Examples / Real-world examples
Real-world examples
What AI already does – six examples at a glance
Six real projects from large organisations. Each example is drawn with one of our three designs: who asks, what the AI does, which systems it works with and what comes out.
Overview
Six projects, six industries
Cloud AI
Example 1 · Meridian
State employees save 95 minutes a day with an AI assistant
Connected systemsUploaded documentsGuidelinesTemplates
Cloud AIExample 2 · Lighthouse
The AI listens to the visit and writes the note
Connected systemsPatient recordNote templates
Cloud AIExample 3 · Compass
Advisers ask the firm's whole knowledge library – and the AI writes the meeting notes
Connected systemsResearch libraryVideo meetingsEmailCRM
On-premise AIExample 4 · Blue Lion
A virtual assistant has answered three billion customer requests
Connected systemsAccountsPaymentsCardsAppointments
Cloud and on-premise AIExample 5 · Evergreen
AI completes 850 million pieces of product data
Connected systemsProduct catalogueSupplier dataProduct images
Cloud and on-premise AIExample 6 · Northstar
An AI agent saves 4,500 years of developer work
Connected systemsCode repositoryBuild systemTest system
People in the organisationThe orchestrator: runs the task step by stepThe AI modelThe gateways: every request, step and answer passes themConnected systems: the software of the organisation, different in every example
Three designs, chosen by size. One platform, for about 25 to 5,000 users: AI gateway and MCP gateway in one product (example 1). Two gateways, for about 1,000 to 20,000 users: the AI gateway and the MCP gateway are separate, and each application has its connector, an MCP server (examples 2 and 3). Group of companies, for about 5,000 users and above: one shared orchestrator and AI gateway, an MCP gateway for each company and a central registry (examples 4 to 6).
The figures come from the organisations' own publications. The diagrams are our simplified illustration: they show how such a task is built with one of our designs, not the systems of these organisations.
Example 1 · Public administration · office work
State employees save 95 minutes a day with an AI assistant
Project Meridian · a state governmentCloud AIDesign: One platform
Show the drawing
State employeeasks in the chat window
↓↑
Orchestratorruns the task step by step
↓↑
AI gatewayWho is asking? · What is allowed? · Logs every step
↓↑
Cloud AIwrites, researches, summarises
and
Connected systems
- Uploaded documents
- Guidelines
- Templates
Steps 3 to 5 repeat until the task is done. Then the answer goes back.
Result: Draft, summary or answer
The employee reviews and decides
Published results
- For one year, 175 employees from 14 agencies used an AI assistant for writing, research, summaries and IT support.
- They reported saving 95 minutes a day on average when using it.
- Over 85% rated the experience as positive. The pilot also found that AI is not the right tool for every task, and that human review stays important.
Example 2 · Healthcare · visit notes
The AI listens to the visit and writes the note
Project Lighthouse · a medical groupCloud AIDesign: Two gateways
Show the drawing
Physiciantalks with the patient as usual
↓↑
Orchestratorruns the task step by step
↓↑
AI gatewaychecks every request and every answer
↓↑
Cloud AIlistens and writes the draft note
and
MCP gatewayWho is asking? · What is allowed? · Logs every step
↓↑
MCP serversone connector per application
↓↑
Connected systems
- Patient record
- Note templates
Steps 3 to 5 repeat until the task is done. Then the answer goes back.
Result: Draft visit note in the patient record
The physician checks and edits the note
Published results
- Over 63 weeks, 7,260 physicians used the AI in more than 2.5 million patient visits.
- It saved an estimated 15,791 hours of documentation time – 1,794 working days.
- 84% of surveyed physicians reported a positive effect on communication with patients. The AI gives no diagnoses, and the physician can edit every note.
Example 3 · Wealth management · adviser knowledge
Advisers ask the firm's whole knowledge library – and the AI writes the meeting notes
Project Compass · a wealth managerCloud AIDesign: Two gateways
Show the drawing
Financial adviserasks a question or holds a client meeting
↓↑
Orchestratorruns the task step by step
↓↑
AI gatewaychecks every request and every answer
↓↑
Cloud AIfinds the answer, takes notes, drafts the email
and
MCP gatewayWho is asking? · What is allowed? · Logs every step
↓↑
MCP serversone connector per application
↓↑
Connected systems
- Research library
- Video meetings
- CRM
Steps 3 to 5 repeat until the task is done. Then the answer goes back.
