Preparing MD reviews
The most time-consuming documentation item according to the DKI. Five steps, three systems, several people involved – before the first line of the answer is written.
What happens in detail
Colleagues are retiring. The documentation stays. So does the research.
The bottleneck is staff, not technology. A third of the working time that is left goes into documentation instead of care, research and teaching — and university medicine carries the resulting deficit itself.
300,000+
hospital staff will reach retirement age by 2035 — at least one in four.
DKG skilled staff monitoring report 2026
2.9 hours
of documentation per doctor per working day, 2.7 hours in nursing. Largest single item: MD reviews.
DKI survey commissioned by the DKG, July 2024
€2 billion
total deficit facing Germany’s 37 university hospitals in 2027 — after more than 800 million euros in 2024.
Association of German University Hospitals, 2026
One hour less documentation per person per day would be the equivalent of roughly 20,650 doctors and 42,870 nurses gaining time at the bedside. The question is therefore not whether AI replaces staff — it does not. The question is what happens for as long as there is no safe internal offering. Then one appears by private routes: findings, fragments of physician’s letters and study documents in outside accounts, with no processing agreement, no log, no control. Bans have changed nothing about this. It stopped being an efficiency problem long ago; it is a liability problem.
Processes that today cost three systems and five working steps
From daily work at a tertiary care centre — medical controlling, nursing, the clinical trials unit. Every process follows the same pattern: the AI pulls the documents together, prepares the draft and backs it with a source reference. The professional decision stays with people. More than 300 agents have been built; which process your hospital starts with is your call.
Draft physician's letter
Discharge letters are written under time pressure from scattered sources. Follow-up queries from the next treating physician are the result.
What happens in detail
Coding review and case completeness
Secondary diagnoses and procedures get lost, billing objections pile up, revenue is left on the table.
What happens in detail
Study costs and third-party funding
Calculations live in inherited spreadsheets. Calls for proposals from federal, state, EU, DFG and foundation sources are searched one at a time.
What happens in detail
Nursing progress documentation
Documentation happens at the end of the shift, from memory, under time pressure. 2.7 hours a day are the lever.
What happens in detail
SOP and in-house standard enquiries
Anyone looking for the work instruction from 2019 asks someone who has been here long enough. When staff change, that knowledge goes.
What happens in detail
Research and teaching
The second and third mission
are not a footnote.
University medicine is care, research and teaching under one roof — with separate budgets, separate purpose limitations and the same overstretched people. Research, registry and study data are opened up only with their own purpose limitation, their own data protection impact assessment and a separate permissions space: what belongs to the clinical trials unit is not visible to administration — and the other way round.
Study cost calculation
Draft based on your cost rates and your separation of accounts — for academic and industry-sponsored studies alike. The clinical trials unit approves the calculation, not the system.
Third-party funding search
Open calls from the DFG, the federal ministry, the EU, the federal states and foundations matched to your research idea, each with source and deadline. Research and fit assessment, not the application itself.
Manuscript and application checks
Formal check against the author guidelines of the target journal or against the funding guideline. Nobody judges the content but you.
Teaching and your own assistants
Prepare curricula, learning objectives and examination formats from your own sources, plus onboarding for final-year students and specialty training. Institutes and working groups build their own assistants through the SDK — whatever they build inherits roles, permissions and the audit trail automatically.
Your institutional knowledge is opened up before you start.
After the Hospital Future Act, no hospital has too few systems. What is missing is the layer that makes them usable. So we do not arrive with an empty system: your in-house standards, SOPs, work instructions, coding and procedural rules and the rulebooks you choose to include are built up, checked and opened up with source references. On top of that, the departments add a layer of their own: processing notes, templates and the documentation of the connected systems — HIS, DMS, coding software, study and registry systems.
300+
preconfigured agents
61
use cases in healthcare, prioritised by ROI, data protection profile and speed of implementation
4
knowledge layers — rulebook, guidelines, professional standards, institutional knowledge
The value is not in the model but in prepared, reusable knowledge. Your systems are connected once and fully opened up in the process: the neuland.ai HUB learns your terms and how they relate — what counts as a case in your hospital, how a coding guideline connects to a work instruction, which document superseded which. This holds at the scale of a tertiary care centre: archives, findings and imaging documentation, research and registry data in the terabyte to petabyte range are opened up and deliver answers with a source reference — even when the evidence is one sentence in a document from twelve years ago. The professional tailoring happens once per process; after that the building blocks can be reused by other departments.
The agent never decides.
It prepares.
Intended purpose: support for administrative, documentary and academic work. No diagnosis, no treatment recommendation. This boundary is drawn deliberately — it makes clear where medical responsibility sits. Four things make that auditable: for data protection, for legal and for employee representation.
A source reference for every answer
Document, passage, version. If the AI finds no evidence, it says so — instead of writing around it. In hospital practice the source reference is the real benefit, not the text.
Medical and nursing sign-off
No draft takes effect without sign-off. The approval levels follow your responsibilities and your signing authority, not our default setting.
Complete logging
Who asked what and when, which sources it drew on, which model answered, what was adopted and what was discarded. Per enquiry, exportable for any audit.
Employee representation from day one
A works council that has been bypassed will sink a project faster than any tender. A template works agreement is available, with performance and behaviour monitoring as a section of its own: no analysis at individual level, no performance rating, purpose limitation set out in writing.
Infrastructure for years. Not a tool for a quarter.
The model is the easy part. What comes with it is the tailored process and the documentation for data protection, procurement and co-determination. The security evidence your supply chain needs under Section 30 BSIG is already in place. All of it stays operable without an AI team of your own: it is used from within everyday medical, nursing, administrative and academic work.
Board questions
Five answers, before you put your name behind it.
The questions come anyway — from the board, from IT, from employee representation and from procurement. Here they are, with the answer we would give in the room.
Getting started
A measurable result before the investment decision.
Data protection and employee representation are involved from phase 1. Procurement clearance, co-determination and the board proposal run in parallel — the pilot does not block these steps, it feeds them.
60 minutes
Initial conversation. Your starting position, our use-case method, a realistic path. On request straight with your IT team — then as an architecture session.
1–2 workshops
Potential analysis. A structured review of your process and system landscape. Result: a prioritised roadmap with ROI estimate, data protection classification and go-live window per use case.
4–8 weeks
Proof of Value. Two or three use cases without patient data, in production. You set the success criteria. If we do not meet them, the pilot ends with no follow-on commitment.
The platform is up in days. What decides matters in large hospitals is what comes afterwards: named contacts instead of a ticket system, onboarding split by medical staff, nursing, administration, IT and research, assistants with prepared input fields for everyone who would rather not write prompts, and hypercare in the first weeks. Professional Services are available for connections to HIS, coding software, rostering and study systems.
Agent Library
An agent for every recurring task.
The Agent Library holds ready-made building blocks for specific tasks — each with its task scope, its boundary of responsibility and its operating model. Four of them cover processes you have read about above:
MD Review and Coding Agent
50%less review effortRaises first-pass coding quality: the only scalable lever before review rates tighten further.
HealthcareMedical controlling, CodingPhysician's Letter Agent
60%less writing effortProduces structured physician's letters from treatment data.
HealthcareMedical staff, DocumentationNursing Planning Agent
30%less documentation effortSupports nurses with structured documentation and shift planning.
HealthcareNursing, DocumentationStudies and Ethics Agent
45%faster contract reviewGuides you through contract clauses, the ethics pathway and research data up to study start.
HealthcareClinical trials unit, Research