Skip to content
Marsh Crest Field
Staff searching through archived documents and procedure manuals
Service 02 · Seven weeks · ¥41,000

Stop rebuilding what your organisation already knows

Years of accumulated procedures, manuals and institutional knowledge sit in files that staff cannot easily search. This engagement makes that material findable — without moving it, restructuring it, or asking anyone to rewrite it.

← Back to Marsh Crest Field
What this delivers

Your existing documentation, made searchable across the organisation

After seven weeks, staff will be able to search across your procedures, manuals and archives using the questions they actually ask — and find answers that already exist in your own material rather than guessing, escalating, or starting from scratch.

The system follows your existing permission structure, so sensitive documents remain visible only to the people who already have access. Nothing is moved, rewritten, or consolidated without your instruction.

Permissions reviewed before deployment

Existing access controls are mapped and honoured. Staff find only material they are already entitled to see.

Evaluation set built with staff input

Before deployment, the system is tested against questions that your staff actually ask — not a standard benchmark written elsewhere.

Written note of where it answers poorly

Every deployment includes honest documentation of the question types the system handles less well — so staff know when to look elsewhere.

The situation

A decade of accumulated knowledge that nobody can find when they need it

Organisations that have operated for more than a few years almost always have the same problem: procedures written in one folder structure, manuals stored in another, archived project notes in a third location nobody remembers. The knowledge exists. Finding it is the difficulty.

When staff cannot locate existing material, they do one of several things. They ask a colleague — which pulls that colleague away from their own work. They escalate to someone more senior — which costs more time. They recreate the information from memory — which introduces inconsistency. Or they simply make a judgement call without the reference they should have had.

None of those options are what was intended when the original documents were written. The material was created so that people could refer to it. The problem is access, not content.

External search tools that are not trained on your own material tend to compound the problem — they return general information rather than your specific procedures, and staff learn quickly not to trust them for anything organisation-specific.

Our approach

Search built on your material, following your existing permissions

We start by inventorying what you have and where it sits. The indexing follows the structure of your existing entitlements — nobody gains access to anything new.

Base layer — existing

Your current documentation estate

Procedures, manuals, archived files, shared drives, internal wikis — wherever your material currently lives, in whatever format it exists. We inventory rather than restructure.

Middle layer — addition

Indexing and search layer

A search index is built across your sources, calibrated against the questions staff actually ask. Permission boundaries are replicated from your existing access controls.

Top layer — oversight

Human review and withdrawal path

The evaluation set is maintained so quality can be checked over time. The withdrawal process — removing the search layer entirely — is documented before deployment begins.

What this engagement does not do

It does not clean up, consolidate, or rewrite your documentation. If your procedures need updating, that is a separate project. The search layer operates on material as it is — which means outdated documents can surface alongside current ones.

The evaluation set we build includes testing for this. Where outdated material consistently appears in search results for a given question type, we flag it in the handover note so your team knows to address it.

The process

Seven weeks, structured in layers

The engagement moves in sequence: understand what exists, set permissions, build and test, then hand over with documentation.

Weeks 1–2

Source inventory and permission mapping

We catalogue what documentation exists, where it lives, what formats it is in, and who currently has access to what. This step determines the scope of the index and establishes which boundaries must be carried into it. No documents are moved or altered.

Weeks 3–4

Evaluation set construction with staff

We work with a small group of staff to collect the questions they actually need answered in their day-to-day work. This set becomes the calibration standard. We are not guessing at what useful search looks like — we are building against real examples from the people who will use it.

Weeks 5–6

Indexing, testing and iteration

The index is built across the sources identified in weeks one and two, then run against the evaluation set. Where results are poor for certain question types, the configuration is adjusted. Permission boundaries are verified before any staff access is granted.

Week 7

Handover and documentation

Deployment is followed by a handover document covering: what was indexed, what was excluded, permission boundaries, known weak areas, how to add new sources, and how to withdraw the system entirely if needed. A walkthrough session is included.

Investment

¥41,000 — fixed across seven weeks

The scope of your documentation estate does not change the fee. Whether you have three hundred documents or thirty thousand, the engagement runs for seven weeks at the agreed cost.

The higher fee relative to the other services reflects the seven-week duration and the permission mapping work, which requires careful attention across sources that may have accumulated over many years without a consistent structure.

¥41,000 JPY

Seven-week fixed engagement

What is included

Full inventory of your existing documentation sources and storage locations

Permission review and mapping prior to any indexing work

Evaluation set developed with staff using their actual day-to-day questions

Search index built across all identified sources, tested before deployment

Written note of question types the system handles less well

Withdrawal documentation — how to remove the system entirely if you choose to

Handover walkthrough session with the team at week seven

Methodology

How search quality is measured and maintained

01 · Evaluation set

Questions from real staff, not synthetic benchmarks

We collect actual questions staff need answered — not questions that make a search system look good in a generic test. The evaluation set belongs to you and can be used to check quality over time.

02 · Honest gap reporting

Where the system answers poorly is documented

No search system covers everything well. We identify the question types and subject areas where results are unreliable and include these in the handover note, so staff know where to look elsewhere.

03 · Reversibility

The withdrawal path is documented before deployment

If you decide to remove the system — for any reason — the steps to do so are written out and handed to you at week seven. The deployment does not create a dependency you cannot exit.

A realistic timeline

In the first weeks after deployment, staff need to learn that the search layer exists and adjust to querying it. Adoption is gradual. The engagement covers the technical work and the handover; broader internal communication about the new capability is yours to manage, though we can advise on how other organisations have approached it.

Commitment

What we stand behind

Permissions are verified before anyone searches

No member of staff will find material through the search system that they could not already access through your existing file structure. This is checked and documented before the system goes live.

The evaluation set reflects actual usage

The test questions come from the people who will use the system, gathered during weeks three and four. The calibration is honest about what your staff need, not optimised to generate a high score on a general metric.

An initial conversation is without obligation

If you are not certain whether your documentation estate is suited to this kind of search layer, get in touch. We are happy to discuss what you have before you decide anything.

Getting started

How an engagement begins

01

Describe your documentation situation

Tell us roughly what you have — types of documents, where they are stored, how many people need to search them, and what the current friction looks like.

02

We respond within two working days

We will ask about format types and storage systems if we need more detail, or suggest a short call to understand the situation before committing either side to anything.

03

Scope and start date confirmed in writing

The sources to be indexed, the permission approach, and the staff involved in evaluation set building are agreed before the seven weeks begin.

04

Seven weeks later: a working system with documentation

Staff can search your material. You have a handover document covering how the system was built, where it works well, where it does not, and how to remove it if you ever need to.

Internal Documentation Search

Make what you already have findable

If your organisation holds years of accumulated documentation that staff cannot easily search, this engagement addresses that directly. Get in touch to discuss what you have and whether this approach fits.

Send us a message

Other services from Marsh Crest Field

Service 01 · Four weeks · ¥31,000

Customer Service Automation Assessment

We analyse your existing customer contacts to assess whether automation suits your specific enquiry mix. The recommendation may be that it does not — and that finding has value too.

Learn more →
Service 03 · Six weeks · ¥30,000

Content Production Workflow

Introducing assisted drafting into your marketing or communications work, with defined review stages and clear rules on where it is not used. Includes style guide encoding and reviewer training.

Learn more →