Frequently Asked Questions
Can't find your answer? Book a call — John answers everything live.
Privacy & Security
Does data ever leave our building?
No. The AI model runs entirely on your own hardware. No API calls are made to OpenAI, Microsoft, or any other cloud provider. There is no internet dependency once the system is activated.
Is this compliant with ABA Model Rule 1.6?
Yes. Because all processing occurs on hardware you control, with no data transmitted externally, this setup satisfies the confidentiality requirements of ABA Model Rule 1.6 and equivalent state bar rules.
Can we use it for client-privileged documents?
Yes — that's the primary use case. Documents never leave your office network, which is the same security model as any other software running on your own computers.
What happens if the hardware is stolen or breached?
The AI model itself contains no client data — it's a general language model. Your documents are stored on your server like any other file. Standard encryption (BitLocker, FileVault) protects the drive.
Performance & Capability
Is it as good as ChatGPT or GPT-4?
For typical legal tasks — drafting memos, summarising contracts, answering questions about your own documents — local models like Llama 3 perform at 85–90% of GPT-4 quality. For complex legal reasoning tasks you'd want to validate outputs, exactly as you would with any AI.
How many documents can it handle?
The Private Business Chatbot ingests up to 5,000 documents by default. For larger document sets (e.g. litigation with 50,000+ files), we build a custom indexing pipeline — pricing varies.
How fast is it?
On a modern machine with a GPU (e.g. RTX 3060), response times are 5–15 seconds per query. On CPU-only setups, expect 20–45 seconds. Both are usable for day-to-day work.
Setup & Support
Do we need an IT team?
No. John handles the full installation remotely. If you can use Zoom, you can participate in the setup call. No prior technical knowledge is required.
What if it breaks after setup?
Every installation includes a 7-day fix-or-refund guarantee. For ongoing support, the Monthly Retainer ($75–$150/mo) includes priority help, model upgrades, and reserved time for changes.
Can we update the documents the AI knows about?
Yes. We provide a simple process for adding new documents (drop files into a folder and run one command). Retainer clients get this done for them.
Do you offer in-person installation?
Yes, for clients in the NYC metro area. Everywhere else, installation is done remotely over Zoom — which works just as well and most clients prefer.
Pricing & Payment
Why is this cheaper than a SaaS subscription?
It's a one-time setup fee, not a recurring license. ChatGPT Teams costs $25/user/month — for a 5-person team that's $1,500/year, every year. A $500 Dexal installation pays for itself in 4 months and then costs nothing.
Do you offer payment plans?
We accept payment via invoice or credit card. For installations over $800, we can split into two payments (50% before, 50% after activation).
What's included in the monthly retainer?
Quarterly AI model upgrades, priority email/Zoom support, security checks, and reserved time for configuration changes. It's optional — your system runs fine without it.
Still have questions?
John answers every question on a free 20-minute call. No sales pressure.