Twenty more Digital FTEs at every stage of readiness — from field-ready products one push from flagship, to early experiments, to the two platform capabilities every unit shares. Scored honestly on AI-native and FTE-fit; we tell you what runs and what's still in the workshop.
An A* Cambridge examiner-tutor that marks against real mark schemes and coaches 24/7. Six LLM agent-roles, pgvector RAG.
A rupee-accurate double-entry kernel with forced Postgres RLS and an immutable journal — the ideal substrate for autonomous finance agents.
Always-on AI that reads, sentiment-scores and drafts replies to every review and mention. Hardened LLM core; one channel live, four scaffolded.
Explainable phishing triage with cost-gated LLM escalation — cheap ML runs first, the model only when it's borderline. Real, runnable code.
CVs parsed, scored and ranked 24/7 before you wake up — an already-autonomous intake loop with human-in-the-loop approval. LLM-central.
An always-on voice tele-sales rep that quotes, qualifies and captures leads. Working voice loop today; the conversation brain is the build.
Brief in, branded editable PowerPoint out by morning — a real engine that emits native decks, not screenshots. The agent wrapper is next.
A legacy MS-Access perfume ERP, reimagined as an AI-run bilingual back-office. A deep planning dossier today — the build is the work ahead.
Plain-language need in; a sourced, ranked PKR top-5 out — across Pakistani retailers, as web UI, JSON, Excel and Markdown. Deterministic today.
Turn a prompt template into a self-running AI worker. A solid full-stack base (35 templates, 11-provider gateway) waiting for an agent layer.
Block a stolen card by voice — Urdu or English — 24/7. Real Whisper + TTS loop; the intent and banking backend are the upgrade.
Type a card BIN, staff instantly see the right bank discount. A deployed webapp whose manual quarterly data ingestion an agent could own.
Autonomous AI authors the chapters; a 24/7 AI tutor teaches the reader. A v1 shipped 3 live chapters; v2 is scaffolding being rebuilt.
An A-Level marking agent grounded in real Cambridge mark schemes. The content moat (past papers + schemes) is the asset; the AI is still to wire.
A Gmail screener for phishing, spam and spoofing. Rules-only today with a working OAuth intake — a clean base for a real AI email-security FTE.
An always-on LinkedIn ghostwriter: on-brand posts, on schedule, while you sleep. A thin skill today with an obvious content-agent expansion.
An AI co-pilot that filters every Gold signal and shows its work. A real MT4→XGBoost forward-test with unusually honest self-correcting analysis.
An AI risk officer that enforces a trading rulebook 24/7 — tiers, loss caps, audit trails. Disciplined ML governance is the real differentiator.
71 vetted capabilities — browser, docs, comms, extraction — the shared library that gives every Digital FTE real hands. The capability moat.
The WhatsApp channel our autonomous FTEs read and reply through 24/7 — a hardened bridge already driving live agents across the portfolio.
These twenty are the ones we've already started. Tell us the seat that's bleeding hours and we'll scope the agent that owns it.