Demo only — scripted, not a live employee, no real work sent or charged. Ora is an illustration of the approval-first idea, not a person working in real time.
Create account — go to dashboard Go to dashboard Sign in
Dashboard link goes to the production dashboard at app.crewora.net. Sign in with your existing account and you land in your workspace dashboard — employees, budgets, approval queue, and credit ledger. No live money, no live Stripe.
CrewOra is the approval-first AI workforce workspace for small and medium businesses: named SI Agents — AI employees who arrive as ready-to-use work packs, work inside one workspace, and cannot publish, send, call, or spend without a human saying yes.
Every SI Agent gets its own credit budget and approval queue. Every pack ends with a 30-day report on what was actually delivered — videos published, bookings kept, research completed — not a chat transcript. That combination is the gap in today's market, and it's what CrewOra is being built to own.
Free tier is draft-only: no publishing, no live calls, no outbound — by design. External actions unlock on paid tiers, and still need your approval.
Maya wants to publish “5 Pricing Mistakes Costing You Clients” — uploaded as Private first. Script claims checked, thumbnail and disclosure flags attached.
of agentic AI projects are forecast to be cancelled by end-2027 — for escalating costs, unclear value, or inadequate risk controls.
Gartner forecast, as cited in market research read 1–2 Nov 2026. CrewOra's answer: hard budget caps, approval gates, and a 30-day outcome report.
of thousands of vendors claiming agentic AI are estimated to offer genuine agentic capability. The rest is “agent washing” — chatbots and RPA rebranded.
Gartner estimate, same research. CrewOra's answer: show receipts — sample approval logs, credit ledgers, and reports, before the pitch.
Three flagship packs specified in full — YouTube, Trading Research (simulation only), and AI Call Centre — plus a seven-pack catalogue in design across support, sales, content, e-commerce, admin, and hiring.
The three flagships were the starting examples, not the whole library. One flagship is proven first in private beta; catalogue packs follow in order of beta demand. See the catalogue below.
Ora is a demo SI Agent — an AI employee, in the term CrewOra started with. Demo only — scripted, not a live employee, no real work sent or charged. Ora proposes, you decide, and the result stays draft-only until you say yes.
Try it in order: Draft → Approve or Deny. Approving here only updates this demo panel — it does not publish anything, anywhere.
3 items waiting · Example items only.
“5 Pricing Mistakes” — Private draft, disclosure set. Example item.
Tue 10:30am, read-back confirmed in script. Example item — no real booking.
Simulation only. Example: 2% simulated risk, journal entry attached. No real money.
Example / planning estimate only — not customers, revenue, or live results.
How to read it: the report judges delivered, approved work and cost per outcome — not activity, chats, or promises.
Example flow only — planning estimate, not a live result.
Choose a ready-to-use work pack. Its SI Agents arrive with their roles already set — Researcher, Maker, Checker, Publisher — each with its own instructions and tool allowlist.
Set a credit cap per employee per week. You see the estimated cost before a job runs and the actual cost after. At zero, the employee drops to draft-only — it never silently spends.
Drafts are free and unlimited to review. Publishing, sending, calling, booking, or spending needs a named human approval: approve, edit, reject, or schedule — with the reason logged.
Every pack reports outcomes and cost-per-outcome: credits per published video, per verified booking, per research report. If it wasn't worth it, the report says so plainly.
The SI Agent market — the market long called the AI agent market — splits into four camps. Each is missing something a business owner actually needs — and the missing pieces are exactly the ones CrewOra combines.
Powerful agent builders and automation tools — but you build the team yourself. No ready-to-run packs, and meters stack up (actions + model credits, seats + credit packs).
Voice platforms, support agents, outbound SDR tools: deep in one job, silent on everything else. A business ends up with five subscriptions and no single place to manage the team.
General AI assistants and open-ended agents: strong task completion, but no employee identity, no per-employee budget, no approval queue, and unpredictable per-task billing.
Named “AI helpers” you chat with, sharing one credit pool: a team feeling, but not a managed team doing end-to-end jobs — no approval workflow and no outcome reporting.
Each pack is staffed by SI Agents — AI employees working from pre-set roles, standard checklists, approval tables, and a Day-30 scorecard — not bare prompt templates. The three flagships below carry full specifications from the market research. Integrations are listed as “via” and are only claimed live once the connection is tested; status is shown per pack. Seven more packs are in design in the catalogue — tell the beta which one your business needs next.
For businesses running a channel without a media team. Research → Script → Video → Approved Publish → Comments → Growth, with a human gate at every step that can hurt you.
