Skip to content
The AI Platform

Enterprise AI, built for trading businesses.

Not a chatbot. A complete AI infrastructure — RAG, prompt orchestration, contextual intelligence and workflow automation — wired into every module of the BSCT.Cloud platform.

Grounded retrieval Role-scoped memory Auditable actions Model-agnostic
AICORE
Terminal
Strategy
KB
Academy
Analytics
CRM
WL
Automation
AI Vision

Practical AI. Not hype.

Five principles that shape every AI capability we ship. Read them as our contract with you — what we’ll do, and what we won’t.

AI should assist — not replace.

Every AI feature we ship starts with a human in the loop. Assistance amplifies expertise; automation only fires with explicit trust.

AI should understand context.

Ungrounded AI is a party trick. Ours reads the module, ticker, portfolio, role and history — and responds accordingly.

AI should learn from your business.

Your playbooks, retros, decks and code become the AI’s memory — role-scoped, versioned and auditable.

AI should reduce manual work.

From onboarding to reporting, AI removes the mechanical steps so operators focus on judgement calls.

AI should improve decisions — with citations.

Every claim cites its source. Confidence is scored, not asserted. If the model isn’t sure, it says so.

AI Architecture

How every AI request actually flows.

Eight discrete stages between the operator asking and the grounded answer landing. Each stage has a single responsibility, is separately governable, and is separately upgradable.

Semantic retrieval over your role-scoped documents and chunks. Freshness policies keep it current.

AI Mentor

Your senior analyst — on demand.

AI Mentor is the conversational surface that stitches every module together. It reads the current screen, retrieves the right context, and answers with citations — not with confident hallucinations.

Context-aware conversations

The mentor sees the same screen your operator sees. No copy-paste, no lost thread.

Guided workflows

Ask “walk me through launching my white-label” and follow a step-by-step trail with checkpoints.

Suggested prompt starters

Live suggestions derived from your KB — not generic template questions.

Business-specific answers

Grounded in your playbooks, your desk retros and your policies — not the open internet.

Follow-up recommendations

Every answer ends with next-step chips — open a chart, run a scan, invite a teammate.

Multi-step reasoning

For complex questions, the mentor plans, retrieves, computes and cites in one coherent trail.

app.bsct.cloud/mentor
live
AI Mentor · sessionmodule=strategy · ticker=MES
Why is my MES strategy underperforming this week?
Regime shifted to low-vol range-bound on Wed. Your breakout entries are firing 3× more but win-rate dropped from 58% → 41%.
2 citations · 1 chart · 1 action
Ask the mentor anything…
Knowledge Base · RAG

Your knowledge, grounded in retrieval.

BSCT.Cloud’s RAG layer ingests your documents, embeds them, and retrieves role-scoped chunks at query time. Every answer cites the source it came from — auditable, testable and correctable.

Ingested sources
Documentation

Auto-generated platform docs kept fresh by change detection.

Platform guides

Best-practice guides for every module — written by our team and yours.

Trading knowledge

Curated market microstructure, strategy patterns and risk primers.

Academy content

Every lesson, quiz answer and worked example indexed for retrieval.

Internal business docs

Your playbooks, retros, decks, sheets and code — scoped by role, encrypted at rest.

Every chunk carries provenance metadata — source, version, timestamp, ACL scope. Freshness policies re-embed on change; audit logs record every access.
Retrieval flow

From your library to a grounded answer.

01
Query is embedded
The user’s question is turned into a dense vector using text-embedding-3-small.
~40ms
02
Vector search
Top-k semantic matches are pulled from the tenant’s vector store, filtered by ACL + freshness.
k = 6 · Pinecone
03
Rerank + condense
A cross-encoder rerank pass picks the most relevant chunks. Long chunks are condensed to fit the prompt budget.
rerank + de-dup
04
Grounded prompt + answer
Chunks are stitched into the prompt with citations. Model responds; citations are returned inline.
gpt-4o · grounded=true
Every answer includes citations — no hallucinations, no anonymous claims.
Prompt Intelligence

Every prompt is a reproducible artefact.

Prompt orchestration is the boring part of AI — versioning, context injection, tool schemas, memory. We take it seriously so your AI stays predictable, testable and safe.

Prompt orchestration

A registry that composes system + user + context + tool schemas into a single, versioned call.

Prompt templates

Per-use-case templates — strategy analysis, portfolio review, tutor mode — all editable, all diff’d.

