Bold Group
Part of the Transformation Blueprint
You invested in AI. Where is the business result?
AI pays only when the work around it changes. We fix the work first, then decide where AI should scale it, or whether it should be used at all.
A 14-day Constraint Assessment on one process.
The same first step as the Transformation Blueprint, with AI in scope.
- You get
- The result and its baseline, where the work really breaks down, and one of five decisions for your use case.
- It asks of your team
- An executive owner and a process owner, a 1 to 1.5-hour interview with each key person, the documents that run the work, and a site visit.
- Then you decide
- Whether to go further, with us or with your own team.
- Cost
- Fixed scope, quoted before we start.
Does your AI pilot need this?
- Can you name the business result it should change?
- Is there a baseline for that result?
- Is one person accountable for the number?
- Has the workflow around the tool changed?
If any answer is no, start here.
How the work runs
Five steps, starting from the result, not the tool.
It begins with one process where speed, cost, capacity, quality, risk or customer value would matter. The first question: if AI works here, what result should change, and what else must change for it to happen?
- 01
Define the result
The business result, its baseline, who owns the number, and why act now.
- 02
Diagnose the current work
Workflow, decisions, roles, data, exceptions, delays and the human knowledge involved.
- 03
Decide the right role for AI
Where it may assist, recommend, generate, classify or automate, and where it should not be used.
- 04
Redesign the work around it
The future workflow, where people review, who decides, controls, adoption and vendor requirements.
- 05
Decision package for your board
The use case, how the result will be counted, risk controls, a validation plan, and a brief your technical provider can quote against.
The decision
The outcome is not “use AI”. It is one of five decisions.
- 01Proceed with a defined AI-enabled workflow
- 02Validate a critical assumption first
- 03Change the process before introducing AI
- 04Choose a different technical or non-technical intervention
- 05Stop the use case, because the value, risk or conditions are not there
Stopping a weak use case is also a result, when it prevents an investment that would not have paid.
Appropriate technology
Fix the work first. Then choose the technology.
Good design, scaled, multiplies value. A poor process, scaled, multiplies the problem.
Often enough on its own.
- Remove steps that add nothing
- Make responsibility clear
- Improve the standard forms
- Standardise the process
Where the fix is proved.
- Connect data you already hold
- Automate repeated work
- Bring in AI support
- Build or buy a new system, if still needed
See how service needs map to digital tools
| Service need | Digital tool | What it helps scale |
|---|---|---|
| Check completeness faster | Validation | First-time-right at higher volume |
| Search and compare information | Search support | Access to evidence |
| Sequence and route work | Workflow | Volume and coordination |
| Report progress | Notification | Communication with many users |
| Produce standard documents | Template automation | Consistent quality |
| See the queue and performance | Dashboard | Learning and service improvement |
| Support analysis | Decision support | Expert capability, not decision authority |
How we diagnose
A public proposal to DIP: the backlog was in the process, not the technology.
Our proposal to Thailand’s Department of Intellectual Property (DIP) for its trademark-registration work, where applications were arriving faster than they were registered. Implementation is DIP’s decision.
- The steps assumed applicants do not understand the law or the process.
- Incomplete applications were let in, then corrected inside the system.
- Several steps duplicated one another.
- Every step through e-filing, with online payment and an applicant self-check at filing.
- Duplicated steps removed.
- Only then, AI support: image and sound-alike search, so examiners compare similar marks faster.
As the work ran
- FilingIncomplete applications let in
- CorrectionFixed inside the system, back and forth
- Duplicated stepsThe same work done more than once
- ExaminationExaminers compare similar marks
- BacklogApplications arrive faster than they are registered
As proposed
- 1 · Fix the worke-FilingOnline payment and an applicant self-check at filing
- 1 · Fix the workDuplicates removedEach step done once
- 2 · Then AIAI supportImage and sound-alike search for examiners
- RegistrationImplementation is DIP’s decision
Who does what
We design the result. Others build the technology.
Frames the result and baseline with you, redesigns the work, challenges premature technology choices, defines roles, controls and measures, and writes the blueprint. Supports vendor evaluation when contracted.
Appoints an executive owner and a process owner, makes risk, policy and investment decisions, brings in legal, privacy, security and data functions, selects technical providers, and owns adoption and results.
Model and platform, data engineering and integration, software, testing, infrastructure, security and support, under their own contract.
What we do not do
Where legal, privacy, security, data or engineering review is needed, you appoint qualified specialists.
- Sell AI licences or tokens
- Train or fine-tune models
- Build production software or integrations
- Certify that an AI system is compliant, private, secure or unbiased
- Guarantee model outputs or financial results
- Force AI into a process where something else fits better
Where we stand
We do not yet present a completed AI implementation case. What we offer is how we redesign any process: start from the result, question the assumed solution, fix the work, then validate the value. Our closest published example: a finance team was scoping an app to capture expense records that its bank already produced for card payments. The evidence on why AI projects stall is in field note 08.
We already have AI pilots and vendor proposals. What happens to them?
Bring them. We test one against the five decisions: proceed, validate first, fix the process, choose something else, or stop. We do not replace your vendors.
Are you neutral on vendors?
Yes. We do not sell AI licences, models or platforms, and take no commissions from AI or software vendors. The requirements we write are yours, and any qualified provider can quote against them.
Does every engagement end in an AI project?
No. The work may show that the process should change first, another technology fits better, or the use case should stop.
Can you help us choose a platform?
We can define requirements, evaluation criteria and the business case, and support vendor evaluation when contracted. Final technical due diligence stays with your technical, security, data, legal and procurement functions.
How is the result counted?
The baseline, target, data source, the person who owns the number and the review period are agreed before the build. Projected value is kept apart from implemented and measured results.
Start here
Bring one AI use case. Leave knowing what it should change.
Bring the business result, the workflow that produces it, what has been tried, and who owns the decision.
Teerasak WongpiyaCo-Founder and Managing Director
1 hour with a founder, in Thai or English, online or at your office.