Works via: YouTube Data API v3 + YouTube Analytics API (default general quota 10,000 units/day; uploads sit in their own bucket — quota headroom is planned for) · keyword/competitor research tools · LLM + text-to-speech on a paid commercial plan · video assembly editor · thumbnail design · Drive/Docs for scripts and assets. Status: in private-beta build — connections are claimed live only after OAuth and quota testing.
| Topic | You approve the backlog before any scripting starts. |
| Script | You approve script + claims; Checker verifies each claim has a source. |
| Voice & visuals | You approve voice and thumbnail. No cloned voice of another person; no recognisable real person, brand, or logo without permission. |
| Publish | You approve the Private draft and set disclosure: realistic synthetic content must be disclosed via YouTube's altered/synthetic setting; AI help with scripts, titles, or captions needs no disclosure — stated per video, in the approval record. |
| Monetisation safety | Original insight in every video — YouTube's July 2025 “inauthentic content” policy targets repetitive, mass-produced content. Cadence capped (max 1–2/day for a business channel) and a human perspective stays in the video. |
8 approved videos published (2/week) from the backlog · median approval turnaround under 48 hours · zero policy strikes, Content ID blocks, or undisclosed-synthetic publishes.
Reported: impressions, click-through rate (benchmark 4–10%; 8%+ is strong), average view duration / retention (35–50% is strong for a small channel; the cited average is ~23.7%), watch time, subscribers per video, top retention drop-off per video, comments answered vs escalated — and credits per published video. Success is a repeatable cadence and a baseline for month 2. CrewOra never promises views, subscribers, or monetisation.
There is no broker connection, no brokerage API key, and no order-routing code in this pack at all. Removing the code removes the accident — what remains is disciplined research, paper trading, and an honest journal.
Works via: a market-data provider (delayed or real-time per licence — the delay is always labelled) · charting · news/filings feeds · spreadsheet/CSV export so your journal survives outside CrewOra · earnings calendar. Explicitly not included: brokerage APIs, payment rails, custody, or any live execution path.
| Strategy | You approve rules and risk caps before any paper trade; caps can't be raised mid-day without a second confirmation. |
| Paper trade | Each ticket approved by you, or covered by a pre-approved rule set with a daily cap — never unlimited auto-trading, even simulated. |
| Language | Banned: “guaranteed,” “will profit,” “buy now,” personalised recommendations. Required on every output: “simulated,” “hypothetical,” “past/simulated performance is not indicative of future results.” |
| Honesty | Simulated P&L is never presented as live, audited, or real. Exports are watermarked “SIMULATED — NO REAL MONEY.” |
| Eligibility | 18+. You confirm paper trading is permitted in your jurisdiction. This pack's copy is legally reviewed before launch — the research behind it is not legal advice. |
20+ approved paper trades (or your actual count, reported as-is — never padded) · journal completeness 100%: every trade has thesis, risk, and review.
Reported: total simulated P&L and return, win rate, profit factor, expectancy, average win/loss, average risk:reward, max drawdown ($ and %), results by strategy/symbol/session, streaks, rule-breach count and trend, and Sharpe — with the caveat that 30 days is too short for a stable Sharpe, so it's reported, never headlined. Plus credits per research brief / per paper trade. Success is discipline and a report you can learn from — even if simulated P&L is negative, and the report says so plainly.
For businesses losing calls they can't afford to miss. One use case proven first — appointment booking — then expanded. No cold AI dialling, ever: outbound runs only from a consented list you upload and attest.
Works via: one primary voice stack (telephony + speech-to-text + LLM + text-to-speech, latency target under ~800ms, ideally ~500ms) with one fallback — not five stitched vendors per customer · Google/Outlook/Calendly calendars · HubSpot/Salesforce/Zendesk or Sheets as a minimum CRM · SMS/email confirmations · recording storage with retention controls. Voice is metered separately from credits because its real cost dwarfs text work (see Pricing). Status: in private-beta build.
| Disclosure | Every call opens with AI disclosure + business name + “say ‘human’ anytime” — aligned with EU AI Act Article 50 (applies 2 Aug 2026) and TCPA identification rules. |
| Consent gate | No prior express consent on file (written, for marketing) → the number is undiallable. DNC scrub at most 31 days old, real-time preferred. Calling window 8am–9pm recipient local time; stricter state windows win. |
| Caller ID | A valid callback number, STIR/SHAKEN attested. |
| Voice cloning | Only your own verified voice, a licensed library voice, or a stock voice — with written consent and verification on file. Never a third party, employee, or public figure. |
| Recording | Notice at call start; retention period stated; EU/UK lawful basis confirmed before capture. HIPAA is offered only on a BAA tier — it is not claimed on self-serve plans. |
Week 1 sets your human / missed-call baseline; weeks 2–4 run on AI. Reported: calls answered, containment (resolved by AI without transfer) and CSAT-validated resolution (contained and confirmed resolved — the honest number; best-in-class cites 60–75% containment but only 50–65% CSAT-validated, and a 40–60% containment start on one use case is a strong start), booking conversion and kept-booking rate, escalation rate, 72-hour re-contact rate (the counter-metric that catches silent failures), latency, hallucination rate on a weekly 100–200 call sample, missed-call capture — and cost per verified resolution against a human-call baseline of ~$7.16/call (third-party cited figure).
