Q-01Integrations
How do you connect to WhatsApp, email and our existing CRM?
Short answer
Through official APIs and webhooks, with every field mapped before we build.
How it works
Inbound messages arrive as verified webhooks, are queued, de-duplicated and written to one contact record. Outbound replies and CRM updates go through the same layer, so a failed call is retried and shown in a failure queue instead of disappearing. [Confirm the WhatsApp provider per project]
- Channels
- WhatsApp, email, web forms
- Pattern
- Webhook in, queue, then one record
- Failures
- Retried, then shown to a person
Q-02AI and data
Which AI models do you use, and is our data used to train them?
Short answer
Hosted model APIs, chosen per workflow. Rules handle what rules can, and AI handles the rest.
How it works
We call provider APIs for classification, extraction and drafting, keep prompts and outputs in your audit log, and can strip personal details before a call. Low-confidence results go to a person instead of triggering an action. [Confirm each provider's data-use terms before launch]
- Models
- Provider APIs, selected per workflow
- Low confidence
- Routed to human review
- Training on your data
- [Confirm per provider and plan]
Q-03Security
How is access controlled, and what gets logged?
Short answer
Roles are enforced on the server, and every change and automation run is logged.
How it works
Permissions are checked in the API, not only in the interface, so a hidden button cannot be bypassed. Each record change and automation run is written to an audit trail with who, what and when. [Add hosting region and encryption details]
- Access
- Role-based, server-enforced
- Audit
- Change and run history
- Approvals
- A person signs off on payments and exceptions
Q-04Reliability
What happens when an automation fails or the AI is unsure?
Short answer
Nothing is silently dropped. Failures are retried, then shown to a person.
How it works
Every step has a status. Temporary errors retry with backoff, and anything still failing lands in a visible queue with the input, the error and a retry option. Uncertain AI results never trigger customer-facing actions without review.
- Retries
- Automatic, then a manual queue
- Visibility
- Failure queue and run log
- Customer-facing AI
- Reviewed when confidence is low
Q-05Ownership
Who owns the code, the data and the infrastructure?
Short answer
You do, with no lock-in.
How it works
Code lives in a repository you own, infrastructure runs in accounts you control, and you pay hosting costs directly. We document the data model and deployment steps so another engineer could take over. [Confirm in contract]
- Code
- Delivered to your repository
- Hosting
- Your accounts, your costs
- Documentation
- Data model and deploy steps
Q-06Privacy
How do you handle UK GDPR and US privacy requirements?
Short answer
We design for data minimization and can sign a data processing agreement. We do not claim certifications we do not hold.
How it works
We collect only the fields a workflow needs, agree retention rules with you, and support access and deletion requests from your records. [Confirm DPA availability and any certifications before publishing]
- DPA
- [Confirm availability]
- Retention
- Agreed per workflow
- Certifications
- None claimed
Q-07Delivery
How are builds scoped, released and supported?
Short answer
Fixed scope in writing first, then staged releases, with support from $75 a month.
How it works
The $99 Workflow Audit produces the scope. Builds start from $650, are developed in a staging environment and released in versioned steps so a bad release can be rolled back. Support covers monitoring, fixes and small changes. [Confirm response times]
- Audit
- $99 fixed, credited toward a build
- Build
- From $650, scope agreed in writing
- Support
- From $75 a month