Advisory
Guidance on a decision, not a project
Advisory here covers three things: framing a use case, reviewing an architecture and agreeing the next useful piece of work. It is deliberately bounded. There is no implementation commitment, no delivery package and no open-ended scope attached to it.
A scoping conversation agrees what is in and what is out before anything else.
Who this is for
You own the decision, and the review is thin
Three moments bring people here. A team is about to give an agent authority to act and nobody in the room has argued the boundary before. A design already exists and needs a challenge from someone with no stake in it. Or there are six candidate use cases and the first one has to be chosen for reasons that hold up later.
It applies when the decision is yours to make and stays yours. If what you actually need is to build the capability across a group, that is a team programme, not advisory.
What it covers
Three things, and the list ends there
Framing a use case
What the system is for, where it stops, and whether the work needs an agent at all. Often the useful outcome is that a fixed workflow was the right answer and the question was never about the model.
Reviewing an architecture
A design you already have, examined against the decisions it has to answer: evidence before action, tool authority, uncertainty, repeated actions, and what you can see afterwards.
Next-step guidance
Which piece of work comes next and why. Perhaps a test. Perhaps a narrower permission. Perhaps a clearer explanation of the trade-off to the people funding it.
A worked example
What a review actually asks
A returns assistant reads a policy and recommends a refund. The team wants to connect a tool that issues it. On the diagram that is one more box; in production the system can move money. Illustrative, and no customer incident is claimed — but the questions are the ones a real review works through, in this order.
- 01
Does this need an agent at all?
If the steps can be written down before the request arrives, a workflow is cheaper to run, easier to test and easier to defend.
- 02
What evidence has to hold before the action?
Named, listed and checked by the tool itself — not asserted in a prompt and hoped for.
- 03
What is each tool allowed to do?
One job per tool, the narrowest scope that does that job, limits the tool enforces rather than limits it is told about.
- 04
What happens when the evidence is uncertain?
Three outcomes rather than two: act, refuse, and hand over with the specific gap named. A confidence threshold is not a plan.
- 05
Can the same action happen twice?
A timeout with an unknown outcome is the ordinary case, not the edge case. The system needs a way to examine the state before repeating anything.
- 06
What can you see afterwards?
Which evidence was used, what each tool did, and whether the run stayed inside its boundaries. Without that, an evaluation cannot point at a design change.
Those six are written out in full, with the reasoning, in the published guides. Reading them costs nothing and is the fastest way to tell whether this way of working suits your team.
The boundary
What advisory is not
It is not delivery. No implementation is committed, guaranteed or included. If a review leads to work you want someone to build, that is a separate conversation with a separate scope, and it does not begin by default.
It is not a package. No tiers, no day rate and no minimum term are published, because none has been agreed. Anyone quoting you one from this page would be quoting something that does not exist yet.
It is not the fractional CAIO retainer. That is a longer, board-facing engagement with its own scope, described on its own page. A fixed-scope diagnostic with a board-ready roadmap is the AI readiness assessment. Both are separate from what this page describes.
No response time is promised. A service target is under discussion and will be published only once it has been agreed — not before.
Next step
Start with the decision in front of you
Write to apply@thelivingcraft.ai with the decision you are facing, what you have already tried and who else has to be convinced. That is enough for a first conversation, and it is enough to say honestly whether this is the right thing at all.
Please do not attach confidential architecture or company documents to a first enquiry. What material may be shared, and under what terms, is part of what a scoping conversation settles.
Written . Service facts pending Sunil’s confirmation before publication.