Knowledge base
Holds internal policies, product information and other company knowledge in one place.
Case Study
We built a permission-aware AI knowledge assistant that answers staff questions from company documents, while keeping every answer inside the limits of the user's role and department.
Built for an organisation whose policies, product information and internal knowledge needed to be easy to ask about, but not open to everyone.
Enterprise and business operations. Client and project names are kept confidential.
- Industry
- Enterprise and business operations
- Project type
- Internal AI knowledge hub and permission-aware assistant
- Used by
- Employee, Department user, Management, Admin
- Technology
- Kept private at the client's request
Overview
An AI knowledge assistant that respects who is asking.
From scattered documents to one secure assistant
Internal policiesProduct informationManagement files
Permission-aware assistant
The organisation had a lot of useful knowledge: internal policies, product information and the kind of company detail people ask colleagues about every day. Finding it meant knowing which folder to open or which person to message.
We built a permission-aware AI knowledge assistant inside the staff dashboard. People ask questions in plain language and get answers grounded in company information, but only from the knowledge their role and department allow them to see.
The problem
Company knowledge was hard to find and hard to share safely.
Staff needed quick answers about policies and products, yet the information sat across documents that few people knew how to search. Questions went to colleagues, who then had to stop their own work.
A general chatbot was not an option. Finance, HR and management information all carry different sensitivity, and an assistant that could see everything would risk showing the wrong answer to the wrong person.
Answers buried across many documents
Colleagues fielding the same questions
Sensitive information needing tight limits
Departments needing different knowledge
Risk of exposing management information
Time lost hunting for the right policy
What needed to change
What needed to change
The assistant had to be useful to everyone while showing each person only what they are allowed to know. Permissions had to shape the answer itself, not sit on top as an afterthought.
Questions we worked through
- How should a user's role and department decide which knowledge an answer can draw on?
- Where should the line sit between departmental knowledge and management-only information?
- Would source references help staff trust an answer and check the original policy?
- How could admins spot questions the knowledge base cannot yet answer?
- Would audit logs of questions and access be needed for sensitive areas?
The solution
A secure knowledge layer behind the AI assistant
We brought company documents into one central knowledge base and connected it to an internal AI chatbot on the user's dashboard. Every user has a role and a department, and those decide which knowledge the assistant can use when it answers.
Document ingestion
Central knowledge base
Role and department permissions
Internal AI chatbot
Role-appropriate answer
Document ingestion
Brings company documents into the knowledge base so the assistant can answer from them.
AI assistant
Takes natural-language questions and answers from company information.
Roles and departments
Gives each user a role and a department that define what they can access.
Access rules
Applies finance, HR and management-level restrictions to the knowledge an answer can use.
Dashboard integration
Places the assistant inside the dashboard staff already use.
How the workflow works
How the AI knowledge assistant works
For the user it feels like asking a well-informed colleague. Behind the scenes, permissions are applied before any knowledge reaches the answer.
- 01
Documents added
Company documents are ingested into the central knowledge base.
- 02
Access defined
Each user is given a role and a department with matching permissions.
- 03
Question asked
An employee opens the assistant from their dashboard and asks in plain language.
- 04
Permissions applied
The assistant limits itself to knowledge the user's role and department allow.
- 05
Answer returned
The user receives an answer grounded in company information they are entitled to see.
Who uses it
Who uses it
The same assistant serves everyone, but what it can tell each person depends on where they sit in the organisation.
Employee
Opens the internal assistant, asks a question and receives an answer appropriate to their role.
Department user
Gets answers from their own department's knowledge, such as finance or HR.
Management
Can access sensitive information that has been authorised for management.
Admin
Manages documents, permissions, users, departments and knowledge sources.
Permission-aware answers
What each department can ask about.
The assistant only draws on knowledge the person asking is allowed to see, based on their role and department.
| All staff | Finance | HR | Management | |
|---|---|---|---|---|
| Company policies and product information | ||||
| Finance knowledge | If authorised | |||
| HR knowledge | If authorised | |||
| Restricted management information |
Key features
Key features
Internal AI chatbot
An assistant built for staff, not the public, available from their dashboard.
Central company knowledge base
Policies, product information and company knowledge held in one place.
Document ingestion
Company documents are brought in so the assistant can answer from them.
Natural-language questions
Staff ask the way they would ask a colleague, with no search syntax to learn.
Grounded answers
Responses are based on company information rather than general knowledge.
Role-based permissions
Each user's role decides which knowledge an answer can draw on.
Department-level access
Finance users reach finance knowledge and HR users reach HR knowledge.
Management-level restrictions
Sensitive management information stays limited to those authorised.
Business value
Business value
The organisation gets the convenience of an AI assistant without loosening control over who can see what.
Before
- Knowledge spread across documents
- Questions sent to busy colleagues
- No safe way to open up sensitive areas
- The same access for everyone
- Slow answers on policy and product
After
- One central knowledge base
- Answers on demand from the dashboard
- Access shaped by role and department
- Management information kept restricted
- Answers grounded in company information
The outcome
The outcome
Staff can ask the assistant first, and the answers they get reflect both what the company knows and what they are allowed to see.
- Employees find policy and product answers without chasing colleagues.
- Finance and HR knowledge reaches the people who need it.
- Sensitive management information stays behind the right permissions.
- Admins manage documents, users and departments from one place.
- The assistant sits inside the dashboard staff already use.
Services in this project
The services behind this system.
More case studies
All case studies Other systems we have built.
AI Sales Companion
AI Sales Companion with Real-Time Sales Intelligence
An AI sales companion that finds relevant company projects during meetings, suggests talking points, then tracks follow-ups, client requests and internal dependencies afterwards.
View case study
AI Document Review
AI Loan Application Processing with Human Review
An AI workflow that parses loan documents, scores each application and routes it to human reviewers, helping a lender handle 100+ applications with reduced manual workload.
View case study