Flagship engagement
A bounded four-to-six week engagement that gives leadership a shared picture before another tool is bought. You receive a written note, a shortlist, and a session to argue about the sequence in the same room.
Field notes before the findings pack
Purpose
A pilot spent on the wrong process teaches the organisation that AI is theatre. A review is cheaper than that lesson. It is also slower than buying another subscription, which is the point.
Mid-sized firms in Singapore often already have several assistants in use. The board then asks for a strategy, or counsel asks what personal data is leaving the building. Neither question is answered by adding a sixth tool. The review gathers the facts that those conversations need: which processes have volume, which files are fit to search, who would own a live system, and what the Personal Data Protection Act (PDPA) implies for the data you would send outside.
We write the findings so an operations lead and a director can both use them. Recommendations are sequenced over twelve months. We say what to delay. We also say when we would not start, including when the documents are too scattered or when nobody is available to own the result. Established in Singapore in 2026, the practice designed this review for organisations of about fifty to five hundred people, where a few interviews still reach the people who do the work.
Five areas
Each area produces a short chapter in the findings note. We do not score you with a vanity index. We describe what we saw and what it implies for the next quarter.
Where the files live, how they are named, who may open them, and whether a search would retrieve the right version. Retrieval-augmented generation, which means searching your corpus before a model answers, fails when this layer is weak. We look at a sample, not at an abstract architecture diagram.
Which steps are repeated often enough to matter, where a draft would save time, and where a person must still decide. Volume without a stable process is a poor place to start. We ask for numbers the team already trusts rather than inventing a baseline.
Who would run a live assistant on a Tuesday morning when the vendor is quiet. If the answer is “the innovation committee”, we treat that as a finding. Ownership needs a named role close to the process, with time in the diary after the launch week.
What personal data would move, to which service, in which region, and on what retention clock. We help you pose those questions clearly for counsel. Guidance on this site does not constitute legal advice. Sector overlays such as financial or clinical rules are flagged so you can take them to the right specialist.
Licence lines, usage fees, staff time to review drafts, and the cost of keeping documents in a searchable state. A cheap chat seat can become expensive once retrieval, logging, and human review are counted. We put those lines next to the process you hope to change.
Calendar
The plan flexes with your diary. The sequence of work does not. We always start with people and files, and we always end with a written note and a session.
Kick-off with the sponsor, confirm the processes in scope, and collect access to three to five people who carry those processes. We agree the sample of documents and the list of current subscriptions. A one-page plan goes back the same week.
Interviews and a walk-through of the files. If a tool is already in use, we watch it on your material. Weekly notes keep the sponsor current. We raise early if a process is out of scope or if the sample is too thin to support a finding.
Draft of the findings note, the shortlist, the risk register, and the twelve-month sequence. You see a working draft before the session so the meeting is about decisions, not about discovering the document.
A ninety-minute working session. We walk the sequence, record the decisions in the room, and issue the final note. Fieldwork copies of sample files are deleted unless you ask us in writing to retain a defined set.
Inputs and outputs
From you we need access to three to five people who carry the processes in scope, a sample of the documents those processes use, and a list of current AI-related subscriptions. We do not need a data warehouse tour. We do need someone who can explain a folder the way a new colleague would hear it. If an NDA is required before samples move, we sign it.
On the way out you hold a findings note of about fifteen to twenty-five pages, a prioritised shortlist of scenarios, a risk register, a twelve-month sequence, and the record of a ninety-minute working session. The note is written to be forwarded. It does not depend on us remaining in the room.
Fit
If you already have a funded pilot, a named owner, and a clean path through the documents, a review will feel like delay. Ask for a supervised pilot or a vendor assessment instead. If the organisation cannot spare three people for interviews in the next month, wait. If the only question is legal interpretation of the PDPA, speak with qualified counsel first; we can join afterwards to map the tools. If you need production software written, this practice is the wrong place to spend the quarter.
A review is timely when several tools are already in use without a rule, when the board has asked for a strategy paper, or when a demo has failed on real files and nobody agrees why.
Short answers
Can we skip the review and start a pilot next week?
You can. We will still ask who owns it, which files it will read, and how you will know it worked. If those answers are already written, a review adds little. If they are not, a short review is usually cheaper than a stalled pilot.
Will you need production data?
We prefer a small sample with personal data removed where you can do so. If a sample must include personal data, we agree the handling in the scoping note and minimise what we hold.
Is the twelve-month sequence a project plan we must follow?
It is a recommended order with reasons. You may rearrange it. The point is to make the trade-offs visible, including the work you should not fund yet.
Next
If a readiness review sounds like the right spend, send a short note. We reply within one working day, Monday to Friday, and we will tell you if the review looks premature.
