Service Level Agreement

Standard, included with every ops retainer ($1,000–3,000/month): email and WhatsApp support, Sunday–Friday 09:00–18:00, with response targets of 4 business hours for P1, 8 business hours for P2, and the next business day for P3.

Priority, for retainers of $2,000/month or more, or as a $500/month add-on: adds a dedicated WhatsApp group and phone support, Sunday–Friday 08:00–20:00 with Saturday on-call for P1, and response targets of 1 hour for P1, 4 hours for P2, and the next business day for P3.

Enterprise, for custom contracts of $3,000/month or more in ops spend: adds a named engineer and a quarterly on-site option in the GCC, coverage 24/7 for P1 and Sunday–Friday 08:00–20:00 otherwise, with response targets of 30 minutes (24/7) for P1, 2 hours for P2, and 8 business hours for P3.

Response means a qualified human engaging with diagnosis, not an automatic acknowledgment. Resolution targets are stated per incident class in the runbook rather than promised as fixed hours, because we control our response — not every root cause, some of which sit with upstream model providers.

P1 means the production system is down or materially harming customers — for example, a voice agent not answering any calls, or an agent sending wrong prices at scale. P2 means a degraded state where a core function is impaired but a workaround exists — for example, handoff-to-human is broken, or latency is three times the baseline. P3 covers minor issues: cosmetic problems, single-conversation errors, or general questions.

For hosted agents (WhatsApp, voice, scribe), the Standard tier commits to 99.5% monthly uptime, with service credits of 5% of the monthly fee per 0.5% below target, capped at 30%. The Enterprise tier commits to 99.9% monthly uptime, with credits of 10% per 0.5% below target, capped at 50%. Uptime applies to our orchestration layer — the honest boundary of what we control — and excludes scheduled maintenance (announced at least 72 hours ahead), upstream model-provider or platform outages, and client-side infrastructure failures.

Level 1 is the on-call engineer, reached via WhatsApp or email, who owns the P1 acknowledgment within SLA and starts the incident log. Level 2 is the Head of Delivery, who is auto-escalated to if a P1 is unresolved after 2 hours or a P2 after 1 business day, and who owns the client communication cadence — an update every 2 hours until a P1 is resolved. Level 3 is the Founder/CEO, escalated to for any P1 lasting more than 8 hours, any data or privacy incident, or any client request for escalation.

For every P1 incident, a written incident report is sent to the client within 3 business days, covering what happened, the impact window, root cause, and prevention steps.