AI IN BMS · VERIFIED WORKFLOWS
AI that uses approved tools and answers from verified results
AI in BMS interprets intent, selects a tool that fits the permission and context, then explains results from real services. The model does not decide prices, stock, money or permissions.
Content reviewed
The path to a grounded answer
- 01Message
- 02Intent
- 03Tool
- 04Permission
- 05Service
- 06Verified data
- 07Explanation
- 08Audit
Translate natural language into known work
AI helps identify intent and context without treating model inference as a business fact.
- Understand varied questions and recent context
- Select only capabilities the system exposes
- Ask for clarification when a request is incomplete
Call approved tools instead of the database
AI does not generate SQL or bypass the service layer. Tools validate input, tenant, permission and audit requirements before real work.
- Customers receive only customer-safe tools
- Staff tools follow permissions that the server checks again
- Sensitive work can be proposed but requires confirmation
Explain the result without inventing the result
After a service returns data, AI summarizes it in clear language and states the limits when evidence is incomplete.
- Do not guess price, stock, status or payment methods
- Do not turn unknown information into a confirmation
- Clinical decisions remain with a licensed pharmacist
Boundaries that keep AI inside its authority
Tenant scoped
Every read and write uses the shop established by the server, never a tenant claimed in user text.
Tool catalog
AI can use only tools already wrapped with validation, permission and audit controls.
Propose before mutate
Sensitive changes show the proposed mutation and server-composed arguments before confirmation.
Known limits
When data is missing, unavailable or unsupported, the system states the limitation instead of creating an answer.
Questions about AI in BMS
How is AI in BMS different from a general chatbot?
AI in BMS is connected to system workflows and approved tools, so it can explain verified data and prepare work for backend services rather than relying on text alone.
How does BMS reduce AI hallucinations?
Business facts must come from real tools and services. If evidence is missing or incomplete, the answer must state the limitation instead of inventing a price, stock level, status or confirmation.
Does AI query the database directly?
No. AI does not generate SQL and can use only backend tools that apply validation, tenant scope, permissions and auditing.
Do customer AI and staff AI have the same authority?
No. Customers receive only customer-safe tools. Staff tools depend on staff permissions, which the server checks again before execution.
Which actions require a person to confirm?
Sensitive changes and consequential transactions must show what will change and require explicit confirmation from an authorized person before execution.
Further reading and verification
BMS prepares this explanation from its product boundaries and operating guides. Examples illustrate workflows; integration availability and access depend on shop configuration and permissions. Contact the team to report information that needs correction.
See the behavior before choosing a workflow
Try a sample conversation or read the guide to AI reporting and operating boundaries.