AI employees that work inside rules you set

A CEO that comes first

• Every new workspace is offered its CEO first. It introduces itself and walks you through setting up the business — each step checked against what is really in place — then moves on to running it with you.
• Link your own WhatsApp and the CEO sends you a daily briefing at your own hour, tells you when work needs your approval and answers questions — typed or as voice notes — from your business's records. Routine approvals need your PIN; anything involving money or employee data is only approved in Skofa.

In Skofa: Home, Settings → WhatsApp

Hire, configure and supervise

• Create AI employees from discovery recommendations or from scratch; each change is versioned, and agents can be paused and resumed at any time.
• Five autonomy levels, from recommend-only to operating within defined financial, legal and operational limits.
• Permission scopes are granted per AI employee and can never exceed what an owner or admin has authorised.
• Per-employee model policy and budgets: which model tiers it may use and how much a single task may cost.

In Skofa: AI Workforce

Goals, not prompts

• Give an AI employee a goal; it plans the steps against the capabilities your connected systems actually offer, and validates the plan before anything runs.
• Execution is bounded: a limited number of observe-then-act rounds and at most one re-plan after a failure, with per-step timeouts — no runaway loops.
• Workflows run sequential and conditional steps with retries, timeouts and failure handling, and AI employees can hand work to each other in collaboration threads.

In Skofa: Goals, Workflows, Collaboration

Humans stay in the loop

• Financial, destructive and outbound-message actions always wait for a person to approve them, whatever the autonomy level.
• Approvals arrive as in-app notifications and email, with a badge in the sidebar — and on WhatsApp, for people who link it.
• A dry run of any workflow shows each step's real authorization decision, which connected system would serve it and which approvals it would need — without calling anything or creating a run.

In Skofa: Workflows, Notifications

Grounded in your business

• Every model call made on your behalf receives one capped context pack built only from your own business: your profile, business policy, and knowledge with source attribution.
• Content from emails, documents, websites and connected systems is treated as data, never as instructions.
• Outcomes of past recommendations are remembered and fed back into future ones.
• First-party abilities: sending email through Skofa (approval-gated, daily cap), reading the records Skofa has synced from your systems, and replying to customers on WhatsApp (approval-gated) — free text within WhatsApp's 24-hour window, your approved templates outside it.
• Your knowledge is kept up to date for you: facts from your profile, products and posting history are refreshed on their own, web pages you added are re-read, and anything learned from the work is suggested for you to approve — never added on its own.

In Skofa: Knowledge, Inbox

Content that gets made and posted

• AI employees produce real content — text posts, flyers and video ideas that become videos — which lands in Ready to post for you to post in one click.
• If you allow it, an AI employee posts on its own to the accounts you choose, within the limits you set and each platform's rules (TikTok posts are always confirmed by a person).
• Goals can repeat on a schedule, so recurring work keeps happening without being asked again.

In Skofa: Ready to post, Video Studio, Goals

Cost under control

• Tasks are routed to cheaper or stronger models by complexity, cached where safe, and stopped by per-business and platform-wide budgets.
• Every model invocation is recorded with its cost, so spend is visible per employee, per goal and per model.

In Skofa: Analytics

What is not there yet

• WhatsApp replies need your own WhatsApp Business number connected; outside the 24-hour window only templates approved for your number can be sent.
• Posting to Facebook and Instagram through Skofa's own Meta connection waits on Meta's app review; posting through the posting partner works today.
• Voice notes, photos and documents customers send on WhatsApp are made readable for your AI employees once Skofa's updated Privacy Policy describing it is published; until then staff open them in WhatsApp.