Work — AI Strategic Partner
A second chair at the decision table.
Adopting AI is not a question of tool selection. It is a chain of judgments.
A year-long advisory engagement for companies, held alongside leadership.
Abstract
Most AI consulting delivers a deck and leaves.
The usual shape of AI implementation support is a situational audit, a roadmap, a handful of use-case examples — delivered as artefacts, at which point the relationship ends. What the organisation is left with is the fact of "having adopted AI", and then — all through implementation — an endless series of moments of judgment that the roadmap did not contain. Each of those moments shapes the outcome.
AI Strategic Partner is designed against that shape. A year-long engagement that returns, repeatedly, to the places where the judgments actually happen. Initial audit, roadmap, implementation review, and finally the organisational absorption of the practice — the same person, at the same table, for a full year.
A second chair
at the decision-making table.
Philosophy
Quality rises only through repetition.
What more than a thousand prompt reviews taught this Lab is a very plain thing — nothing settles in a single pass. Write, run, take the criticism, rewrite. It is inside that loop, and only there, that a prompt becomes sharp.
Business decisions have the same shape. AI strategy is not set in the kickoff. It is technical debt that surfaces six months in, unexpected demands that come up from the field, new options that appear when a competitor moves — and the organisations that can actually implement AI are the ones with a standing capacity to update their judgment as those things arrive.
The annual contract is the form that makes this iteration contractual. It is not a spot consultation. The judgment six months from now, and the retrospective twelve months from now, involve the same person. Teaching becomes, in the end, research — the same structure that runs through Learn Premium runs through Work.
Practice areas
What the engagement covers.
Twenty-eight years of systems design, and more than a thousand prompt reviews: these are the two foundations. What they add up to is not a narrow prompt specialist but an architect of the product and the business around it. Four practice areas.
| AI adoption strategy Strategy | Work backwards from the business objective, look at the full operation, and identify where AI belongs. Not "where can we put AI" but which decisions should AI update. Designed across departments, not inside one. |
|---|---|
| AI product diagnostics & tuning Diagnostic | Existing AI implementations, prompts, or agent architectures that are not quite working — reviewed live, by running them. What is merely weak; what is genuinely fatal; in what order to fix them. Particularly useful for the "we shipped, but it is not delivering" situation. |
| Accelerating non-AI development Acceleration | Software that is not an AI product — made sharper and faster by using AI as a tool inside the engineering practice. Code generation, review automation, specification drafting, all designed to fit the team in front of us. Includes consultation on retrofitting AI into existing systems. |
| Process redesign Process | Cross-departmental workflow review, identifying where AI can compress or strengthen the process. Not point-tool adoption — rewriting the chain of work. Who decides what, where, when — reconsidered. |
Scope
What the annual engagement covers.
Not a one-off audit or a standalone roadmap — a year-long companion engagement. Four layers, traversed and re-traversed across the contract term.
| Strategic audit Audit | Business objectives, current AI usage, organisational structure, technical assets — all put on the same map. Not only what is missing, but also what should be stopped. Four to six weeks initially; re-run as needed. |
|---|---|
| Roadmap Roadmap | Against the audit, a one- to three-year plan. Not product-adoption-first — designed around the sequence of decisions. Reviewed each quarter and rewritten as reality comes in. The roadmap is treated as a living document. |
| Implementation review Implementation Review | What the team is actually running is reviewed by running it. Prompt and agent diagnostics for AI products; design work for embedding AI into non-AI products; review of AI acceleration across the existing codebase. All of it is run-and-surface audit. Monthly cadence; intensive when it needs to be. |
| Organisational absorption Knowledge Transfer | Personalised know-how turned into organisational form. prompt-as-code as an in-house standard, CriticChain's discipline as the posture for review culture. A combined contract with Learn Premium is available. |
Engagement
The shape of the contract.
| Contract form | Annual engagement as the default. Monthly retainers and spot consultations are not, as a rule, accepted. |
|---|---|
| Term | Twelve months per cycle, renewable by mutual agreement. Early termination is possible, but the design assumes short terms will not pay off. |
| Cadence | Monthly standing meeting (two hours) + implementation review (once or twice per month) + Slack/email as questions arise. Extraordinary sessions for major decisions or incidents, as needed. |
| Who is in the room | Hatanaka is in every session, directly. Work is not delegated to subcontractors or associates. This is the reason the number of concurrent clients is capped strictly. |
| Fees | Quoted individually. Varies with scope, expected outcome, and whether Learn Premium is bundled. Presented after the initial conversation. |
| Combination | A bundled engagement with Learn Premium (the human-reviewed learning program for teams) is available. Recommended for organisations that want strategy and learning run together. |
To prospective clients
A handful of companies each year, held deeply.
