MMOOZZGG.com

AI solution map

AI for data analytics: from manual reports to answers in a minute

7 min read2 sources

AI for data analytics answers questions about money, sales and stock in plain language, using figures from 1C, your CRM and your bank. McKinsey (November 2025) reports that 88% of organizations use AI, and 39% see an effect on operating profit (EBIT). Below are three levels of such analytics, the data they need, ways to check the numbers and the price of a ready solution.

On this page · 6 sections

Three levels of analytics with AI

The levels go from simple to complex, and each one builds on the last. For each, below is an example of a manager’s question and the answer they get.

1. Answers to questions about data. Question: “How much do we owe suppliers, and to whom the most?” Answer: the debt total this morning, the overdue part and the three largest suppliers, with links to the documents in 1C.

2. A daily report with deviations. Question: “What went wrong yesterday?” Answer: one store’s revenue is below its four-week average, some CRM deals have had no reply for more than a day, and a fast-selling product will run out in three days.

3. A forecast. Question: “Will we have enough cash for payroll at the end of the month?” Answer: based on planned payments and past receipts, a gap is expected in the last week. Payment from two large debtors will close it, if they pay on time.

Level one replaces a request to accounting that used to wait until evening. Level two removes the manual morning summary. Level three is useful when the data has at least a year of history and it is clear what to do after a signal.

Different people ask questions. The owner asks about money and debts, the sales director about deals by manager, the finance director about payments for the week. The agent answers all of them from the same data, so everyone sees the same numbers.

What data you need

Data analytics with AI is built on the systems where the company already keeps records: 1C for documents, stock and debts, the CRM for deals, the bank for money, and spreadsheets for plans and everything not yet in the systems.

Quality matters more than volume. The model will make mistakes if one customer has three records in the CRM, deals have no dates and payment statuses are updated once a week.

A forecast needs history without gaps: if the CRM was introduced six months ago, there is nothing yet to base a deal forecast on.

The agent connects to the systems through its own account with read-only rights. It does not need staff passwords, and you can revoke access at any time.

Before launch we check three things: reference lists have no duplicates, dates and statuses are filled in, and each number has one source system. What counts as data for AI is explained on the page on data for AI, and the check order is in the section on data readiness.

If your main analytics still lives in spreadsheets, start with the page “AI for Excel”: it has ready prompts and rules for spreadsheets with personal data.

How to check that AI calculated correctly

A manager makes decisions based on numbers, so an AI answer cannot be taken on trust. Three checks make the calculation verifiable.

1. Reconciliation with accounting. We take a closed month and compare revenue, debts and stock from the AI answer with the accounting report. While there are differences, the agent does not go live.

2. A query log. Each answer is stored with its question, its source and the database query used to calculate it. You can open any number and recalculate it by hand.

3. Read-only access. The agent sees 1C, the CRM and the bank, but it cannot change an entry, a deal or a payment. An error in an answer stays an error in the answer, not in the books.

We repeat the reconciliation after each change in the records: a new warehouse, a new bank account, a new sales funnel in the CRM. Otherwise the agent keeps calculating by the old scheme.

We also check the questions for which the agent has no data. The right answer then is “no data”, not a plausible number.

Forecasts and predictive analytics

A forecast answers the question “what will happen if nothing changes”: how much product will sell, when the money will run out, which customer is about to leave. It needs history for the model to learn from and a clear action after the signal.

A useful forecast ends in an action: order stock, move a payment, call a customer. If nobody changes anything after a signal, the forecast becomes one more report that nobody reads.

How models notice the first signs of a breakdown or of a customer leaving is explained on the page about predictive analytics. Forecasts of sales, purchasing and production load are covered separately, on the page about demand forecasting.

We do not advise starting with a forecast. First the agent must answer questions about the past correctly: a forecast on wrong data is wrong with more confidence than a report.

A ready solution: the manager dashboard

For the owner of a chain or a company with several locations, all three levels come in one product, the manager dashboard. It brings 1C, the CRM and the bank into one window, sends a report every morning, answers questions and warns about a cash gap or an overdue deal.

You can view the report for the whole company and for each location. If AI agents already work in sales, support or bookings, the dashboard shows their results in the same numbers as the money.

The dashboard is part of the “Business turnkey” plan: from $18,000, term 2–3 months. The contents of the report are fixed in the contract.

You can test the approach on a smaller budget: a pilot with one agent and a morning report from one source costs from $1,499, term 14 days. The pilot metric: the time to prepare a report and the number of manual summaries per week.

How such a report fits management accounting, and what stays with the accountant and the finance team, is on the page about management accounting with AI.

Data and security

Company finances and customers’ personal data in one system need rules before launch, not after. Under 152-FZ, Russia’s personal data law, data of Russian citizens is collected and stored in databases in Russia (Part 5, Article 18).

Since 1 March 2023, an operator must notify Roskomnadzor, the Russian data regulator, before a cross-border transfer of personal data (Part 3, Article 12). So a CRM export with customer contacts cannot simply be sent to a foreign chat service.

Public chat services do not fit financial questions for the same reason: exports from 1C and the CRM almost always contain customers’ names and phone numbers, and nobody will remove them by hand before each question.

In our projects, personal data is stored on servers in Russia, the agent gets read-only rights, and the query log shows who asked what. All the rules are on the page about data security.

Sources and fact-checking

Checked by the editors on 16 October 2026.

  1. IT Brief, 11 November 2025 (itbrief.co.uk)
  2. Brace Law Firm, checked 16 October 2026 (brace-lf.com)

Found a mistake? Tell us

Publications

Contact us