FREQUENTLY ASKED QUESTIONS
Frequently asked questions about Alicanto Analysts
How the agent that connects your data works: how it plugs into your internal databases, where the numbers come from, how it avoids inventing figures, and how to try the public sample.
This page describes how the product works. Any figure the agent returns should be validated with your team before making a decision.
Frequently asked questions
- What is Analysts and what is the public sample I can try free?
- The product is an agent that connects to your organization's internal databases so your team can ask them in plain language, without depending on an analyst having time to build the report. What you try free is a public sample of the agent over open Central Bank and CMF data: it lets you see how it answers before connecting your own.It starts in finance, where the public catalog gives it a clear edge, but connecting to internal data works for any industry with databases of its own to query.
- How does it connect to my internal databases?
- We build search tools tailored to your sources, which add to the agent's catalog. On top of that run transformation tools that are cross-cutting to any source: you just declare the shape of the data to the system, such as a time series or panel data, and the transformations become available.That way the agent keeps calling controlled functions instead of improvising the number, and the same traceability you see over public data applies to your internal data.
- How does Alicanto Analysts avoid inventing data?
- The language model does not produce the figures. It interprets your question and decides what to look up, but the numbers come from deterministic code that queries a closed catalog of real series. If the data is not in that catalog, the system says so instead of filling in a plausible number.The component that writes the answer is forbidden from typing numbers by hand: it leaves a reserved placeholder that the code fills with the already-computed value. This holds the same way when the catalog includes your internal data.
- Where does the data come from?
- In the institutional version, from your own internal databases, connected with tools built for your sources. The public sample works on official sources: indicators from the Central Bank of Chile, such as the CPI, the monetary policy rate, the IMACEC and GDP, and the accounts that banks report to the Financial Market Commission (CMF).The system does not go out to quote prices or scrape random sites. It works on a curated, dated catalog, whether yours or public.
- Does it show where each figure came from?
- Yes. Every answer carries its own traceability: which source the series came from, over what time window and what transformations were applied, written in plain language right next to the figure.You can open that traceability and export both the chart and the raw data to review on your own.
- What kind of questions does it answer?
- Natural-language questions about your data: how one of your own indicators evolved, a cross-reference of your operations against the system, market share or a change over time. In the public sample, about the Chilean financial system. It returns prose, charts and tables you can export.The chart type is chosen by a rule based on the shape of the data, not by the model.
- What does the agent not do?
- It does not give investment recommendations or answer about data that is not in its catalog. When it lacks context, it asks instead of assuming.It is a deliberate limitation: the system prefers to hold back rather than improvise a number that looks right.
- Can it make projections?
- Yes, in the institutional version. We build projections tailored to each client's requirements, grounded in official macroeconomic indicators and scenarios. They are not part of the public sample.Since each projection is modeled with the client, it is part of the institutional rollout, not a button you try for free.
- Is the AI model trained on my questions?
- No. Your conversations are not used to train models. In the public version the provider operates under an agreement that excludes training, and in the version for institutions, under contractual zero retention.The institution's internal data does not travel to the model: what is sent is the question and the structured answer.
- Is it tied to a single AI provider like OpenAI or Google?
- No. The engine is designed to be provider-neutral, so the model can be changed without rewriting it.That neutrality avoids getting locked into a single AI provider.
- Can I audit or reproduce an analysis?
- Yes. Every query is logged and the computation is deterministic, so the same analysis on the same data always gives the same result.The traceability of each step lets someone on your team review how the figure was reached.
- Which channels does it work on?
- Web, WhatsApp and Telegram.The same logic of traceability and abstention applies on all three.
- Do I need to connect my internal data to try it?
- No. You can start with the public data from the CMF and the Central Bank. There are 150 free credits, no card required, to try it.Connecting the institution's internal data is a later step, under contract.
Glossary
Product and financial-data terms that appear in these answers.
- Traceability
- The list of steps that accompanies each answer: where the series came from, over what period and what transformations were applied to the data.
- Deterministic tool-calling
- The agent invokes controlled functions over data that already exists, instead of generating the number itself. Each call is logged with its result.
- CMF
- Financial Market Commission, the regulator of banks, insurers and other entities in Chile. It publishes the accounting figures that banks report.
- Central Bank
- The Central Bank of Chile, which publishes macroeconomic indicators such as the CPI, the monetary policy rate, the IMACEC and exchange rates.
- IMACEC
- Monthly Index of Economic Activity. A monthly proxy for how GDP is evolving.
- Monetary policy rate
- The benchmark interest rate set by the Central Bank of Chile.