Comparison · Approaches · Honest Assessment
Different approaches to AI in business — and what each actually delivers
There are at least three ways a business can introduce AI into its processes. Each has genuine uses and genuine limitations. This page sets them out without favouring one by default.
Back to homeThree paths, different outcomes
Most businesses considering AI end up choosing between off-the-shelf software, conventional technology consulting, and a scoped integration approach. The choice depends on what you actually need — not on which option has the stronger marketing.
Off-the-shelf AI tools
Subscription products applied directly to business tasks without customisation.
Scoped integration approach
Analysis-first engagement with a defined scope and documented handover.
Conventional consulting
Large-scale technology projects managed by external teams over months or years.
Monthly subscription. Appears low initially but accumulates over time, and savings depend on staff actually using it consistently.
Fixed fee per engagement: ¥30,000–¥41,000. You know the total at the start. No hourly billing or scope extensions.
Project-based, often variable. Costs can extend as scope is discovered. Typically better suited to larger budgets.
Immediate access. But meaningful integration with existing workflows takes weeks of internal effort regardless.
Four to seven weeks depending on service. Fixed endpoint. The work is done when the engagement closes.
Months to years. More appropriate when the problem itself is not yet well-defined.
Generic. The tool is built for broad applicability, not your specific enquiry mix or document structure.
Built from your own data — past contacts, existing documents, current style guides. The output reflects how your organisation actually works.
High customisation potential but requires sustained collaboration and longer discovery phases.
Vendor documentation for the product. No record specific to your deployment or how it was configured.
Every engagement ends with documentation you keep: what was found, what was built, what it does not handle, and how to withdraw it.
Typically extensive, but documentation scope varies and may not cover operational specifics once staff turnover occurs.
The layer view — making the change visible
One element of our approach that clients find useful is a static diagram of each process: what exists now, what the integration adds, and where human review sits. Nothing is removed without being named.
Existing process
We document what already happens — how your team handles the task today. This is the foundation. Nothing here changes without agreement.
What the integration adds
The specific function being introduced — search, drafting, or categorisation — shown as an addition, not a replacement.
Human oversight arrangement
Where a person reviews, approves, or can intervene. Documented before deployment, with a clear withdrawal path included.
Realistic results, not projections
The outcomes of any AI deployment depend on the condition of your data, the consistency of your processes, and whether staff use the system as intended. We prefer to describe what we measure rather than what we project.
A Off-the-shelf tools — realistic picture
-
Works well for individual productivity tasks. Effectiveness drops when applied to processes requiring institutional context.
-
Requires staff to adapt their habits to the tool. Adoption rates vary significantly between teams.
-
No documentation of what changed. Difficult to assess impact or reverse if needed.
B Scoped integration — what we deliver
-
Evaluation set built with your staff. Results compared against real queries before deployment, not after.
-
We document what the system handles poorly — this is part of the deliverable, not an omission.
-
Quality sampling included in the first month. You have a written record of how the system performs in practice.
What you receive for the fee
¥31,000
Customer Service Assessment
Categorised view of enquiry volume. Proportion estimate for automated handling. A recommendation that may conclude no change is needed.
¥41,000
Documentation Search
Permissions reviewed and applied. Evaluation set built with staff. Written note of question types the system answers poorly.
¥30,000
Content Production Workflow
Encoded style reference. Category list of material requiring human authorship. Quality sampling across the first month.
What the engagement actually looks like
With a large consulting firm
- Extended discovery phase, often several months before any build
- Multiple stakeholders, frequent handoffs between teams
- Scope changes common as the project evolves
- Budget revisions not uncommon
- Appropriate for complex, organisation-wide change
With Marsh Crest Field
- One named contact throughout
- Scope written and agreed before week two
- Regular written updates, no surprise discoveries
- Fixed cost from the first conversation
- Appropriate for a specific, well-bounded process
What stays after the engagement closes
The written record
Documentation covering findings, configuration, limitations, and withdrawal procedure. Useful if staff change or a review is needed later.
Staff understanding
Training happens with the people who will actually use the system. They know what it does and what falls outside its scope.
A withdrawal path
Every deployment includes documentation on how to stop using it. No dependency on our continued involvement to exit.
Things worth clarifying
Often said
"AI will handle this automatically."
Automation handles the repetitive, well-defined portion. Judgement, exceptions, and edge cases still require a person.
Often assumed
"More AI means lower headcount."
Our services are scoped around process improvement, not workforce reduction. These are separate questions with separate implications.
Common expectation
"The result will be visible immediately."
Impact becomes visible over weeks, particularly for document search and content workflows, as staff develop familiarity with the system.
Sometimes believed
"AI is always the right answer."
Our assessment engagements may conclude that a process does not suit automation. That finding has the same value as a deployment recommendation.
When a scoped approach makes sense
A scoped integration engagement is worth considering when you have a specific process in mind, want to understand the impact before committing to a larger change, and prefer written documentation over verbal assurance.
You have a specific process — customer enquiries, internal search, or content production — rather than a broad initiative.
You want to know whether automation suits your situation before making a larger commitment.
You need documentation for internal governance, future reference, or simply to understand what you are running.
Have questions about which approach suits your situation?
We are happy to talk through what you are working on without any obligation to engage us. Write to us and describe the process you have in mind.
Get in touch