Seven more packs are in design, scoped from the roles the market research found businesses already pay specialists for — support resolution, outbound sales, voice, content. Honest status: these are planned packs, not beta-ready. Each will be staffed by SI Agents and get the same treatment as the flagships before it opens — named roles, an approval table, a per-employee budget, and a Day-30 scorecard — and beta demand decides the build order.
Ticket triage, knowledge-base-grounded replies, refunds triage, escalation to a human with full context.
Lead research and enrichment, personalised outreach drafts, follow-up sequences, qualification scoring, CRM updates, meeting booking.
Content calendar, short-form scripts and clips cut from long-form, captions, hashtag and trend research, comment triage.
Product listing drafts, inventory and order-status answers, review monitoring and reply drafts, supplier follow-up drafts.
Invoice chasing drafts, expense categorisation, receipt matching, weekly cash summary, inbox and document filing.
Job-post drafts, CV screening against your criteria, interview scheduling, candidate follow-up drafts, onboarding checklists.
A lighter front-desk pack for clinics, salons, trades, and professional services: calls and messages answered, appointments booked, reminders drafted. Pairs with the Call Centre Pack when call volume grows.
That's the point of the beta. Describe the workload in the Hiring Manager finder — if several businesses ask for the same pack, it moves up the build order, with its roles, approvals, and scorecard published before it opens.
Request a pack in the betaAnswer three questions and CrewOra maps your workload to the packs that cover it, with the approvals, budget shape, and 30-day success metric for each. Then generate a one-page brief — packs, costs, controls, and what success looks like — that you can download or copy and hand to a hiring manager or partner for a decision.
The decision view: every pack against the department that owns the job, the approvals it requires, the budget shape, and the metric a hiring manager judges it by in 30 days.
| Pack | Department / owner | Jobs covered | Approvals required | Budget shape | Day-30 success metric |
|---|---|---|---|---|---|
| YouTube Pack Flagship | Marketing / owner | Research, scripts, production, publish, comments, growth | Topic, script, voice/visuals, publish + disclosure | Credits per employee wallet | 8 approved videos; credits per published video |
| Call Centre Pack Flagship | Operations / front desk | Inbound calls, booking, summaries, approved outbound | Consent gate in code; booking rules; payment/refund/legal always escalated | Voice minutes (separate meter) + credits | CSAT-validated resolutions; cost per verified resolution |
| Trading Research Pack Flagship | Owner / research | Briefs, strategy lab, paper trades, journal, risk | Strategy + caps; every simulated ticket | Credits per employee wallet | Journal completeness 100%; rule-breach trend (never a profit claim) |
| Customer Support Pack Planned | Customer Experience | Ticket triage, KB replies, escalation | Refund/credit/policy promises | Credits per employee wallet | Verified resolutions; cost per resolution |
| Sales Outreach Pack Planned | Sales | Lead research, outreach drafts, qualification, booking | First-touch messages + campaign lists | Credits per employee wallet | Qualified meetings per 100 approved outreaches |
| Social & Short-Form Pack Planned | Marketing | Calendar, clips, captions, comment triage | Every post before publish | Credits per employee wallet | Posts published on cadence; cost per post |
| E-commerce Ops Pack Planned | Operations | Listings, order answers, reviews, supplier follow-ups | Prices, publishes, delivery/refund promises | Credits per employee wallet | Order questions resolved; updates shipped |
| Bookkeeping & Admin Pack Planned | Finance & Admin | Invoice chasing, expenses, cash summary, filing | Anything moving money or filed externally | Credits per employee wallet | Admin hours returned; invoices chased on schedule |
| Recruiting & HR Pack Planned | People | Job posts, CV screening, scheduling, onboarding | Rejections and offer-stage messages are human-sent | Credits per employee wallet | Application → interview time; screening hours saved |
| Reception & Bookings Pack Planned | Operations / front desk | Enquiries answered, appointments, reminders | Booking rules + templates, weekly log audit | Credits (+ voice minutes if calls) | Missed enquiries captured; bookings kept |
Select the packs you're considering, add your business and the decision you need, and generate a one-page brief: what each pack does, what it may never do without approval, what it costs on beta pricing, and the 30-day metric it will be judged by. Download it or copy it — it's written to be forwarded as-is.
