Ask your documents instead of searching them.
The internal AI assistant from encorp.ai answers employees' questions from the company's policies, procedures and documents and points to the source of every answer. It respects the access rights we set together and is built on Microsoft Copilot, ChatGPT, Claude, Gemini or as a custom solution.
Employee, sales department
How is a business trip abroad approved?
You fill in a travel request and your line manager approves it. After the trip you report your expenses with the receipts.
Sources
- Business travel policycurrent version
- Expense reporting procedurecurrent version
The answer uses only documents you have access to. Documents for management and HR stay closed to your role.
Each answer points to the document it came from, so it can be checked.
The answers exist, but nobody finds them quickly.
The answers exist, but they are scattered across dozens of folders, emails and document versions. People ask colleagues or lose time searching, and new employees are hit the hardest.
One assistant that knows where the answer is.
We build the assistant on your documents and on the tools your team already works with.
Choosing the sources
Together we choose which policies, procedures and documents go into the assistant.
Access rights
The assistant respects the access rights we set, and everyone sees only their own documents.
Answer with a source
Each answer comes with a reference to the document it was taken from, so it is easy to check.
Your tool
Microsoft Copilot in Microsoft 365, a project in ChatGPT, Claude or Gemini, or a custom solution.
Security and data rules
Before launch we run a security check and agree which data goes into the assistant.
Current sources
In AI Ops we keep the sources current and improve the assistant every month.
What changes when an employee has a question
The documents stay where they are, but an answer no longer takes a search.
| Moment | Today | With an AI assistant |
|---|---|---|
| Question about a procedure | Searching folders and emails | An answer with a reference to the document |
| Versions | It is unclear which version is current | The assistant works with sources we keep current |
| Colleagues | Experienced colleagues answer the same questions | The assistant takes the repeat questions |
| New employees | They do not know whom to ask | They ask the assistant from day one |
| Access | Documents are forwarded by email to reach the right people | Everyone gets answers from the documents they can access |
From the first call to an assistant in real work
The first measurable results usually come within 2–4 weeks of the start.
Sources and rights
We choose the knowledge sources, the access rights and the tool the assistant will run on.
- List of sources
- Access rights
- Metrics and a baseline
Setup and checks
We build the assistant, connect the sources and run a security check.
- Answers with references
- Security check
- Data rules
Test with real users
We roll the assistant out to a group of employees with real questions and measure it against the baseline.
- Real questions
- Comparison with the baseline
- A decision on whether to continue
Current sources
We keep the sources current and improve the assistant every month.
- Response within 24 hours
- Monthly improvements
- New sources when needed
Where an internal assistant helps the most
These are generalised scenarios, with no client names and no real data.
Policies and procedures
An assistant on internal rules and procedures for the whole office.
Service team
A technical assistant built on the product documentation.
Sales team
An assistant with information on products and terms.
Companies we work with
Econt
Implemented a custom AI agent for Econt, the leading logistics provider in Bulgaria, to automate customer support.
vivacom
Vivacom is one of the leading telecommunications companies in Bulgaria, offering a full range of services including mobile, fixed-line, internet, and digital television. As part of the United Group, Vivacom is recognized for its innovation, reliable network infrastructure, and customer-centric approach. The company serves millions of residential and business clients, delivering high-speed connectivity, advanced telecom solutions, and cutting-edge digital services across the country. With a strong focus on digital transformation, 5G technology, and smart business solutions, Vivacom continues to play a key role in shaping the future of Bulgaria’s telecom sector.
Forbes Bulgaria
Delivered AI-focused educational trainings for Forbes Bulgaria to empower business leaders with practical knowledge.
| For whom | Companies where answers are scattered across folders, emails and document versions |
|---|---|
| Sources | The company's policies, procedures and documents |
| Answers | With a reference to the source |
| Access | The assistant respects the access rights we set |
| Built on | Microsoft Copilot, ChatGPT, Claude, Gemini or a custom solution |
| First result | 2–4 weeks from the start |
| Price | On request |
| Support | AI Ops: current sources and improvements every month |
What is an internal AI assistant?
An internal AI assistant is a tool for your employees, not for your customers. It answers questions about internal rules, procedures, products and terms, using the company's own documents. The question can be about leave, a business trip, a pricing policy or the steps of a procedure. Instead of searching dozens of folders and emails or interrupting a colleague, people ask the assistant and get an answer with a reference to the document it came from.
The reference gives the employee a way to check the answer in the document itself before using it. The assistant does not replace the documents, it leads to them. The better organised and more current the sources, the more useful the answers, which is why choosing them is the first step. New employees, who do not yet know whom to ask or where to look, benefit the most. For them, see also the page on faster onboarding of new employees.
How does the assistant respect access rights?
The assistant respects the access rights we set together with you. Not every employee sees every document: someone in sales gets answers from the sources they have access to, while management or HR documents stay closed to them. The rights are defined in the first week, together with the choice of sources, and are built into the first version of the assistant.
Before launch the assistant goes through a security check and data rules: which documents go in, which stay out and who is responsible for them. This work runs together with AI governance and the EU AI Act, where the company gets an AI policy and rules for its vendors. When new documents or new teams are added, the rights are reviewed along with the sources, so information does not reach people it is not meant for.
What is the assistant built on?
The assistant is built on the tool that fits your company. If your team works in Microsoft 365, it can be Microsoft Copilot with agents in Microsoft 365. If you use other tools, the assistant can be a project in ChatGPT, Claude or Gemini. When neither fits, we build a custom solution connected to your documents. In all three cases the access rights and the source references are set up to suit the chosen tool.
We are not tied to a single vendor and choose by the task. In medium-sized and large companies we often combine the two tracks by team: Microsoft Copilot for the employees in Microsoft 365 and projects in ChatGPT or Claude for managers and advanced users. We do not sell licences; we help you use them. The choice is made in the first week, together with the sources and the rights, because it decides how both are set up.
How does the project start, and how do the sources stay current?
The project starts with a free 30-minute call, after which you know whether an internal assistant makes sense for you. If it does, in the first week we choose the sources, the rights and the metrics we will measure success by. The test runs with real users and real questions, and the first measurable results usually come within 2–4 weeks of the start. If the assistant does not deliver against the baseline, it does not reach production.
The assistant can be a standalone project or part of a full AI transformation with the Encorp AIM (AI Implementation Method), where it is chosen at Gate 1 and reaches production in weeks 11–12. Documents change, though: new versions come out and old ones stay in the folders. That is why after go-live the assistant moves to AI Ops, where we keep the sources current and improve it every month, with response within 24 hours.
Frequently asked questions
Will everyone see every document?
Can it live in Microsoft 365?
Do we have to use Microsoft Copilot?
Where do the answers come from?
How are the sources kept current?
How soon will we see a result?
How much does it cost?
Book a call with Martin Kuvandzhiev
30 minutes to find out whether and how Encorp can solve this problem for you.
Free call · No commitment