Last updated: May 20, 2026 · CP Global Dynamics LLC d/b/a Beverly Hills Growth
The short version. We use AI to power the chat widget on our site, to draft parts of our audit reports, and to produce the UGC video creative in our top-tier plan. Every AI output is reviewed by a human before it leaves our hands. AI is a tool, not a licensed professional. Our deliverables that involve AI are clearly labeled as such, both on our own surfaces and on any ad platform we publish to.
The bottom-right chat widget on beverlyhillsgrowth.com is powered by Google's Gemini large-language model (model gemini-2.0-flash) via the Gemini API. When you type a message, your text is sent to Google's API and an automated reply is returned. We do not store the transcript on our servers unless you confirm at the end of the conversation that you want a human follow-up. We do not use chat content to train any AI model, neither ours nor Google's: our Gemini API project runs on the paid tier (verified 2026-05-22 via the API's serviceTier: standard response field), and per Google's published Gemini API terms the paid tier does not use prompts or responses to improve Google products.
The audit you receive after submitting our form is computed deterministically (PageSpeed Insights API, Google Places API, on-page checks). The narrative parts that explain the findings in plain English are produced with help from a large-language model and reviewed by our team before delivery. The numbers are not invented; the prose is AI-assisted.
For clients on the $999/month tier, we produce 8 short (15-second) video ads per month using generative AI tools: text-to-video models, AI voice synthesis, and (if applicable) AI lip-sync. If a video features a real-person likeness (your face, your employee's face, a paid actor), that person must sign a model release before we produce the video. We provide a release template at onboarding. Videos that feature a synthetic avatar, a generated face that does not correspond to any real human, do not require a release.
If your plan includes weekly Google Business Profile posts or monthly review responses, drafts are produced with AI assistance. We review every draft before it publishes.
We operate a public AI Voice Receptionist demo line at +1 (833) 695-2183 and deploy the same service for paying clients on our Voice Receptionist subscription tiers. The voice agent is built on Retell AI as the orchestration layer, with real-time voice synthesis by ElevenLabs and language-model inference by OpenAI (model gpt-4o by default; Concierge-tier deployments may use Anthropic's Claude family on request). Calls handled by the agent are AI-generated, may be recorded, and a written summary is logged.
The agent identifies itself as AI. The opening greeting is configurable per Client deployment, and on our public demo line the agent opens with "Beverly Hills Growth, this is Olivia, how can I help?". The agent does not impersonate a human. This is required for any commercial bot used in California to incentivize a transaction under California Business & Professions Code § 17941 (SB 1001, "Bolstering Online Transparency").
Caller can request a human transfer at any time. Saying "human", "real person", "transfer me", "speak to someone", or analogous phrases triggers the agent's escalation path, which routes the call (or a written ticket) to the Client's configured human contact. On our public demo line the escalation goes to audit@beverlyhillsgrowth.com.
Limits. Voice agent output can mishear caller names, misread service requests, or misroute calls. We design escalation paths to catch high-stakes cases (emergencies, large-ticket consults, complaint handling) and ask the caller for confirmation before booking. Voice agent transcripts may also contain misheard words; the booking metadata (name, phone, requested service) is the authoritative record.
AI output can be wrong, outdated, biased, or made-up. Treat AI-generated content as a starting point for human review, not as a final answer. Specifically:
For any AI-generated UGC video creative we produce for you to run as an ad: we instruct you (in your Statement of Work) to turn on the platform's "made with AI" / "digitally created or altered" toggle when uploading the creative. Meta and TikTok both require this for AI-generated content; failure to comply is a platform policy violation and may result in ad disapproval or account suspension. This is your responsibility as the advertiser. We will remind you each time we deliver new creative.
Per EU AI Act Art. 50, we disclose that our chat widget is an AI system, and that any synthetic or AI-manipulated audio, image, or video we produce will be clearly labeled as such on publication. The labeling is satisfied (a) on our site by the first-bubble notice in the chat widget, and (b) on ad platforms by the platform-provided "made with AI" toggle described in Section 4.
We do not make exaggerated claims about the capabilities of AI tools. We do not use AI to create fake reviews, fake testimonials, or fake endorsements. Anything we produce that looks like a real-person testimonial is either a real testimonial (with consent) or clearly labeled as a dramatization.
The only third-party AI service we send your data to is Google's Gemini API (chat widget). We do not send your audit form contents, your name, your email, or any payment data to any AI provider. We may use locally-running or self-hosted AI tools in our internal workflow; in those cases your data does not leave our control.
Question about how we use AI: audit@beverlyhillsgrowth.com.