How accurate is the savings projection?+
The projection uses public deflection benchmarks from Salesforce, Gartner, McKinsey, Crescendo, and Zendesk, and applies a 15% conservative haircut for residual escalation, AI runtime overhead, and supervision drag. Expect ±20% accuracy on the mid estimate when your ticket volume and handle time are reasonably well-known.
Why does the calculator recommend SaaS sometimes when buzzi.ai builds custom?+
Honesty is the moat. The recommendation engine is rule-based and audited in CI: SaaS must be recommended in at least 40% of synthetic scenarios. If your inputs favour SaaS, that is what the tool says — even though we build custom-trained chatbots.
What ticket complexity counts as simple, medium, or complex?+
Simple = FAQ, account lookups, status updates (the 70-80% AI-deflectable bucket). Medium = bookable steps such as cancellations, address changes, or refunds. Complex = multi-system or judgement calls that humans still own.
Why is voice deflection lower than chat?+
ASR error rates and intent ambiguity cap voice deflection. We model voice at 65% of chat effectiveness — improving rapidly but still significantly harder.
What hidden costs does the tool include?+
Token / API consumption (10-15% of gross savings), helpdesk integration ($25K-$35K one-time), training data prep ($12K-$18K one-time), and ongoing QA / supervision (~8% of agent cost). All are subtracted before showing net savings.
How does the build-vs-buy logic work?+
A rule-based score: low volume + simple tickets + single language → SaaS. High volume + complex tickets + regulated industry → custom. Mid-range or mixed signals → hybrid (SaaS for tier-1, custom workflows for tier-2/3).
Can I change the assumptions on the results page?+
Yes. The editable assumptions panel lets you override AHT, agent cost, ticket volume, and the deflection rate. The model recomputes in the browser as you change values — no server round-trip.
Do you store my submission?+
Only if you sign in to unlock the full breakdown + PDF. Anonymous previews are not persisted. Your row is private to your account; we never sell or share it.
How is currency handled?+
The tool detects your country and pre-fills the currency (USD, GBP, EUR, AED, SGD, AUD, INR, or CAD). All cost figures are formatted in the chosen currency on results, PDF, and emails.
How often are the benchmarks refreshed?+
Quarterly, from Salesforce State of Service, Gartner CIO Survey, McKinsey State of AI, Crescendo CX Benchmarks, and Zendesk CX Trends. Each refresh re-runs the SaaS≥40% distribution gate.
Is the methodology open?+
Yes. Every benchmark, channel modifier, and hidden-cost range is published on the methodology page with sources cited. The build-vs-buy distribution gate is in CI as a public test.
Can buzzi.ai help me build a custom chatbot?+
Yes. If your projected savings exceed $300K/year and the recommendation is custom or hybrid, we will deliver a written deployment plan in 14 days. The contact form on the results page routes directly to a senior delivery consultant.