Trust isn't a slogan here — it's three artefacts your business can inspect: the approval log, the credit ledger, and the outcome report. Samples below are labelled sample; beta customers get the real ones, exportable.
| Step | Detail | Credits |
|---|---|---|
| Brief approved | Topic #4 from approved backlog · estimate shown first | est. 145 |
| Research | Keyword + 3 competitor teardowns, sources attached | 38 |
| Script draft | Hook, retention beats, 3 titles, thumbnail brief | 52 |
| Checker pass | 11 claims verified · 1 flagged unverified → removed | 19 |
| Voice + assembly | Licensed stock voice · captions · rights logged | 31 |
| Approval | Private draft approved by owner · disclosure set | 0 |
| Published | Actual vs estimate: 140 vs 145 (−3%) | 140 |
These are CrewOra's recommended beta tiers, structured to stay viable against real multi-agent costs. They are to be validated during the private beta before public launch — if beta changes them, beta customers keep a founding discount (20% for 12 months), stated as such.
CrewOra calls its own agents SI Agents, following the new U.S. federal term defined in the FAQ below. Other vendors still use “AI workforce” and “AI employees” language, and CrewOra won't call itself “the first AI workforce platform” or “the best SI Agents” — those claims are unprovable and, under FTC doctrine, an objective claim needs its evidence held before it's published. CrewOra's claim is narrower, and it's checkable in the demo: the approval-first workforce workspace where every SI Agent (AI employee) has its own budget, approval queue, and 30-day outcome report — stated with its scope, small teams, alongside it. Until each element ships and is proven in beta, it stays a goal on this page, not a boast.
SI = Super Intelligence — the new U.S. federal term for AI, set by Executive Order 14434 (September 2026) for federal executive-branch communications. An SI Agent in CrewOra is an AI employee under that new name: the same underlying AI technology, working in a pack with its own budget and approval queue. The name is not a claim of superhuman or superintelligent ability — and no SI Agent can do anything external without your approval.
Not for anything external. Drafts, research, and internal preparation happen autonomously inside the workspace. Publishing, sending, calling, booking, and spending are default-deny: each needs a named human approval, and the decision is written to an audit log you can export. If an employee's wallet hits zero, it drops to draft-only — it cannot publish or call on credit.
No, and it cannot. A research and paper-trading simulator for education. No real money, no real securities, no brokerage, no advice. The pack contains no brokerage connection and no order-routing code. Simulated results are watermarked as simulated on every export, and the 30-day report measures discipline and journal quality — never a profit promise. It's 18+, and you confirm paper trading is permitted where you live.
No. Outbound runs only from a list you upload and attest is consented. Consent is enforced in code: no consent record, no dial. Numbers are scrubbed against do-not-call lists, dialled only inside the recipient's local calling window (8am–9pm, stricter state rules win), and an in-call “stop” suppresses the number immediately. Every call opens by disclosing it's an AI assistant and offering a human.
Your scripts, journals, call summaries, and ledgers export in one click (CSV/PDF/JSON) — your data leaves with you. Customer data is never used to train models without explicit consent, PII is redacted in logs and summaries where feasible, and providers are used under no-training / zero-retention terms where available. Exact retention periods, sub-processors, and backup/restore procedure will be published on the trust page at launch; certifications CrewOra doesn't hold yet won't be claimed (see “What CrewOra will not claim — yet”).
CrewOra is in internal alpha now, moving to an invite-only private beta of 20–50 businesses. Beta applicants are onboarded pack by pack, with a baseline measured first so the 30-day report means something. Apply below — applications are reviewed in order, and honest fit matters more than speed: the beta needs businesses with a real job for a pack, not curiosity signups.
Status: CrewOra is in internal alpha. Next is an invite-only private beta of 20–50 businesses, onboarded in order of application — no dates are promised.
Businesses are onboarded 10–15 per pack. Each gets a 20-minute onboarding call, a measured baseline (your current videos/month, missed calls, or journal habit), and a 30-day outcome report at the end — the same report format shown on this page.
Beta terms are stated up front: expect rough edges, your data is handled per the commitments in the FAQ, and beta converters keep a 20% founding discount for 12 months at public launch.
This form prepares your application on this page — nothing is sent anywhere. Copy the summary or download it and send it through whatever channel you reached CrewOra on.