This engagement is designed to sit alongside leadership's decisions for a full year. Because Hatanaka is in every session personally, the number of companies that can be held in parallel is very small.
Because of that, a relationship that is patient enough to grow the quality of judgment over a year is more useful — for both sides — than a relationship aimed at quick wins. The five perspectives below are a thinking aid, not a qualification test.
Perspectives
Five perspectives.
| Intent | Is AI being held, inside the organisation, as a question of strategy — not as one more tactical project? Is there a willingness to update the shape of the organisation in response? |
|---|---|
| Time horizon | Beyond short-term ROI, is there space to invest in raising the quality of judgment on a three-year horizon? |
| Budget | Is there a budget appropriate to a year-long engagement, positioned as a strategic decision? Is this a serious contract conversation? |
| Leadership presence | Can a member of leadership attend the monthly standing meeting? When the decision-maker is in the room, the density of the hour rises sharply. |
| Transparency | We work best in a relationship where failures, debts, and internal friction can be discussed plainly. It is not necessary to bring only the success stories. |
First year
The map of the first twelve months.
None of this plays out exactly as written — reality rewrites the plan. This is the initial skeleton, revised each quarter.
Background
Twenty-eight years of systems design.
Takaho Hatanaka. Twenty-eight years as a systems architect — core enterprise systems, the web, the cloud, and now AI. In each era, the same work: translating between the language of business decisions and the language of technical requirements, on the floor, with real stakes. That back-and-forth is the foundation of this engagement.
Since 2023, more than a thousand prompt reviews have gone into a structured method for running LLMs inside real work. That research is what crystallised into the prompt-as-code specification and the adversarial review engine CriticChain. Research becomes the instrument of practice — that circuit is the structure of this Lab.
FAQ
Questions that come up.
| What is the current availability? | Always limited. Because Hatanaka is in every session personally, the number of concurrent annual clients is very small. Availability is confirmed inside the initial conversation. |
|---|---|
| Are spot consultations available? | As a rule, no. Quality rises only through repetition, and a one-off consultation does not fit the Lab's working hypothesis. |
| Can fees be quoted up front? | Fees are not listed publicly. They vary with scope, expected outcome, and whether Learn is combined, so they are quoted after the initial conversation. Requests for a quote before a conversation are declined. |
| How does this differ from Learn Premium? | Work is companionship of the decision; Learn Premium is the learning of the implementers. They are complementary — a combined contract is available, and recommended where strategy and implementation are not held apart. |
| Industry focus? | No particular industry specialisation. Industry-specific facts matter less than the structure of the judgment, which repeats. There are cases that cannot be accepted for ethical reasons; those are declined. |
| Only prompt-engineering topics? | Not at all. Twenty-eight years of systems design cover AI product diagnostics, the AI-driven acceleration of non-AI engineering, and full process redesign. See the practice areas section above for detail. |
| Geography? | Japan-based; online the rest of the world. Monthly standing meetings run hybrid as needed. |
| NDA? | Signed as a matter of course. The material surfaced during implementation review is sensitive; contract paperwork is built into the opening of the engagement. |
Start with a conversation.
The initial conversation is sixty minutes, at no cost. Current situation, interests, the shape of the decision-making group — we walk through those, and together decide whether there is a place this engagement would be useful inside your organisation. It is most useful for people looking for a year-long companion to the judgment, rather than a short-term tool rollout.
A note, before you write. Because the number of concurrent engagements is very small, not every inquiry can be translated into a contract. Hatanaka responds personally, which occasionally means a reply takes a little time. Please bear with us.