How to Keep AI Business Analysis Inside Your Company
Why AI analysis worries finance and IT teams
The quick way to get AI help with business numbers is to export a report and paste it into a chat tool. That file can include customer names, margins or salaries, and once it leaves the ERP none of the permissions set there apply to it any more.
Letting an AI query the database directly is not safer. A model that writes its own queries can be steered by text hidden in the data, can join tables it should not see, and can run expensive queries against the live system.
Safe AI analysis needs controls on four things: what data is read, under whose rights, which AI model receives it, and where the answer can go afterwards. Each one has to be enforced by the system, not left to good intentions.
Why this happens
- Exports carry data past ERP permissions.
- AI tools are given broad database access.
- Sensitive figures reach people who should not see them.
- Answers are forwarded outside the company.
- Nobody records which model received which data.
Biznsbook addresses this through data classes per person, answers only to company users, SiteAdmin-approved models and regions, consent and audit history in AI Analyst, reading your own company data with each figure linked to its source row.
None of this needs more data: the business records are already in the system. What is missing is a quick, trustworthy way to ask about them and see the rows behind each answer.
Step-by-step: Keep AI Business Analysis Inside Your Company
Written for owners, finance managers and team leads using Biznsbook AI Analyst on their own company data.
- Assign the AI Analyst roles. AI Analyst Manager turns analysis on and chooses the model. AI Analyst User asks questions and runs briefs. Company administrators can do everything.
- Choose an approved model. In AI Analyst → Settings, under Set up an approved AI model, choose a model and paste your company's API key. Only models and regions Biznsbook has approved are listed.
- Tick the kinds of data allowed. Under Turn analysis on, choose the model and tick the kinds of data the analyst may send to it. Tick the agreement box and click Save and turn on; the consent is recorded.
- Grant sensitive data rights sparingly. Margins, HR, Payroll and Personal data each need their own right. Give these only to people who need them, in Management → Users.
- Limit who receives scheduled answers. In a scheduled workflow, tick up to 10 recipients under Send the answer to. Anyone who may not see all the data the answer used is skipped.
- Review Runs and history. Runs shows every analysis with the kinds of data, model, region, tokens and cost. Settings → History shows each time analysis was turned on or off, by whom, with which model and data.
- Turn it off if needed. Click Turn analysis off in AI Analyst → Settings. Nothing more is sent to the AI model, and past answers stay in Runs.
Review results after the first full monthly cycle. Adjust roles, mappings, or approvals where the same exception repeats.
Setup and screen steps are in the AI Analyst guide inside Biznsbook. This page focuses on the business question; the AI Analyst page summarises the add-on.
Common mistakes to avoid
- Mistake 1: Exports carry data past ERP permissions. Repeating this each month usually shows up first in decisions made on numbers nobody checked.
- Mistake 2: AI tools are given broad database access. Repeating this each month usually shows up first in decisions made on numbers nobody checked.
- Mistake 3: Sensitive figures reach people who should not see them. Repeating this each month usually shows up first in decisions made on numbers nobody checked.
- Mistake 4: Answers are forwarded outside the company. Repeating this each month usually shows up first in decisions made on numbers nobody checked.
- Mistake 5: Nobody records which model received which data. Repeating this each month usually shows up first in decisions made on numbers nobody checked.
Each of these leaves managers acting on numbers nobody checked. Ask about the data where it lives, and read the source rows before acting on a figure.
Best practices that hold up as you scale
- Assign the AI Analyst roles — AI Analyst Manager turns analysis on and chooses the model.
- Choose an approved model — In AI Analyst → Settings, under Set up an approved AI model, choose a model and paste your company's API key.
- Tick the kinds of data allowed — Under Turn analysis on, choose the model and tick the kinds of data the analyst may send to it.
- Grant sensitive data rights sparingly — Margins, HR, Payroll and Personal data each need their own right.
- Limit who receives scheduled answers — In a scheduled workflow, tick up to 10 recipients under Send the answer to.
Ask specific questions, read The data it used before acting, and save the questions you repeat as briefs.
How Biznsbook supports this workflow
data classes per person is part of Biznsbook AI Analyst and reads the same business records the rest of Biznsbook uses, under the asking person's rights, with every figure checked against the data before the answer is shown.
answers only to company users is part of Biznsbook AI Analyst and reads the same business records the rest of Biznsbook uses, under the asking person's rights, with every figure checked against the data before the answer is shown.
