Crisis routing is the first rule in the system prompt. Any indication of overdose, suicidal ideation, or acute danger triggers the bot to share the SAMHSA National Helpline at 1-800-662-HELP (4357), the 988 Suicide and Crisis Lifeline, and 911 for emergencies, in that order depending on context. The bot does not try to talk a person through a crisis itself. It points to live human help that operates 24/7. This rule applies even when the visitor asks the bot to stay on a program-detail topic, because crisis routing always takes precedence.

No. Level of care (outpatient, IOP, PHP, residential, detox) is a clinical decision based on ASAM criteria and made by a licensed counselor at intake. The bot describes what each level of care looks like at your practice using your published copy, but it never recommends one for a specific case. The boundary is in the system prompt and consistent across every conversation. Recommending IOP versus residential from a chat conversation is exactly the kind of error a bounded chatbot is designed to prevent.

Yes, in general terms. Publish your admissions and insurance policy as a page covering which Medicaid plans you accept, which commercial insurers, your self-pay rate, and any sliding scale. The bot answers from that page and is explicit that final verification of coverage happens at intake, not in chat. This is honest and protects the practice from inadvertently promising coverage that admissions later has to walk back.

Yes. The intake script branches on whether the caller is the client, a family member, or a referring clinician. Family members get a separate path that acknowledges their context, points to Al-Anon and Nar-Anon resources if you publish them, and captures the family's intake without pretending the client themselves is on the line. The eventual goal is still a conversation between the client and a licensed counselor, but the family path respects that someone else often initiates contact.

Conversations live in your WordPress database. Model calls go to OpenAI, Anthropic, Google, or OpenRouter using your own API key, never through a Sleek-hosted service. For a practice handling sensitive substance use disclosures, treat the bot as an administrative tool and avoid asking for clinical specifics in chat. Retention is configurable in plugin settings, and most practices set short retention with administrative summaries forwarded to the EMR through a webhook. 42 CFR Part 2 considerations should be reviewed with your compliance team before launch.

Yes. Medication-assisted treatment intake has different needs from talk-therapy intake (current use, last use, opioid history, whether the patient has tried buprenorphine before, induction setting). The MAT page can run a separate intake bot with the appropriate flow, while never assessing whether MAT is clinically appropriate for the individual. The induction decision belongs to the prescribing provider. The bot captures enough context to route the call, not enough to substitute for clinical evaluation.

If your admissions software exposes a JavaScript API or a webhook, yes. SleekAI's JS API lets the chatbot push captured intake into your admissions tooling, and many practices route the bot's output to a webhook that lands in Kipu, Sunwave, or whatever EMR they use. Practices on simpler stacks often have the bot create a WordPress custom post with the intake fields, which the admissions coordinator reviews each morning.

Yes. SleekAI is multibot. The IOP page, the family program page, the MAT page, and the alumni page can each run their own bot with a different tone, intake script, and routing destination. A family inquiry on the family page should not route to the MAT admissions queue, and trying to handle every program with one bot produces a worse experience across the board. Display conditions ensure each bot only appears on its intended pages.