AI at work
AI assistants that you control
If you wish, AI assistants support your employees and customers directly in the form, in the flow and in the customer portal. You decide which model is used and which data an assistant knows about. Your own API keys are used, and there is no separate AI licence.
- Your own API keys
- No separate AI licence
- The asking person's permissions
- Internal knowledge can be included
In daily work
The AI helps where the work happens
In the form
Suggestions where data is created
From the description of a purchase request, the AI suggests product group and cost centre. The requester confirms or corrects.
In the flow
Preparation for processing
Text is summarised and classified, and field values are suggested. This happens as a step in the process.
In the portal
Answers from your knowledge base
Your customers find the right request and get answers from your knowledge base, without having to call the hotline.
Working with your data
The assistants answer based on your data
A language model that only generates text is of limited use in a business system. It becomes helpful when it can reach your data: search, analyse, open records and run saved queries.
The assistants in nexxcite can do this. They look things up in your data before they answer. They use the same access path as the user interface, so the permissions of the person asking apply.
For a data protection review, this is the important point: the assistant has no separate, extended access to the database. It only sees what the person is allowed to see anyway.
- Access through the same path as the user interface
- Permissions apply automatically, no second rulebook
- Restrictable per assistant: which data areas it knows about
- Restrictable per assistant: which saved reports it may run
- Searches are shown in the conversation and traceable
- The assistant can operate the UI, for example open results in a list
AI at work
AI helps where it saves work
nexxcite offers AI when building, in day-to-day work and in your customers' portal. It works with your data and always with the permissions of the person asking. Where you don't want AI, leave it out. The platform works without it too.
Which open purchase requisitions over €5,000 are still waiting for my approval?
Searches your data
Purchase requisitions · open · amount over €5,000
9 cases
Analyses
Grouped by waiting time and cost centre
3 older than 5 days
Opens the list
As a view in your interface
ready
9 purchase requisitions over €5,000 are waiting for your approval, 3 of them for more than five working days. The list is open, sorted by waiting time.
Searched with your permissions
Ask about your cases …
No tie to one AI provider and no extra licence
- Connected through your own account, billed directly by your AI provider
- All common language models, including open models in your own environment
- No mandatory vendor cloud and no separate AI licence
- Tools from other providers can be connected to the assistants
- The assistants draw on your internal knowledge in their answers
- Set per assistant: which data it knows and what it may do
What the assistants can do
You decide what the assistants may do
Your administrators provide the tools and set their limits. Your employees and customers use them, each within what their role allows.
Questions about your own data in plain language
"Which complaints about line 3 have been open for more than a week?" The assistant searches, analyses and opens the list, with the permissions of the person asking.
- Find and analyse cases
- Open records
- Suggest field values
- Summarise and classify text
In the customer portal
The assistant guides people to the right form, answers questions from your knowledge base and explains the status of their own cases.
Include internal knowledge
Your manuals, policies and knowledge articles are drawn on in the answers, so the assistant relies on your own documents.
Connect further tools
External tools can be connected via the MCP standard, set per assistant and usable in flows too.
As a step in a flow
Suggest field values, summarise and classify text or use an assistant with tool access, directly within the process.
For administrators
The AI can suggest fields, forms, rules and flows, explain existing configuration and find faults in configurations that have grown over time. Changes always go into a draft.
Building with AISovereignty
You choose the model, with no separate AI licence
- Your own API keys. You pay your provider directly for model usage, with no mark-up from us.
- Free choice of model. All common providers are supported, as are OpenAI-compatible endpoints.
- Open models in-house. If you run your own model, you pay no usage fees to a provider and no data leaves your data centre.
- No obligation to use a vendor cloud and no separate AI licence. The assistants are part of the platform.
- External MCP servers can be connected and add your own tools to the assistants.
- Internal knowledge from manuals and policies is included in the answers through indexing (RAG).
AI without special rights
The assistant passes the same permission check as everyone else
The assistant accesses data through the same central permission check as the user interface, interfaces and automation. Data filtered out there is not visible to the AI.
The permission checkRollout
Introducing AI step by step
-
01
Pick a model, add your key
An external provider under your own contract, or an open model in your own environment. Switching later is a setting, not a programming task.
-
02
Set up assistants per workspace
The procurement assistant knows requests and suppliers, the HR assistant knows joiners and training. You decide which data areas each assistant knows about.
-
03
Release tools and knowledge
Saved queries via an allow list, external tools and your internal knowledge, each as much as the use case needs.
-
04
Enable it for roles
First for a pilot group, then for the department, then for the customer portal. Without enablement, no assistant is available.
FAQ
What data protection officers and works councils want to know
Does our data leave the organisation?
That depends on the model you choose. With an external provider, the content of a request goes to its interface, under the terms you agree with that provider. With an open model in your own environment, nothing leaves your data centre. Both options can be set up without programming.
Can the assistant see data a person is not allowed to see?
No. The tools use the same access path as the user interface, and permissions apply there. In addition, you can restrict per assistant which data it knows about.
Can we switch AI off completely?
Yes. Without stored credentials there is no assistant. You can also enable it only for certain roles or certain processes.
What if the model gets it wrong?
The AI makes suggestions, and people decide. A field value suggested by the model is shown and can be corrected. A flow can be built so that a person has to confirm. Where a decision has consequences, that confirmation belongs in the process.
What about logging?
Assistant calls are traceable. The searches the assistant ran are shown in the conversation. For a works agreement, this is usually an important point.
Live demo
Test the AI on a real use case
Ideally with your data protection officer at the table. We know the questions that usually come up there and are glad to answer them before the project starts.
- Your use case, your data
- Rights, model choice, logging
- Data protection welcome