Use case
Document search
In every company that has grown over time, the knowledge sits in folders only two people know. The bottleneck is not searching - it is that nobody searches and asks a colleague instead.
Usually the process accepted fastest
The process
Where a person is needed - and where they no longer are.
- ClaudeLab from here
Step 01Person
Question
Intake
Step 02ClaudeLab
Search
Gathering
Step 03ClaudeLab
Answer
Output
The difference
The same process, once as it is today and once with us.
Step 01 · today
Step 01 · with us
Step 02 · today
Step 02 · with us
Step 03 · today
Step 03 · with us
What we build
- Component 01
- Existing filing is opened up - without you having to tidy it first.
- Component 02
- Answers come with a source: which document, which passage, which revision.
- Component 03
- Permissions stay as they are - what someone may not open today, they will not see here either.
Degree of autonomy
Pure enquiry runs independently, because it changes nothing. Anything that creates or sends a document is a different process.
Where it does not fit
If the records contradict each other, the search makes that visible - it cannot resolve it. Sometimes that is the actual finding.
More specific
What this looks like in your sector
The process is sector-independent, the fit is not. These pages show the systems, the formats and the rules that actually apply there.
Mechanical engineering
In mechanical engineering the question is almost never “where is the file” but “which revision was built”. An answer without a source is therefore worthless.
ViewIndustrial suppliers
At a supplier, document search is a liability matter. What cannot be found within hours in an audit does not exist as far as the auditor is concerned.
ViewLaw firms
In a law firm the permission boundary is part of professional conduct. A search that does not reproduce it exactly is not impractical, it is impermissible.
ViewFrequently asked
What does it cost to automate document search?
From €6,900 for the first process, from €4,900 for each further one in the same project. The firm price follows the audit (€2,500, credited against the project) - before that, any number would be a guess.
How long does implementation take?
3 to 5 weeks, depending on the state of the filing. Building it is usually the shorter half - the time goes into access, coordination and test cases.
What do we have to provide?
Read access to the archives that are to be searched - network drive, mail archive, DMS. A permissions rule: who may see which records. The system adopts it, it does not work around it. A dozen typical everyday questions as touchstones. Nothing more is needed to start.
How will we know it works?
From verifiable criteria agreed before the build: On 50 prepared questions, at least 85 % come back with an answer and a correct source. Every answer names the document it came from - no source, no answer. Nobody sees records through the search that they could not open directly.
Does the AI decide on its own?
Pure enquiry runs independently, because it changes nothing. Anything that creates or sends a document is a different process.
When is this not worth it?
If the records contradict each other, the search makes that visible - it cannot resolve it. Sometimes that is the actual finding.
Which sectors does this suit?
In our experience particularly Mechanical engineering, Industrial suppliers, Law firms. The process itself is sector-independent, though - what matters is whether it repeats.
Tell us where your time goes.
In the audit we record the process, work out what is worth automating, and you get a plan with fixed prices - credited against the build if you go ahead.