Administrative
Ask the way you speak
No fields to fill in, no need to learn where everything lives. Type the question and Polaris answers with what is in the documents it has classified.
// Product
Polaris AI, in detail.
Everything that does not fit on the home page, organized by what you want to know. Pick a topic: each one opens its own panel.
The product’s own screens, with sample data, translated for this page: the product interface is currently in Spanish.
Role
Administrative
Everything that comes in, in one place
Whether it arrives by email or as a photo: it is detected, classified, and filed with its date. And it goes out to Excel or PDF for your accountant in one click.
Administrative
It proposes. It does not execute.
Five proposals awaiting approval; each one states which document it acts on. Four are waiting for manual review, with Approve and Dismiss buttons. The fifth is waiting for an answer on the desktop notification and on WhatsApp: its buttons are grayed out and the card says no manual decision is needed yet. The WhatsApp answer has been demonstrated in a test environment.
Administrative
For your accountant, in one click
The quarter goes out in Excel, CSV, or PDF without searching for anything again, because it was already classified the moment it came in.
Administrative
A yes before moving on
When a new document comes in, Polaris notifies you on the desktop and waits. Approving or rejecting is one click. If no one answers in time, the notice reaches you on WhatsApp: that part has been demonstrated in a test environment.
Management
The business at a glance
The Management role dashboard: today’s orders, pending invoices, monthly spend, and messages processed, with a chart of spending by month and the month’s top sales. The screen carries the DEMO mark.
Polaris does not decide for you. It decides with you.
For the actions configured for approval, approval is a step in the process: Polaris leaves the work ready and waits for a person to say yes.
Polaris is installed on your infrastructure, connects to the tools you already use, and works on whatever comes in.
- 01
In
Business email, through your own Google Workspace, and photos of receipts. Documents arrive the way they already arrive today.
- 02
Polaris classifies it
The model, on your server, reads what each document is — invoice, delivery note, order — who it is from, and when. It files it with its date and detects duplicates.
- 03
A person approves it
Actions configured for approval are held: a desktop notification and, if no one answers in time, a WhatsApp message (demonstrated in a test environment). Afterwards, what has been classified can be queried in plain language and exported to Excel, CSV, or PDF.
How Polaris AI fits into your company
A compute unit installed on your servers, connected to the tools you already use. Everything that comes in is classified on-premise — documents are not sent to an external AI provider — and actions configured for approval wait for a person to say yes.
In
Business email
Photos of receipts
WhatsApp
Telegram
Sales reps in the field
A folder on your computer
On your server
1 Polaris classifies it The model, on your server. It reads what it is, who it is from, and when.
2 A person approves it For configured actions: a desktop notification and, if no one answers, a WhatsApp message (in testing).
Out
A single archive
For your accountant
Answers
Answers over WhatsApp
Your management software
A layer of intelligence over the work you already do.
Polaris AI is an agentic AI platform installed on your company’s own infrastructure, working over your documentation and your processes. The model runs on your server; the connections that go out are listed by name in the privacy panel.
Privacy by design.
Your documents are not sent to an external AI provider: the model runs on your server. No copies in someone else’s data centers to analyze them. Designed to help you comply with GDPR and for professions bound by confidentiality.
External connections: your Google Workspace and WhatsApp
Cost, on the call.
The cost model is explained on the call, with your case in front of us. We do not publish prices on the website.
No prices on the website
It knows your business.
It works over your company’s documents: every answer comes from what is in them, in your own vocabulary. No public models are trained on your data.
Answers grounded in your documents
Four labels, with no mixing, and a single definition for each one: it applies across the whole website. The hours are estimates based on a real case.
- Available
- Works end to end and can be demonstrated. It ran in a real environment during the pilot (July to September 2026).
- Demonstrated in a test environment
- Tested end to end in a demonstration environment. It is not an integration ready to install, nor a service in operation.
- In development
- Partly built or not yet connected: it does not work end to end yet.
- Planned
- Not yet implemented. It is on the roadmap; no date is promised.
Show
Available
- 72 ha year
Organizing documents
CompleteEmail is read automatically, and invoices and delivery notes are recognized with their supplier, number, and date. A photo of a receipt is classified on its own. Duplicates are detected.
- 24 ha year
Preparing the handoff to your accountant
CompleteEverything is classified and dated from the moment it comes in, so the quarter goes out filtered and exported to Excel, CSV, or PDF without searching for it again.
- 6 ha year
Closing the annual accounts
PartialDocuments go out the same way as the quarter. Revenue does not: no sales source is connected. Payroll is listed, but without amounts and without export.