Result: Answer from the firm's knowledge · meeting note and draft email
The adviser edits and sends the email
Published results
- 98% of adviser teams use the internal AI assistant. It answers questions from the firm's own research and knowledge library.
- A second tool takes notes in client meetings, with the client's consent, summarises the key points and drafts the follow-up email.
- It saves the note in the customer system. One adviser reports saving about half an hour per meeting.
How would this look in a specific case? 20 minutes, without obligation.
Discuss a use caseExample 4 · Banking · customer assistant
A virtual assistant has answered three billion customer requests
Project Blue Lion · a large bankOn-premise AIDesign: Group of companies
Show the drawing
Customerasks in the banking app
↓↑
Orchestratorshared by all companies
↓↑
AI gatewaychecks every request and every answer
↓↑
On-premise AIunderstands the question, answers or completes the task
and
MCP gatewaysWho is asking? · What is allowed? · Logs every step · one for each company
↓↑
MCP serversthe connectors of each company
↓↑
Connected systems
- Accounts
- Payments
- Cards
- Appointments
Central registryapproved tools · one policy for all companies
Steps 3 to 5 repeat until the task is done. Then the answer goes back.
Result: Answer or completed task in the app
Appointment with a specialist for complex matters
Financial services: secure and reliable. In our design the AI model runs on-premise. Account and payment data stay within the network of the bank, every request and every answer is checked and logged, and the system is built redundantly: if one server fails, another takes over.
Published results
- Since its launch in 2018 the virtual assistant in the banking app has handled more than 3 billion customer requests – now over 58 million a month.
- Nearly 50 million customers have used it.
- A version for staff is used by over 90% of employees and has cut calls to the IT service desk by 50%.
Example 5 · Retail · product catalogue
AI completes 850 million pieces of product data
Project Evergreen · a large retailerCloud and on-premise AIDesign: Group of companies
Show the drawing
Catalogue teamdefines which product data is needed
↓↑
Orchestratorshared by all companies
↓↑
AI gatewaychecks every request and every answera rule decides: cloud or on-premise
↓↑
Cloud AIseveral language models, in the cloudreads product information and fills in the details
On-premise AIa smaller model for confidential data
and
MCP gatewaysWho is asking? · What is allowed? · Logs every step · one for each company
↓↑
MCP serversthe connectors of each company
↓↑
Connected systems
- Product catalogue
- Supplier data
- Product images
Central registryapproved tools · one policy for all companies
Steps 3 to 5 repeat until the task is done. Then the answer goes back.
Result: Complete, consistent product data
Published results
- Several AI language models created or improved more than 850 million pieces of product data in the catalogue.
- According to the company, the same work done by hand in the same time would have needed about 100 times the staff.
- Better product data helps customers find what they are looking for, and helps staff find the items when they pick online orders.
Example 6 · Software · modernisation
An AI agent saves 4,500 years of developer work
Project Northstar · a technology companyCloud and on-premise AIDesign: Group of companies
Show the drawing
Developernames the application to upgrade
↓↑
Orchestratorshared by all companies
↓↑
AI gatewaychecks every request and every answera rule decides: cloud or on-premise
↓↑
Cloud AIthe strong AI, in the cloudplans the upgrade and rewrites the code
On-premise AIa smaller model for confidential data
and
MCP gatewaysWho is asking? · What is allowed? · Logs every step · one for each company
↓↑
MCP serversthe connectors of each company
↓↑
Connected systems
- Code repository
- Build system
- Test system
Central registryapproved tools · one policy for all companies
Steps 3 to 5 repeat until the task is done. Then the answer goes back.
Result: Upgraded application
Developers review the plan
Published results
- An AI agent upgraded tens of thousands of the company's applications to a current software version.
- Compared with doing it by hand, this saved more than 4,500 years of developer work.
- The upgraded applications run more efficiently: the company puts the saving at 260 million US dollars a year.
Contact
Your AI. Your data. In the EU.
Discuss a use case with us
20 minutes, without obligation: the situation is described, and we say what AI can realistically do, and what it cannot.
Open the form
- contact@ingnetworks.com
- Phone
- +49 1515 3851511
- Address
- Ingenious Networks GmbH
Eisenacher Straße 6
10777 Berlin, Germany