Enterprise Search
One search across systems that each think they are the only one.
What is enterprise search?
Enterprise search means finding information across all the systems an organisation runs, which is difficult less because of search technology than because permissions differ between systems and the same thing is named differently in each.
Search inside one system is a solved problem. Search across eight is not, and the reasons are structural rather than technical.
Each system has its own permissions, and combining them is the hard part. If search returns a document somebody could not otherwise open, you have created a data breach with a search box attached, and it is easy to do accidentally.
The second problem is naming. A customer is an account in one system, a client in another and a debtor in a third, and search that does not know they are the same thing will confidently return a partial answer.
Why permissions come first
Duara builds cross system search in Kenya where information genuinely lives in several places and people cannot find it.
We treat permissions as the first design question rather than a configuration detail. Search that quietly widens access is the most common serious failure in this work, and it is discovered by the wrong person rather than by a test.
Who this is for
- Information lives across several systems
- People ask colleagues rather than searching
- The same thing is named differently in each system
- Nobody knows where a document actually is
Industries we serve
- Financial services
- Professional services
- Manufacturing groups
- Healthcare organisations
- Public sector
- Large corporates
Problems this solves
These are structural problems rather than search quality problems, which is why buying a search product rarely fixes them.
- Information scattered across systems. What somebody needs sits in one of eight places and they have to guess which before they can start looking.
- Permissions that differ by system. Combining access rules is genuinely difficult, and getting it wrong creates a breach rather than an inconvenience.
- The same thing named differently. A customer, an account and a client are one entity across three systems, and search that does not know returns a partial answer confidently.
- Asking colleagues instead. People stop searching and ask somebody, which works and quietly consumes the time of whoever knows.
What we deliver
Search across systems with permissions enforced at the strictest level, entity naming reconciled, results showing where each came from, and honest reporting of what was not searched.
Search across the systems that matter
Permissions enforced at the strictest applicable level
Entity naming reconciled across systems
Results showing which system each came from
Reporting on searches that returned nothing
A list of systems deliberately not included
Features
- Permissions designed first. Search that quietly widens access is the most common serious failure here, and it gets discovered by the wrong person rather than by a test.
- Strictest permission wins. Where two systems disagree about who may see something, the tighter rule applies rather than the more convenient one.
- Naming reconciled. A customer, a client and an account treated as one entity, since search that does not know returns a confident partial answer.
- Source shown per result. Which system a result came from, because the same document in two places usually means one of them is out of date.
- Failed searches reported. What people looked for and did not find, which is usually more useful than the successful searches.
Benefits
- People stop asking colleagues. The informal asking that consumes experienced staff time reduces when searching actually works.
- Access stays correct. Permissions designed rather than configured means search does not become an accidental disclosure route.
- Complete answers. Reconciled naming means a search for a customer finds everything about them rather than one system worth.
- Gaps become visible. Failed search reporting shows what people expect to find and cannot, which frequently points at a real filing problem.
Technologies we use
- Cross system indexing. Building a searchable index across the systems that genuinely hold information people need.
- Permission resolution. Combining differing access rules with the strictest applicable one taking precedence.
- Entity reconciliation. Recognising that different names across systems refer to the same customer, project or person.
- Source attribution. Showing which system each result came from and when it was last updated.
- Failed search reporting. Capturing what people searched for unsuccessfully.
The failure that matters is not poor results. It is search returning something the searcher was never entitled to see, which is a breach created by a convenience feature and is trivially easy to build by accident.
How the process works, step by step
Enterprise search usually takes six to twelve weeks, and permissions take longer than indexing.
-
Step 1
Free systems review
We establish where information actually lives and which systems people currently have to guess between.
-
Step 2
Permissions mapped first
Access rules across systems are reconciled before anything is indexed, since this is where the serious failure happens.
-
Step 3
Naming reconciled
Entities that appear under different names in different systems are matched, which decides whether answers are complete.
-
Step 4
Indexed and tested by role
Search is tested as different user roles specifically to confirm nobody sees what they should not.
-
Step 5
Failed searches reviewed
What people cannot find is reported, which usually reveals filing problems rather than search problems.
Quality assurance and how we work
The commitments attached to making everything findable.
- Permissions are the first design question. Search that quietly widens access is the most common serious failure in this work and it is discovered by the wrong person.
- The strictest rule wins. Where systems disagree about who may see something, we apply the tighter rule rather than the more convenient one.
- We test as real user roles. Confirming that a junior account cannot reach senior material is a specific test rather than an assumption.
- We report what was not indexed. A system left out for good reason is stated, since a search that silently excludes a source misleads about completeness.
- We report failed searches. What people looked for and did not find is usually more useful than what they found, and it is normally discarded.
Frequently asked questions about enterprise search
Why is search across systems hard?
Not for search reasons. Permissions differ between systems and the same thing is named differently in each, and both are structural rather than technical.
What is most likely to go wrong?
Search returning something the searcher was never entitled to see. That is a breach created by a convenience feature and it is easy to build by accident.
How do you handle conflicting permissions?
The strictest applicable rule wins. Where two systems disagree we apply the tighter one rather than the more convenient.
Do you test permissions specifically?
Yes, as real user roles. Confirming a junior account cannot reach senior material is a test rather than an assumption.
What is entity reconciliation?
Recognising that a customer, a client and an account across three systems are one thing, without which search returns confident partial answers.
Will it search everything we have?
Only what we index, and we state what was left out. A search silently excluding a source misleads about completeness.
Why report failed searches?
Because what people looked for and could not find is usually more useful than what they found, and it is normally discarded.
Does it show where results came from?
Yes, and the same document appearing in two systems usually means one of them is out of date, which is worth seeing.
How long before staff can search across systems?
Six to twelve weeks, and permissions take longer than indexing does.
Will it work with older systems?
Frequently, through database or file level indexing where no interface exists, with the limits stated rather than glossed over.
Can it search email and chat?
Technically yes, and we would raise the privacy questions before doing it rather than after somebody objects.
Is this the same as a knowledge base?
No. This finds what exists across systems. A knowledge base is written material maintained deliberately.
Does it answer questions or return documents?
Returns results with sources. Where you want answers generated from them, that is retrieval work and a different scope.
What if our filing is chaotic?
Search helps and does not fix it. The failed search report usually shows filing problems that no search quality would resolve.
Can we restrict search by department?
Yes, and department scoped search is often the sensible starting point rather than opening everything at once.
Does content leave our systems?
No. Indexing runs in your infrastructure and content stays where it is.
Will it slow our systems down?
Indexing is scheduled to avoid load periods, and we would size that against your actual usage rather than assuming.
Can new systems be added later?
Yes, and each addition brings its own permission mapping, which is the work rather than the connection.
Who decides what gets indexed?
You do, and we would recommend starting narrower than feels ambitious since permissions are the constraint.
What does enterprise search cost?
System count and permission complexity determine it, quoted after the systems review.
Glossary of terms
Search terms in plain language.
- Cross system search
- Searching several systems at once rather than each separately.
- Permission resolution
- Combining differing access rules so search never returns what somebody could not otherwise open.
- Entity reconciliation
- Recognising that different names in different systems refer to the same thing.
- Index
- The searchable structure built over your content, which is where permission decisions are enforced.
- Source attribution
- Showing which system a result came from and when it was last updated.
- Failed search
- A query that returned nothing useful, which usually indicates a filing gap rather than a search fault.
Related services in AI Solutions
Talk to us about enterprise search
Ask somebody to find a document they know exists but did not create. How long that takes is the measure.
Start your project