How Truth works: from request to reviewable work
A plain-English walkthrough of how Truth turns a Bahamian financial-services request into skills, agents, source context, drafts, diagrams, documents, and reviewable work.
In plain English
- Truth is the infrastructure around the work: intake, route selection, skills, agents, source context, documents, diagrams, and review.
- A skill gives the system a route. Agents do defined preparation jobs inside that route.
- The useful output is not just text. It is work that shows what is known, what is missing, and what a professional still needs to decide.
5 min read
Truth starts with a request, not with a blank prompt. The product helps turn that request into structured work a legal or financial-services team can understand, improve, and review.
A fund launch, trust structure, digital-asset question, or global counsel request should become a route, a source plan, assigned agents, draft outputs, document tasks, and a visible review path.
1. Start with the real request
A user should not need to start with a perfect prompt. They should be able to describe the real work: launch a fund, prepare a trust structure, review a digital-asset route, explain a family office structure, organize a cross-border counsel pack, or understand what a Bahamas workflow requires.
Truth’s first job is to turn that plain-English request into a useful starting shape. What kind of work is this? Which Bahamas route might apply? What facts are missing? What sources should be checked? What should an agent prepare first?
2. Identify the route
The route is the difference between generic AI output and useful preparation. A trust, foundation, Professional Fund, SMART Fund, ICON, FCSP question, digital-assets activity, holding structure, or global counsel request each needs a different path.
Truth should make that path visible early. The user should see whether the work looks like a source-checking task, an instruction pack, a structure explanation, a document plan, or a review queue before anyone treats the answer as reliable.
3. Use a skill to set up the work
A skill is a reusable way to prepare a type of work. It tells the system what to ask, what to collect, what sources to check, what documents may be needed, what diagrams may help, and what questions should stay open for professional review.
This is why skills matter. A good skill prevents the AI from treating every request like a blank page. It gives the agent a practical job and gives the reviewer a clearer basis for checking the output.
4. Assign agents to defined jobs
Agents are useful when their roles are narrow. A source agent collects authorities and regulator context. A drafting agent prepares memo sections and document outlines. A structure agent maps entities, assets, beneficiaries, and control paths. A Desk agent tracks missing evidence, approvals, and follow-up.
The user should understand what each agent is doing. That is the difference between a black box and infrastructure. The agent is not there to make the final call. It is there to prepare a piece of work that can be inspected.
5. Keep review visible
The review queue is where the product stays honest. Missing documents, source gaps, unclear facts, administrator questions, banking follow-up, tax or foreign-counsel issues, and final legal decisions should not be buried inside a paragraph.
They should become visible tasks. That is what makes the work useful: the system shows what has been prepared, what still needs support, and what a professional still needs to decide.
What this should feel like
For a new user, Truth should feel like a guide through the work. It should explain the route, show the pieces, and make the next step obvious. For an experienced professional, it should feel like a better preparation layer: fewer loose threads, clearer source context, and a cleaner handoff.
That is the product standard. Not more AI text. Better legal AI infrastructure for Bahamian financial services.