Kirameki
Principles and philosophy behind structured AI integration

Our principles

What we believe about how AI integration should work

Not a mission statement. A description of how Kirameki approaches its work and why — written so that prospective clients can decide whether it aligns with how they would like to work.

← Back to home

Foundation

Where we start from

Kirameki was set up on the observation that AI tools are acquired quickly and evaluated slowly — or not at all. That is not a criticism of organisations using AI. It is a consequence of how accessible these tools have become. Signing up is easy; measuring what changes is less straightforward.

The work Kirameki does is built around making that measurement straightforward. Every engagement records something before any change is made, compares it against what the change produced, and delivers those findings in writing. That structure does not require a particular view on whether AI is good or bad for business — it just requires that the question be answered with evidence rather than impression.

Vision

How we think about AI in organisations

Tools serve purposes

AI tools are not ends in themselves. They are worth using when they produce a measurable benefit for a specific task. Whether that is the case should be determined by the organisation using them, based on their own figures.

Decisions belong to clients

Kirameki's role is to structure a pilot and deliver a written comparison. What happens after that is the client's decision — including the decision to stop. That should remain true regardless of how the results look.

Evidence ages well

A written baseline from two years ago is still useful today. It tells you what the process was doing before the change, which is the one thing you cannot reconstruct after the fact. Documentation is worth the effort.

Core beliefs

What we hold to be true about this work

01

Baseline before everything

If you do not know what was happening before a change, you cannot know what the change produced. This is so basic that it is frequently skipped. We do not skip it.

02

Scope protects everyone

A pilot that starts as one process and quietly becomes several is no longer a pilot. Clear scope from the beginning means results can be attributed to what was actually tested.

03

Stopping is a valid outcome

Not every pilot will produce results that justify continuation. That is useful information, not a failure. A decision framework that does not include the option to stop is not a decision framework.

04

Written outputs persist

Verbal summaries disappear when people leave. Written deliverables — reports, dictionaries, policies — stay in the organisation and can be referred to later. We produce documents, not presentations.

05

Data quality is not optional

Automated systems produce outputs shaped by the records they work from. Poor input quality becomes poor output quality. Addressing data before automation — not after — is the reliable sequence.

06

Governance before problems

Policies developed in response to problems carry more difficulty than policies developed in advance. Knowing what tools are in use, and what rules govern their use, is easier to establish before issues arise.

In practice

How these beliefs shape what we do

Every engagement starts with a conversation about scope

Before any agreement is signed, we discuss which process will be examined, what will be measured, and over what period. That conversation produces a shared understanding of what the engagement is — and is not.

Baseline figures are recorded before any change

This is not optional. If we cannot agree on what to measure before we start, we revisit the scope. A pilot without a baseline is an anecdote, not an evaluation.

Parallel operation is standard

During a pilot, the old process is not removed. Both approaches operate simultaneously. This produces a direct comparison and also preserves continuity if the pilot is discontinued.

The written comparison is the product

At the close of a pilot, the deliverable is a written comparison of the baseline figures against what the pilot produced. This is what the client uses to decide what comes next. It is written to be useful to someone who was not involved in the project.

Recommendations are based on the client's situation

Kirameki does not sell or represent software products. When a tool or approach is recommended, that recommendation is based on the client's data and process — not on a preferred vendor relationship or platform partnership.

Fees are fixed before work starts

Each engagement has a stated price known before any commitment is made. There are no usage-based components or fees that accumulate as the engagement progresses.

Human-centred approach

The people in any given organisation are the constant

AI tools come and go. Processes change. But the people who work in an organisation are there through those changes — and they are the ones who use, maintain, and sometimes resist the systems that consultants recommend. Engagements that do not account for that tend to produce good-looking deliverables that are not used.

Kirameki structures its work around the people who will work with any change it proposes. That means involving the people who run the relevant process in scoping what will be measured. It means explaining findings in plain language, not specialist vocabulary. And it means delivering documentation that a non-specialist can read and work from later.

On innovation

New approaches are worth testing, not assuming

The consulting industry has a tendency to describe new methods as improvements before those methods have been tested in the client's context. Kirameki tries to resist that. A pilot is not premised on the assumption that automation will be better. It is an opportunity to find out whether, in this particular organisation, for this particular process, it is.

This matters for the client because it means the result of a pilot can genuinely go either way. And it matters for the quality of the work because it means the scoping, measurement, and comparison have to be done carefully — there is no predetermined answer to work backwards from.

Integrity

Transparency about what we can and cannot tell you

What we can tell you before a pilot

  • — What will be measured and how
  • — What the pilot period will cover
  • — What the written deliverable will include
  • — The fee for the engagement

What we cannot tell you before a pilot

  • — Whether the automation will be faster
  • — Whether the cost will be lower
  • — Whether the output will be more accurate
  • — Whether continuing is the right decision

What the pilot will tell you

  • — What the baseline figures were
  • — What the pilot produced over the same period
  • — Where discrepancies or exceptions arose
  • — A framework for deciding what to do next

Working together

How Kirameki works with clients

Kirameki is not a large consultancy with a standard playbook applied to every organisation. Engagements are small, focused, and involve direct contact with the people doing the work. The scope conversation at the start of each engagement is not a formality — it is where the specific context of the organisation shapes what happens next.

Clients are involved throughout. Baseline figures are agreed jointly. Exceptions and edge cases identified during parallel operation are discussed as they arise, not accumulated into a final report. The written comparison at the end reflects a shared understanding of what was tested, not a unilateral assessment by Kirameki of how things went.

Long-term view

What an engagement leaves behind

An engagement with Kirameki closes with written outputs that are owned by the client. A data dictionary is not useful only during the engagement — it is the reference document for anyone working with that dataset in the future. A governance policy does not expire when the project closes. A baseline figure is still relevant when the organisation is considering its next change to the same process.

The value of structured AI integration is not only in the immediate comparison it produces. It is in the institutional knowledge that accumulates from doing things in a documented way — and that remains in the organisation long after the engagement is complete.

For prospective clients

What this philosophy means for working with Kirameki

You will be asked to agree on what is being measured before work starts

This is not optional. If we cannot agree on what success looks like and how it will be measured before the pilot begins, the engagement cannot proceed. This protects both parties: it means the result of the comparison cannot be disputed on grounds that the wrong thing was measured.

You will receive written documents, not verbal summaries

Every engagement delivers written outputs: a comparison report, a data dictionary, or a governance policy — depending on the engagement. These are delivered in formats the client can open, edit, and keep without any dependency on Kirameki after the engagement closes.

You will not be pressured toward a particular outcome

Kirameki does not have a commercial interest in the client continuing with any particular tool or expanding scope after a pilot. The comparison report is written to inform a decision, not to make one. Discontinuing is presented in the decision framework as a reasonable and legitimate outcome.

Engagements are bounded and priced in advance

The scope of each engagement is defined before work starts, and the fee is fixed. There are no components that accumulate based on usage, scope drift, or additional deliverables not agreed at the start. If the scope changes substantially, that conversation happens explicitly — not through a bill at the end.

Next step

If this approach seems relevant to your situation

A short conversation can determine whether any of Kirameki's engagements are a practical fit — or whether they are not, which is also a useful thing to know early.

Get in touch