GenAI on Your Data. With Safeguards.
What it does
Four Services. One Platform.
Content generation
Create.
Picks topics from your own search data, drafts in your brand voice and glossary, and checks every claim against your live website before your approver sees it.
- SEO articles
- Social media posts
- Short videos
- Ad copy
Customer engagement and knowledge
Assist.
Answers customers and staff from your own pages and documents, shows where each answer came from, and hands over to a person when it is not sure.
- Website customer assistant
- Staff knowledge assistant
- Proposal and quotation drafting
Conversational analytics
Insight.
Ask your marketing data a question in plain English. Every number in the answer is checked against the data, and it explains what drove a change.
- Website and search analytics
- Social media performance
- Competitor analysis
- Website audit and conversion advice
Operations automation
Intake.
Turns enquiries from email, web forms and chat into clean records, checks them against your business rules, and queues anything uncertain for a person.
- Enquiry automation
Sentrify
Add-ons. Work with any service.
Add-on
Sentrify Advisor
An AI agent your staff can ask about the data on their screen, which also sends weekly advice.
Add-on
Sentrify Connect
A read-only API and MCP connector, so the AI tools you already use (Claude, ChatGPT, Microsoft Copilot) can work with your Sentrify data.
How it behaves
It Knows What It Doesn't Know
Most AI tools will answer anything, confidently. Sentrify answers only from your sources and says which one it used. When your pages don't cover a question, it hands over to your team with the conversation attached, instead of guessing.
The same rule runs through every service. Articles are checked claim by claim against your live site. Every number in an analytics summary is checked against the data. Enquiries it can't read with confidence go to a person.
Example conversation, website assistant
- CustomerDo you open on Sunday mornings?
Governance
Safeguards Built In
Personal Data Removed First
NRIC/FIN numbers, phone numbers, email addresses, card and passport numbers are replaced with placeholders before any text reaches an AI model.
Prompt-Injection Screening
Hidden text is stripped from web pages and uploads, and instruction-like text is quarantined, so a document cannot quietly take over the AI.
Toxicity and Bias Checks
Every output is checked, with a bias check that covers race, religion, nationality, gender, age, disability and sexual orientation, tuned to Singapore.
Grounded in Your Sources
Answers cite the page or document they came from. Claims are checked against your live site, and an answer it cannot support goes to a person.
Your Approver Signs Off
Nothing is published until the approver you name signs off that exact version. Approvals are kept in a tamper-evident ledger.
Every Step Logged
An audit log records each step, and the type of personal data found, never the value. Staff see only the documents their role allows.
Getting started
Live in About Six to Eight Weeks
Ready to connect
- Google Analytics 4
- Google Search Console
- Microsoft 365 (SharePoint)
- Google Docs
- Strapi CMS
- Git-hosted websites
- CSV export
- CRM webhook
Weeks 1–2
Scoping
Choose your services, agree the measures of success and today's baseline, and confirm data access, roles and approvers.Weeks 2–4
Configuration
Your brand kit and glossary, knowledge sources, metrics, record fields and business rules. Settings, not new code.Weeks 3–5
Integration
Connect your website or CMS, analytics, Microsoft 365 or Google Workspace, and your CRM.Weeks 4–6
Safeguards and testing
Tune the personal-data, bias and injection checks to your content, and build a test set from your own cases.Weeks 6–7
Testing with your team
Your staff try it on real work, then a workshop on daily use, approvals and governance.Then
Go live
A period of supported use with monthly reviews, ending in a report of results against your baseline.
Questions
Before You Ask
Anthropic Claude for drafting, independent review, answers and analytics summaries, and OpenAI's moderation model for toxicity screening. Both are used through their APIs on LOMA's accounts, and Anthropic does not train on API inputs or outputs by default.
No. Nothing goes out until the approver you name signs off that exact version, and every approval is recorded.
Singapore personal data such as NRIC/FIN, phone numbers, email addresses and card numbers is replaced with placeholders before any text reaches an AI model, and outputs are checked again before publishing. The website assistant asks before keeping anyone's contact details and records the answer.
No. You choose one or more services and pay only for those. The add-ons need at least one service.
Connectors for analytics, Microsoft 365, Google Docs, Strapi and Git-hosted sites are ready. Others, such as WordPress, Shopify or a CRM, are built for your company during setup.
Typically six to eight weeks from kick-off, followed by a period of supported use with monthly reviews.
Case studies
Our Work in Software Development
See Sentrify on Your Own Data
Tell us which part of your work you'd hand to AI first. We'll show you what Sentrify would do with it, and what it wouldn't.
Checkout