen never spends an AI action.
Meridian answers, and shows its sources where we have them.
Nothing joins your project until you save it. A suggestion you ignore changes nothing.
What AI can see
What Meridian uses depends on where you ask. It may use your company profile, your products, your markets, your projects, files you have added, information shared with you, public trade data, and sources it looks up for that question.
It does not see everything. Meridian sees what you can see. It follows your permissions and never widens them. This matters most when an advisor works with a client, or an agency works with a company on its roster.
The apps never hold a key to an AI provider and never contact one directly. Every AI request goes through our servers.
What Meridian can use by account type
Promote:Global has six kinds of account, and Meridian answers each one as itself. In every case it works only from what that account can already see.
| If you are a | Meridian works from |
|---|---|
| Company | Your profile, products, markets, projects and the files you have added |
| Service provider | Your own listing, your service context and the requests sent to you |
| Advisory practice | A client project you have been given access to, and nothing outside it |
| Trade or investment agency | Your roster’s consented profiles and your own queues |
| Chamber of commerce | Your members and your missions |
| Bank or financier | What your institution can already see on its own screens |
Which of these you are is settled by your account, not by what you type. You cannot ask Meridian to treat you as something else, and it will not assume. If your role is unclear, it answers with less rather than more.
What AI will not do
Some decisions stay with people. Promote:Global’s AI does not:
- approve or decline financing
- make credit, KYC, anti-money-laundering or sanctions decisions
- make a binding customs or tariff determination
- decide legal or tax compliance
- commit you to an introduction, or send one for you
- publish a change to your project without your approval
- override your sharing permissions
- speak for a government agency
On the financier console this is not just policy. The server sends this sentence with the answer: “Meridian does not make credit, KYC, AML, sanctions or guarantee decisions.”
Where answers come from
An answer separates what Promote:Global holds from what Meridian worked out. The last one matters most:
- Promote:Global data — already held in the platform
- Your own records — what you or your organization entered
- Your documents — files you added
- External sources — retrieved for that question, with the date they were read
- AI inference — an interpretation by Meridian, not a fact we hold
An inference is never shown as something Promote:Global knows.
AI can make mistakes
AI can misread a question. It can work from incomplete information. It can rely on a source that has since changed. It can assume something you did not say. It can state something wrong clearly enough to sound right. It can miss a change in law, tariffs or regulation.
Confirm anything that matters with the relevant authority or a qualified professional. Trade, customs, legal, tax and financing decisions all matter.
Your data and AI
| Question | Answer |
|---|---|
| Is my data used to train AI models? | No. Promote:Global does not use your data to train AI models, and we do not authorize our AI providers to use it for model training. There is no model-training pipeline in Promote:Global. |
| Does asking Meridian teach it anything? | No. A question you ask today does not change what Meridian knows tomorrow, for you or for anyone else. It reads your records at the moment you ask and nothing is learned from the asking. |
| What reaches the AI provider? | Your question, and the records the feature needs to answer it. Nothing is sent that the feature does not use. |
| Where does AI processing happen? | Model inference and the embeddings behind search run on European infrastructure. |
| How long is AI-related data kept? | AI call logs — the feature, the model and token counts, not your words — for 12 months. Your conversations are kept until your account is deleted. |
| What happens to it if I delete my account? | Your assistant conversations are deleted outright, not anonymized. So are the summaries written about your business and the record of what you were shown. |
| Can another member read my conversations? | No. Meridian conversations are private to your account. |
The privacy policy is the fuller account, including every subprocessor.
Providers
| Provider | Purpose | Receives | Training use |
|---|---|---|---|
| evroc | Model inference, and the embeddings behind search | Your question and the records the feature needs | None |
evroc is a Swedish company and our AI model provider. Processing runs on European infrastructure. Web search is handled separately and is listed with our other subprocessors in the privacy policy.
This list is kept current and updated when our AI infrastructure changes.
AI labels in Promote:Global
These are the exact words the product shows.
Shown on every Meridian surface, including one with no answer on it yet.
Shown when the answer was built from your own records rather than general knowledge.
Shown on the financier console, with the answer.
Where Meridian writes prose, you can report it. We receive the answer and the message that prompted it. We never receive the rest of your conversation.
Change log
First publication. Added the AI Nutrition Facts and set out where Meridian is used, what it can see, and what stays with people.