Foan is a phone agent for normal businesses, not a substitute for human judgement in high-stakes situations. Some uses are off-limits, regardless of plan.
We won't run agents for…
Mental health, crisis, or suicide-prevention lines
Even with the most carefully-tuned prompt, an AI cannot reliably triage someone in mental-health crisis. The downside risk (a caller in distress getting bad advice or being mis-escalated) is unbounded. We won't deploy here. If you run one of these services, our advice is to keep humans on the line and use Foan only for non-clinical adjacent tasks — appointment reminders for already-engaged patients, billing, intake of non-distressed callers.
Medical advice, prescriptions, or diagnoses
Same reason. Medicine is regulated, mistakes harm people, and the AI doesn't know what it doesn't know. Foan can take messages for clinics, schedule appointments, handle billing — it won't tell anyone whether their chest pain is heartburn.
Legal advice on specific cases
Foan can answer "what areas of law do you cover?" and "what are your fees?" — public-facing FAQ. It won't analyse a caller's situation and tell them what to do. Solo lawyers absolutely benefit from Foan as a receptionist; the agent must always say "I can pass this to one of the lawyers — what's the best number to call you back on?"
Financial advice, tax recommendations, or investment guidance
Same logic — these are regulated activities. Foan can take details and route, never advise.
Election interference, political robocalls, or impersonation
We refuse outright. Automated political voice calls are heavily restricted in most jurisdictions and in some countries (US TCPA, parts of the EU) outright illegal without specific opt-in. The reputational risk to Foan and our other customers is unacceptable.
Debt collection beyond a friendly reminder
A neutral "you have an outstanding invoice — can I help with payment?" is fine and within most jurisdictions' rules. Aggressive collection scripts (threats of escalation, repeated dialing, anything that could be construed as harassment) we won't run. Use a human collector for that.
Pretending to be human
If a caller asks "is this AI?", the agent must say yes. This is non-negotiable, regardless of how natural the voice sounds. We hard-code this behaviour into every system prompt.
Cold-calling without a verified opt-in list
See the Calling-rules guide. We gate cold-calling templates behind a consent attestation in the campaign wizard. Lying on the attestation = breach of our Terms.
Spam, scams, or fraudulent campaigns
Obvious. Reported instances result in immediate workspace suspension and we keep the call logs as evidence.
What this means in practice
Most businesses don't bump into any of these. Roofers, salons, contractors, e-commerce, real estate, fitness, hospitality, dentists (for booking and billing only), insurance brokers, lawyers (for intake only), B2B SaaS — all fine.
If your use case is in the grey zone, contact us before you launch the campaign. We'd rather have the conversation up front than yank a workspace.
Why we're public about this
Two reasons:
- Trust is a moat. Customers who care about doing things right will choose us over a vendor that doesn't think about this. Cowboys self-select away.
- Legal posture. We have a documented record that we don't enable these use cases — useful if a regulator ever asks.
If you want to suggest something we should add to this list, email hello@foan.me.