- ≈ 50 ha year
The day-to-day trickle
Two of threeLooking up an expense and finding an invoice, yes: you ask the way you speak, from the desktop. Checking a due date, not yet.
Demonstrated in a test environment
-
Queries over WhatsApp
Asking over WhatsApp about the data that lives on the company’s server, with role-based access and limits: each role reaches only what concerns it.
-
Management over WhatsApp
For the Management role: queries about documents, reports, and emails; WhatsApp confirmations when the desktop notification is not approved in time; and automatic status reports.
-
Automatic status reports over WhatsApp
Status reports that arrive on their own over WhatsApp, without having to ask for them.
-
Appointments and reminders over WhatsApp
Built. They were not adopted in the first deployment.
In development
- ≈ 96 ha year
Sales and payments received
Downloading every payment received, posting it to the books, and categorizing it by service. It is the same work as organizing invoices, with a different entry point. It is built; today no sales source is connected.
-
Due dates
Detecting them in email and sending a warning before they pass, with an entry in the calendar. It is built and currently paused.
-
Business queries on the Management panel
Asking about the business in plain language from the Management panel, for the Management role only. The screen exists; the query feature is pending activation.
-
Catalog over WhatsApp
A customer asks about a service or a price and Polaris answers. Built and shelved: a customer-facing chatbot was not wanted.
-
Search across the entire archive
Asking in plain language about the entire archive, beyond the prepared documents and queries, is the improvement with the greatest effect on what already exists.
-
Orders over WhatsApp
An order comes in over WhatsApp and is recorded, without anyone typing it in.
Planned
-
Reflective Module
Polaris would watch how it is used and propose improvements: a purchase worth reviewing, a recurring question that deserves a template, two reports that could arrive together. It is a planned feature, not yet implemented.
-
Sales reps asking from the field
The information they need, on their phone, without calling the office. In a three-person business it is not needed; in a company with people in the field, it changes the whole calculation.
Up next.
Polaris is under active development, and this is what is coming. If any of this is exactly what you need, tell us in the first conversation: what you ask for weighs on the order.
Telegram, alongside WhatsApp
The conversation with Polaris over WhatsApp has been demonstrated in a test environment. Telegram would be the same work done a second time: same agent, same human approval, another door. For those who prefer not to mix business with their personal number.
A sales role
Today Polaris distinguishes between management and administration: each sees only what belongs to them. The next role is the sales rep: checking an order or a client file from the field, with no access to anything beyond their remit.
A folder that monitors itself
You drop documents into a folder on your computer and Polaris picks them up, classifies them, and files them without anyone pressing a button. The component exists and is paused on purpose: we are still cataloging how many different ways the same document can arrive before we let it loose on yours.
Your management software
A3, Sage, Odoo, or whichever you use. Today Polaris leaves the work ready for your accountant in Excel, CSV, or PDF; the next step is for it to flow directly into the system where you already keep it.
Planned
Reflective Module proposals
This is how the window where Polaris would propose improvements could look, each one with its priority, to accept or dismiss. The feature is not implemented: this image is a mock-up.
What runs on your server, what leaves it, and what a person decides. Named explicitly, with no absolute promises.
What runs on-premise
- The AI model runs on the company’s own server: your own server or a private virtual machine on your network.
- Document classification, the archive, and answers to queries all happen there. Documents are not sent to an external AI provider.
- The model does not depend on a provider or on an internet connection to work. The connections to email and to WhatsApp do need internet.
Which external connections there are
- Your own Google Workspace, to read business email. Using OAuth2 with the minimum permissions required. The company authorizes the connection.
- WhatsApp, through Meta’s Cloud API, for queries and notifications. Demonstrated in a test environment: it is not yet an integration ready to install.
- You can audit the outbound traffic and verify it yourself.
Which actions a person approves
- Human approval applies to the actions configured for it. For example, filing an invoice that arrives by email: it waits until someone approves or dismisses it.
- The notification reaches the desktop. If no one answers in time, it arrives over WhatsApp (demonstrated in a test environment). Pending actions are also shown in the Suggestions panel.
- Not every operation requires approval. In the demo we show you which ones are held.
Regulatory compliance
- Designed to help you comply with GDPR, Spain’s LOPDGDD, and the EU AI Act. Processing on your own server reduces the number of third parties involved.
- Designed for professions bound by confidentiality: fewer third parties involved in processing your documents. The duty of professional secrecy remains with the professional.
- For special categories of data, processing happens on your server; the legal obligations remain yours, and we help you meet them.
Copilot and ChatGPT are good tools. The difference lies in where your documents are processed, who the service depends on, and how you pay. We describe it without figures: third-party prices change and we do not publish them.
