Engagements
Three engagements, each addressing a specific part of how organisations work with automated systems
Data preparation, model monitoring, and governance setup. Each is a separate engagement with its own scope, timeline, and written output. They can be undertaken individually or in sequence.
Back to homeHow to read this page
Each engagement is described in two registers
For management
The descriptions below explain what problem each engagement addresses, who it is appropriate for, and what you receive at the end. Prices and timelines are stated for each one. Nothing on this page requires a technical background to evaluate.
If you are not sure which engagement is the right starting point for your organisation's situation, the initial conversation is the right place to discuss that — before any scope is agreed.
For technical staff
Each engagement description also covers the technical scope: what is assessed at intake, the specific decisions made during the engagement, and the format of the deliverable. The measurement panel below each engagement states what is checked at start and at close.
The three engagements are independent. Undertaking the data preparation engagement does not commit you to the monitoring or governance engagements, and vice versa.
Engagement 01
Data Preparation
¥40,000 JPY · Seven weeks
Work on the records that any automated system will depend on, addressing duplicates, inconsistent formatting, gaps in historical fields, and definitions differing between departments. Frequently the necessary step before any other project can begin.
For management
Most organisations that have been operating for several years have records that have accumulated across multiple systems, sometimes with different naming conventions, missing fields, and duplicate entries that were never resolved. Automated systems that depend on these records inherit their problems.
This engagement is appropriate for companies whose records have accumulated across several systems and reorganisations. The result is records your automated systems can read consistently and a set of rules that keep new entries consistent going forward.
What is included
- —A documented data dictionary in an agreed open format
- —A cleaned dataset delivered alongside the original
- —Validation rules for keeping new records consistent
- —A review session at close with your team
For technical staff
The intake assessment covers duplicate record rate, field completion percentage across target tables, definition discrepancies between departments, and any existing documentation of data structure. These figures form the baseline.
Cleaning decisions are made in sessions with the staff who know the data — not applied independently. Each decision is recorded in the dictionary with its rationale. Validation rules are written in a format applicable to new records without specialist intervention.
Measurement panel
At intake
Duplicate rate, field completion, definition conflicts
At close
Same metrics against cleaned dataset and validation rules
Engagement 02
Model Monitoring Setup
¥36,000 JPY · Five weeks
Establishing ongoing monitoring for systems already deployed, covering accuracy drift as conditions change, distribution shifts in incoming data, alert thresholds, and the review routine that follows an alert. Intended for organisations running systems introduced a year or more ago without continuing measurement.
For management
Systems that were accurate when deployed can become less accurate over time as the data they process changes. Without a monitoring routine in place, this drift goes undetected until a failure surfaces — often at a point where it has already affected decisions.
This engagement sets up the measurement infrastructure your team needs to catch problems before they affect operations, including a clear routine for what to do when an alert fires and named staff responsible for acting on it.
What is included
- —A monitoring dashboard with defined alert thresholds
- —An escalation routine naming responsible staff
- —A written assessment of current performance against original figures
- —A review session at close with your team
For technical staff
The intake assessment draws a baseline from current system logs: accuracy against original deployment figures, distribution of incoming feature values, and any existing alert infrastructure. Where original accuracy figures are not available, that limitation is documented and a current baseline is established instead.
Threshold values for alerts are decided with your team, not set independently. The dashboard specification is written to your monitoring infrastructure. The escalation routine is written for your organisational structure, naming actual staff roles rather than generic titles.
Measurement panel
At intake
Current accuracy, data distribution, existing monitoring coverage
At close
Dashboard operational, thresholds set, escalation routine documented
Engagement 03
Governance and Policy Setup
¥30,000 JPY · Five weeks
Establishing internal rules for the use of automated tools, covering approved applications, information that may not be entered into external services, review requirements for output used in decisions, and record keeping for oversight. Intended for organisations where staff have begun using tools ahead of any policy.
For management
In many organisations, staff have begun using AI tools informally — for drafting, summarising, or analysis — without any internal guidance about what is appropriate. The risk is not that the tools are being used; it is that there are no rules covering what data can be submitted to external services or what review is required before output is acted upon.
This engagement produces a written policy sized for your organisation, a staff briefing session, and a register template for tracking tools in use. The result is a documented position your organisation can point to and maintain.
What is included
- —A written policy sized for the company, with legal and operational input
- —A staff briefing session covering the policy provisions
- —A register template for tracking systems in current use
- —A review session at close with your team
For technical staff
The intake assessment covers the number of AI tools in current use, whether any have documented approval, what data categories are being submitted to external services, and what review process exists before output is used in decisions. This forms the compliance gap baseline.
The policy is drafted with legal, technical, and operational input across five weeks. Policy provisions are explained during drafting so that the staff responsible for enforcing them understand the reasoning. The register template is designed for ongoing maintenance without specialist input.
Measurement panel
At intake
Tools in use, undocumented usage count, staff awareness level
At close
Written policy complete, briefing delivered, register template in place
At a glance
Engagements side by side
| Data Preparation | Model Monitoring | Governance Setup | |
|---|---|---|---|
| Price | ¥40,000 JPY | ¥36,000 JPY | ¥30,000 JPY |
| Duration | Seven weeks | Five weeks | Five weeks |
| Appropriate for | Records accumulated across systems and reorganisations | Systems deployed a year or more ago without monitoring | Staff using tools ahead of any internal policy |
| Main deliverable | Data dictionary, cleaned dataset, validation rules | Monitoring dashboard, escalation routine, performance assessment | Written policy, staff briefing, register template |
Starting point
How to identify where to begin
If you are not sure which engagement fits your situation, the following questions give a rough guide. They are not definitive — the initial conversation is the right place to work through the specifics of your situation before any scope is agreed.
Start with data preparation if:
Your organisation's records have been accumulated across several systems, or you are planning to build or improve an automated system and are not confident the underlying data is consistent enough to be reliable.
Start with monitoring if:
You have systems already in production that have not been checked against their original accuracy figures, or you have no alert mechanism for catching changes in model performance over time.
Start with governance if:
Staff are using AI tools without documented rules, or an audit or compliance review has raised questions about your organisation's position on AI tool use that you cannot currently answer in writing.
Undertaking more than one
The three engagements are independent and can be undertaken in any order or separately. However, there is a natural sequence for organisations building from minimal infrastructure:
Data preparation addresses the foundation that other automated systems depend on. Problems here affect everything downstream.
Model monitoring makes deployed systems measurable. It is most useful once you have a clear picture of your data quality.
Governance can be undertaken at any point and is often the first engagement for organisations where compliance is the pressing concern.
If you are considering more than one engagement, the initial conversation is a useful point to discuss sequencing and whether anything changes between them.
Next step
Discuss which engagement suits your situation
The initial conversation is a discussion, not a commitment. We cover what you are currently working with, which engagement is the right starting point, and what the scope would look like. No obligation until scope is agreed in writing.
Get in touch