Context injection

The context manager slots current module + ticker + role into the prompt automatically.

Conversation memory

Multi-turn context with configurable retention — short-term working memory, long-term semantic recall.

Structured outputs

When the answer must be JSON, an action, or a form fill — outputs are validated against schemas.

Multi-step prompting

Chain-of-thought + tool calls + follow-up prompts for questions that need planning.

Prompt anatomy · mentor-v14
system
You are the BSCT.Cloud mentor. Cite sources. Never invent modules.
tools
open_module, run_scan, backtest_strategy, invite_teammate
context
module=strategy, ticker=MES, role=trader, portfolio=live
retrieved
[6 chunks] regime-shift-guide · kelly-sizing · mes-microstructure …
user
Why is my MES strategy underperforming this week?
Every field is versioned. A/B tested. Cost-tracked.
Workflow Automation

AI that does the boring work.

Every automation is opt-in, guarded, and reversible. Humans stay in the loop for high-consequence actions; low-consequence work runs quietly in the background.

Onboarding guidance

New operators are walked through their first week with adaptive checklists and role-specific hints.

Learning recommendations

The mentor recommends Academy lessons based on gaps observed in the operator’s workflow.

Trading assistance

Setup detection, sizing sanity-checks and risk warnings surface as inline suggestions.

Support workflows

Tier-1 tickets auto-triaged, drafted and (with approval) resolved by the AI — humans handle exceptions.

Business operations

Renewal outreach, licence expiries, expansion signals — the AI schedules them; the operator approves.

Notification generation

Cross-channel notifications are drafted contextually — tone matched to the audience and event.

AI Across the Platform

Not a chatbot bolted on. A capability woven in.

Hover any module to see how AI shows up there. Every capability inherits the same context manager, prompt engine and KB — no duplicated pipelines, no inconsistent behaviour.

AICORE
AI Mentor
Trading Terminal
Strategy Builder
Backtesting
Scanner
Trading Academy
Analytics
CRM
White-Label
Enterprise AI

AI that passes procurement.

Every AI capability inherits the platform’s security, tenancy, RBAC and audit posture. Nothing is bolted on — the AI is part of the platform, not adjacent to it.

Context awareness

Every request carries workspace + module + role — responses honour all three.

Role-aware responses

A trader’s question yields a trader’s answer. Admin questions never leak into operator conversations.

Secure access

Row-level scoping, encrypted transports, audit logs — the AI inherits the platform’s RBAC.

Modular architecture

Prompt engine, retriever, context manager and tool layer are separately upgradable.

Configurable knowledge

Choose which sources feed the KB per tenant, per role, per freshness policy.

Future extensibility

Bring your own model, retriever or evaluator. The layer is pluggable, not vendor-locked.

AI Use Cases

Real scenarios. Real outcomes.

Seven ways operators, analysts and admins use BSCT.Cloud’s AI every day. Each maps to a module you’ll actually use.

Answer platform questions

AI Mentor

“How do I add a new module to my tenant?” → grounded walk-through from the admin console guides.

Explain trading concepts

Academy

“What is a walk-forward optimisation and why does it matter?” → pulled from the Academy KB with cited examples.

Guide onboarding

Customer Journey

New tenant admins get an interactive tour of the launch sequence — branding, domain, modules, launch.

Recommend Academy courses

Academy

Based on observed workflow gaps, the mentor suggests the next lesson to unlock.

Assist strategy creation

Strategy Builder

“I want to fade breakouts in low-vol regimes” → proposes signal / filter / risk blocks + a validated backtest.

Summarise analytics

Analytics

“What drove MRR growth this quarter?” → the mentor decomposes into cohorts, module usage and pricing changes.

Help administrators manage workspaces

Admin Console

Bulk seat provisioning, module toggles, role changes — explained and (with approval) executed.

Frequently asked

The questions most buyers ask about AI.

If yours isn’t here, our team can answer in a 30-minute call.

Only within your tenant. Retrieval is role-scoped, transport is TLS 1.3, storage is AES-256 at rest, and every high-consequence action is audit-logged. You keep ownership of your knowledge base and can export or purge at any time.

Ready to see it in action

Empower every user with intelligent assistance.

BSCT.Cloud’s AI meets your team where they work — in the terminal, in the strategy builder, in the academy, in the admin console. All grounded. All auditable. All included.

Model-agnostic·Grounded retrieval·Role-scoped memory·Auditable actions