SiteAdmin-approved models and regions is part of Biznsbook AI Analyst and reads the same business records the rest of Biznsbook uses, under the asking person's rights, with every figure checked against the data before the answer is shown.
consent and audit history is part of Biznsbook AI Analyst and reads the same business records the rest of Biznsbook uses, under the asking person's rights, with every figure checked against the data before the answer is shown.
AI Analyst is an add-on ($10 per company per month) and needs no other module; each analysis area needs its module, scheduled analysis also needs Business Automation, and your company pays its own AI provider for tokens.
Suggested implementation timeline
- Day 1: Add an approved AI model and API key in AI Analyst → Settings, tick the kinds of data allowed and turn analysis on.
- Week 1: Give the AI Analyst User role to two or three people and try data classes per person on real questions.
- Week 2: Open The data it used on each answer and check the figures you plan to act on.
- Month 1: Save the questions people repeat as briefs, and schedule the weekly ones if you use Business Automation.
- Ongoing: A quarterly access review.
The data classes, and who should hold them
Every column the analyst can read belongs to a data class: Operations (quantities, counts and statuses), Customers (customer names and balances), Finance (ledger balances, cash, receivables and payables), Margins (cost price and gross margin), HR, Payroll and Personal data (names, phones and emails of people).
Operations, Customers and Finance follow the module rights a person already has. Margins, HR, Payroll and Personal data are granted on purpose and are not added to existing roles, so an owner or finance manager may hold Margins while a sales user does not.
Two filters apply to every analysis: the classes the person holds and the classes the company allowed when it turned analysis on. If nothing is left, the analyst reports that there is no kind of data both this person may see and the company allows, and the fix is to give the right or tick the kind of data.
Metrics to track monthly
- Questions asked per week and by whom
- Analyses refused or failed, and why
- Recommendations acted on
- Tokens and cost per month in Runs
Start with three metrics; trend direction matters more than a single point-in-time snapshot.
Manual analysis vs Biznsbook AI Analyst
Compare common manual analysis with Biznsbook data classes per person and the related AI Analyst capabilities.
| Capability | Manual / Spreadsheet | Biznsbook |
|---|---|---|
| What the AI reads | ❌ Whatever is in the export | ✅ Only the asking person's permitted data |
| How data is fetched | ❌ Copied by hand or by AI-written queries | ✅ Server data packs and reviewed queries; the AI never writes SQL |
| Sensitive figures | ❌ Visible to anyone with the file | ✅ Margins, HR, Payroll and Personal data need their own right |
| Where answers go | ❌ Anywhere they are forwarded | ✅ Only company users who may see the same data |
| AI provider | ❌ Any tool someone signs up for | ✅ Approved models and regions only, with recorded consent |
| Record of use | ❌ None | ✅ Runs and settings history |
A quarterly access review
Each quarter, list who holds the AI Analyst roles and each sensitive data right, and remove any that are no longer needed. Read Settings → History for changes to the model or the kinds of data, and check Runs for refused analyses that point to a missing or unexpected right.
Write down which questions each role asks every week and who acts on the answer.
Frequently asked questions
Does Biznsbook see or sell our data?
No. The data goes from your company to the AI model you chose, on your own account. Biznsbook keeps only the answer and the tables in your company's own records.
Does the AI write SQL?
No. Data comes from server-side data packs and from SiteAdmin-reviewed, isolation-tested queries that run on a read-only login with row-level company isolation.
Can someone see data through the analyst that they cannot see in Biznsbook?
No. The analyst uses the asking person's rights, and an answer is shown only to people who may see all the data it used.
Can answers be sent outside the company?
No. Answers can be emailed only to your own users who may see the same data, never to outside addresses, by SMS, WhatsApp or to other systems.
Why is a model missing from the list?
Only models Biznsbook has approved for analysis can be used, so data only goes to providers and regions that have been checked.
What does Agree again for the new model mean?
The model, its approval or its data terms changed. Read the model's terms in Settings and agree again before analysis continues.
How this differs by industry
Retail
Store managers analyse sales and stock without ever seeing cost prices, while the owner holds the Margins right.
Wholesale & distribution
Sales staff ask about their customers' balances, while payroll figures stay with the people who hold the Payroll right.
Manufacturing
Plant managers see operational figures, and only the finance team's answers include margins on finished goods.