Servly is the orchestration engine behind your AI persona. It knows your content, runs your e-services through a visual workflow engine, and shows up as a humanoid avatar, a copilot inside your web pages, a chat or a voice on the phone. Set it up once, and every surface stays in sync.
Diagram: your documents feed one persona at the center. A ring of workflow steps orbits it, and the persona answers through four surfaces: a humanoid avatar, a browser copilot, messaging and phone.
The workflow engine
Licence renewals, bookings, applications, change requests: design each e-service once on a visual canvas, and the persona runs it in conversation on every surface, checking every answer before it calls your systems.
Animation: a request moves through a service. It fails validation once and retries, detours to answer a side question, passes a one-time passcode check and a condition, calls your system's API, and finishes by sending a webhook.
Pick a question. Watch the engine decide what it is, find the answer or start the right workflow, and hand the same response to each surface in the form that surface needs.
Someone asks
Playing through examples. Pick one to stay on it.
The engine decides
One answer, adapted to each surface.
Every surface answers
I can renew that with you now. What's your licence number?
“…please say or key in your licence number.”
Real-time faces and voices are improving every month, and plenty of excellent companies compete on them. We focus on what sits behind them: knowledge, workflows, memory and guardrails. Plug in the models and renderers you trust, and change them without rebuilding anything.
Each knowledge base can carry its own provider keys, so one team can run on one model and another team on a different one.
Four systems work together on every message, whichever surface it arrives on.
Upload PDFs, Word files, spreadsheets and text. The engine reads them (with OCR for scans), turns them into question-and-answer pairs, audits each pair for clarity, rewrites the weak ones and indexes every knowledge base separately. When people ask something it can't answer, that becomes a knowledge gap on your to-do list, not a made-up reply.
How knowledge and gaps workDesign e-services in a drag-and-drop flow builder: steps, conditions, API calls and endings. The persona proposes the right flow mid-conversation, validates every answer (national IDs and phone formats across the GCC included), checks identity with a one-time passcode, calls your systems, and picks the flow back up after a side question.
Inside the workflow engineThe agent keeps the thread across turns: it asks a clarifying question when a request is ambiguous, remembers what someone already told it, and switches between Arabic and English mid-conversation.
The conversation agentAnswers stay inside your own content. Daily AI spend caps, phone numbers and emails redacted from saved transcripts, audit logs, domain allow-lists, and an evaluation harness that tests the agent before changes go live.
Security and guardrailsMost assistants guess when they're unsure. Ours turns every unanswered question into a gap your team can close, so the persona gets measurably better each week.
A question arrives on any surface: avatar, copilot, chat or phone.
Instead of guessing, the persona says so and offers a person or another route.
The question lands in your console, grouped with repeats and ranked by how often it's asked.
Write the answer, or approve one the engine drafts in Arabic and English. It's indexed straight away, and every surface knows it.
Same persona, same knowledge, same workflows. Turn surfaces on as you need them.
A face on your site or kiosk, rendered by the lipsync provider you choose.
A cursor that moves through your pages and talks people through them.
The same persona in the messaging apps your customers already use.
The same persona answering calls, with the same knowledge and services.
Need a person in the loop? The copilot hands over to a human agent through Pairly, with the whole conversation attached. Running Mendix? The persona ships as a pluggable widget for Mendix apps.
Each persona speaks the dialect your audience speaks, understands people who mix Arabic and English, and switches language mid-conversation without losing the thread.
Each embed is scoped to one knowledge base and works only on the domains you allow.
<script src="https://avatar-cdn.syntiva.tech/v1/YOUR_WIDGET_KEY/avatar-widget.iife.js"></script>
<avatar-widget
api-url="https://avatar-api.syntiva.tech"
widget-key="YOUR_WIDGET_KEY"
position="bottom-right"
theme="light"></avatar-widget><script src="https://file.syntiva.tech/latest/cobrowse-webcomponent.umd.js"></script>
<cobrowse-widget
copilot-api-url="https://avatar-api.syntiva.tech"
knowledge-base-id="YOUR_KB_ID"
api-key="YOUR_WIDGET_KEY"
language="ar"></cobrowse-widget>Full options are in the widget guide and the API reference.
Send us a few documents and one e-service. We'll show you the persona answering on the surfaces you care about.