Privacy: where the model runs and who processes the data
| Polaris AI | Cloud tools (Copilot, ChatGPT) | |
|---|---|---|
| Where the model runs | On the company’s own server | On the provider’s servers |
| What leaves the company | Only the connections the company configures: its Google Workspace (OAuth2, minimum permissions) and WhatsApp through Meta’s Cloud API (demonstrated in a test environment) | The content sent to the service |
| Provider commitments | Not applicable: no AI provider processes the documents | The major providers state that they do not train on business data by default and offer data residency in the EU or by region |
| Where the difference lies | Processing happens on the company’s infrastructure | Processing happens on a third party’s infrastructure, under that third party’s contractual guarantees |
Polaris AI runs the model on your server. In the cloud, the provider processes your documents on its servers, under the guarantees it offers. Choosing one or the other is deciding who handles your information.
Dependence on the cloud
| Polaris AI | Cloud tools (Copilot, ChatGPT) | |
|---|---|---|
| To work | The model does not depend on the provider or on an internet connection. The connections to email and to WhatsApp do need internet | The service needs a connection to the provider |
| Terms and availability | The model is installed on the company’s server: it does not change because a provider changes | Set by the provider and subject to change: it publishes plans and options that evolve, and sometimes promotional offers priced differently from the usual rate |
With a cloud service, its availability, terms, and prices are decided by the provider. With Polaris, the model lives on your server.
Variable costs
| Polaris AI | Cloud tools (Copilot, ChatGPT) | |
|---|---|---|
| How it is purchased | The cost model is explained on the call | Per user and, depending on the product, by usage or by credits. Copilot Studio (agents) is billed by credits, with packs or pay-as-you-go, and requires an Azure subscription. ChatGPT Enterprise offers credit-based or usage-based options with custom pricing |
| What spending depends on | Explained on the call, with your case | On how many users and how much usage; it can vary from month to month and with changes in terms |
In the cloud, spending follows the number of users and the usage, and it can change with the provider’s terms. The Polaris cost model is explained on the call.
Sources for what is said about third parties
- OpenAI, ChatGPT Business and Enterprise pricing page, accessed October 16, 2026.
- Microsoft Spain, Microsoft 365 Copilot for business page, accessed October 15, 2026.
- Microsoft, Copilot Studio pricing page, accessed October 16, 2026.
Third-party products change. If anything here is no longer accurate, write to us and we will correct it.
If your question is not answered here, email us at contacto@polaris-ai.es or call us at +34 602 02 58 77.
We talk with you and look at how you work: which tasks repeat, what comes through email, WhatsApp, or Excel, which tools you have, and which processes are still manual. From that we identify what Polaris can automate, assist with, or speed up, estimate the impact using your data, and prioritize what would make sense to deploy first. It is carried out by a person, not generated automatically.
The goal is not to replace what you already use, but to sit on top of it: Polaris works as a layer of intelligence and automation over your tools where that makes sense. Today it leaves the work ready in Excel, CSV, or PDF, and writing directly into management software such as A3, Sage, or Odoo is on the roadmap. What can be integrated in your specific case is determined during the assessment: we do not promise that everything integrates with everything.
The AI model runs on your server: documents are not sent to an external AI provider. Two connections do go out, and the company configures them: your own Google Workspace, for email (using OAuth2 with minimum permissions), and WhatsApp, through Meta’s Cloud API, which has been demonstrated in a test environment. You can audit the outbound traffic and verify it yourself.
It proposes and leaves the work ready. Actions configured for approval — for example, filing an invoice that arrives by email — wait until a person approves them, on the desktop or over WhatsApp (the latter demonstrated in a test environment). Not every operation requires approval; in the demo we show you which ones are held.
Setup is carried out by an integration partner at your premises, and training is tailored to your team and your schedule. The specific timeline is set during the preliminary visit, once we know what we are working with.
It depends on the size of your company and the workflows you want to automate. We guide you through choosing and purchasing the right server and, depending on the case, we can supply it ourselves as a turnkey solution.
The model can be swapped: your hardware and your documents stay the same; only the engine changes. The cost model is explained on the call.
The difference lies in where documents are processed and how you pay. In the cloud, the provider handles your documents on its servers, under its guarantees, and spending depends on how many users and how much usage. Polaris runs the model on your server; the cost model is explained on the call. The full comparison, with its sources, is in the Compare panel.
Available: document classification, export, and plain-language queries about documents. WhatsApp has been demonstrated in a test environment. Everything else, in development or planned, is listed separately in the What is ready panel.