Case Studies
Supporting 4,000+ hospital employees with a private digital assistant
Building first-line support inside a closed healthcare network during a critical software rollout
Sector
Healthcare
Client
Large Dutch hospital group
Challenge
Support thousands of employees during the launch of a new core software system without increasing helpdesk capacity
Solution
Private desktop assistant providing step-by-step first-line support inside the hospital network
01
The challenge
Thousands of employees were about to start using a new core system.
A major hospital group was preparing for the rollout of new core software.
A pilot revealed a predictable problem.
As thousands of employees moved onto the new system, the IT helpdesk would face a sharp increase in support questions during the first weeks after launch.
The expected volume would exceed the available support capacity.
Hiring a temporary support team was not an option.
And the launch date could not move.
The hospital needed another way to provide first-line help at scale.
02
An additional constraint
The assistant could not rely on the public internet.
Healthcare infrastructure comes with stricter requirements than a normal consumer application.
The hospital required the solution to:
Run inside its closed network
Work from managed Windows desktops
Avoid open-internet dependency
Avoid external third-party services
A conventional web chatbot or cloud-based assistant was therefore not suitable.
The support experience needed to operate entirely inside the hospital environment.
03
What we built
A private first-line support assistant.
WeAreBrain developed a custom desktop assistant employees could use when they encountered common problems with the new software.
Instead of immediately contacting the IT helpdesk, employees could ask the assistant for help.
The system provided step-by-step guidance for common support questions.
The assistant ran locally within the hospital’s closed network.
No connection to the open internet was required.
1. Help where employees were already working
No new platform to learn.
The assistant was available directly on employees’ Windows desktops.
That mattered during a major software rollout.
Users were already adapting to one new system.
Support should not require them to learn another complicated tool.
The interaction was deliberately simple:
Open the assistant
Describe the problem
Follow the instructions
Escalate if needed
2. Turn recurring questions into self-service
Many support requests during a software launch are repetitive.
Employees may ask:
How do I complete this task?
Where can I find this function?
Why am I seeing this message?
What should I do next?
The assistant provided predefined step-by-step instructions for the most common issues.
This allowed simple questions to be handled without immediately consuming helpdesk capacity.
3. Keep support content easy to update
The content needed to change right up to launch.
During a system rollout, support teams keep learning.
New questions appear.
Instructions change.
Unexpected issues emerge.
The hospital therefore needed a way to update the assistant without requiring a new software release every time.
We created an Excel-based content-management approach that allowed editors to continue updating help texts and support wizards until shortly before launch.
That gave the operational team control over the knowledge inside the assistant.
4. Connect self-service to second-line support
Automation needs an escape route.
The assistant was not expected to solve every issue.
When a question required further help, the solution connected with the hospital’s existing Topdesk second-line support environment.
That created a simple support model:
Employee has a problem
Assistant provides first-line guidance
Common issue is resolved
Complex issue moves into existing support
The assistant reduced pressure on the helpdesk without replacing the support organisation around it.
04
Private by architecture
Security requirements shaped the product from the beginning.
The assistant could not simply be built as a cloud application and secured afterwards.
Its deployment model was defined by the hospital environment.
The solution operated:
Inside the closed network
On managed desktops
Without open-internet access
Without dependency on external services
This is an early example of a principle that has become increasingly relevant with enterprise AI:
Deployment architecture has to match the sensitivity of the environment.
05
The result
4,000+ employees supported during the critical rollout period.
During the first three weeks after launch:
2,500
assistant interactions were recorded.
And the solution supported:
4,000+
hospital employees.
The assistant helped absorb the initial support peak without requiring the hospital to scale its first-line IT team to the expected demand.
Once the initial wave of rollout questions had passed, the assistant had completed its job.
It was retired and removed from employees’ desktops.
06
Temporary technology can still create lasting value
Not every successful product needs to live forever.
This project had a very specific purpose.
Help the organisation through a high-pressure transition.
The assistant was designed for that period.
Once employees became familiar with the new software and the volume of support questions fell, the need for the tool decreased.
Retiring it was part of the plan.
That is useful product thinking.
Success should be measured against the problem the technology was created to solve, not simply by how long the software remains online.
07
Why this case still matters
The technology has changed. The enterprise problem has not.
This solution was built in 2018.
It predates the current generation of large language models and generative AI assistants.
But several principles behind it are highly relevant to enterprise AI today.
Keep sensitive workloads private
The deployment matched the hospital’s security environment.
Put knowledge close to the user
Employees could get answers where they were already working.
Let domain teams control the content
Support teams could update knowledge without engineering intervention.
Escalate when automation reaches its limit
The assistant connected to existing human support.
Use AI and automation for a defined problem
The technology served a specific operational need rather than becoming a goal in itself.
08
How we would approach this today
A modern private enterprise assistant could extend the same concept using:
Private language models
Retrieval-augmented generation
Internal documentation
Role-based permissions
Source citations
Automated ticket summarisation
Tool integration
Multilingual support
Conversation analytics
Human escalation
A current system could understand a much broader range of employee questions than the rule-based support assistant built in 2018.
The architecture principle remains the same:
Sensitive organisational knowledge stays inside a controlled environment.
09
Beyond IT support
The same pattern can apply wherever employees repeatedly need access to controlled internal knowledge.
Examples include:
Software rollouts
Help employees adopt new enterprise applications.
HR
Answer recurring questions around policies, processes and benefits.
Operations
Provide instructions and procedures at the point of work.
Compliance
Help employees find approved guidance and internal policies.
Healthcare administration
Make operational knowledge easier to access without exposing sensitive information externally.
Field support
Guide employees through technical or operational procedures.
The common challenge is turning internal knowledge into useful assistance without losing control over the environment.
AI services used
Software Engineering
Building a custom assistant around enterprise security, desktop deployment and support integrations.
Product Design
Creating a simple support experience for employees already dealing with a major software transition.
AI Adoption
Designing technology around security requirements, controlled knowledge and human escalation.





Need employees to access sensitive company knowledge without sending it outside your environment?
Private AI does not have to mean a poor employee experience. We design secure assistants that connect people to the information and workflows they need while keeping organisational requirements in control.