Enterprise AI workforce platform
AI employees that do the work, not just the talking.
Wetel puts AI employees to work in real roles: screening applicants, tracking orders, answering from your own policies, keeping records current. Each one follows your procedures, works inside the systems you already use, and checks with a person before anything consequential.
Already at work in recruitment, logistics and higher education.
Your team has become the router.
The applicant asks whether the role is still open. The customer asks where the parcel is. A colleague asks what the leave policy says. Each answer lives in a different system, so someone opens three tabs, copies, pastes, and types it back. Then does it again forty times before lunch.
Most AI tools only fix the typing. They answer from a document uploaded last quarter and leave the actual work — the lookup, the update, the filing — to a person.
An AI employee takes the routing off your team, and still checks with them before it commits you to anything.
Roles you can staff this quarter.
Every role below is built from the same parts and runs on the same platform. The live ones are running in production today; the rest are built with you.
Sixty CVs in the inbox, and the role closes Friday?
AI Recruiter
Takes applications on chat or the web, reads the résumé, checks it against the roles that are actually open, files the candidate into your HR system, and answers applicants' questions any hour.
Your recruiters meet the shortlist instead of reading the pile.
The same “where is it?” question, forty times a day?
AI Operations Coordinator
Looks up orders and shipments in the systems that hold them, reads a tracking code straight from a photo or a voice note, gives a real status, and hands the exceptions to your team.
Your ops team works the exceptions, not the lookups.
The answer is on page 40, and nobody can find page 40?
AI Knowledge Officer
Answers staff and member questions from your own documents and systems, grounded in what they actually say, and handles the simple follow-through: a booking, a renewal, a request.
People get the right answer the first time, with the source.
The CRM is always a week out of date?
AI Sales-Ops Assistant
Reads and updates CRM records through reviewed operations, prepares deal summaries, and logs follow-ups, so the pipeline reflects what actually happened.
A pipeline you can forecast from.
Design a role for your team
Every AI employee is the same seven parts, configured for a different job. We scope and build new roles with you during a pilot.
- HR & talent
- Operations
- Sales operations
- Finance
- Education
- Service operations
- Procurement
- IT
What every AI employee is made of.
Seven parts, the same for every role. That's why a new role reuses everything the last one proved.
01 A role and its procedure
Its job description and the steps it follows, drawn as a visual workflow you can read and edit — not a prompt you hope holds. The model chooses the branch; the graph decides which branches exist.
How workflows work02 Your knowledge
Your policies, product data and documents, isolated per organisation and searched on every answer — plus live reads from the systems that hold today's truth.
Knowledge bases03 Skills in your systems
It acts through reviewed operations with a known contract, known failure mode and known credential scope — 170+ certified operations today, plus your own APIs and MCP connectors. Credentials never enter the model's context.
Certified operations04 Identity and permissions
It can act as the signed-in person through single sign-on, only within what that person may do. Every organisation's data is isolated at every layer, with a dedicated database when a deal requires it.
How isolation works05 A place to work
The same AI employee on chat, voice, a 3D avatar, Telegram, your website, or inside the platforms your team already uses. Adding a channel never means rebuilding the employee.
Agents and channels06 A supervisor's record
Every run is traced step by step, drafts are tested against real conversations before they go live, conversations are scored, and spend is capped per organisation.
Observability07 Initiative
Starting work on a schedule or a trigger, not only when someone writes in — the part that turns a responsive assistant into an employee who picks up recurring work.
On the roadmapHow you hire one.
Nobody should buy an AI employee off a slide. Start with one role, on your own systems, measured on your own numbers.
- 1
Pick one workload
We sit with the team that owns it, map the procedure, and agree the one number that says whether it worked.
- 2
Set it up on your systems
Your knowledge, your integrations, your permissions — configured into the role, with the procedure laid out as a graph your team can review.
- 3
Go live, supervised
It runs as a draft against real conversations first. Once live, every run is on the record and anything consequential goes to a person.
- 4
Measure, then add the next role
If the number moved, the next role reuses the same knowledge, skills and permissions. Nothing gets rebuilt.
Under the hood, each role is a deterministic graph around the model's reasoning — reproducible, auditable, and readable by the people accountable for it.
See the workflow canvasSmall stuff, it just does. Big stuff, it asks.
That's the honest answer to “can we trust an AI employee?” An AI tool is risky when it's all-or-nothing — either too limited to be useful, or free to invent a 40% discount at 2am. This is neither.
It can only do what someone already approved.
Actions come from a reviewed catalogue, not from handing a model an API spec and hoping. Each one is a contract: this call, these arguments, this failure mode, this credential scope.
Everything it did is on the record.
Every run is traced step by step — the input, the output, the branch taken. Conversation quality is scored. Spend is capped. Nothing is a black box after the fact.
Consequential things come to a person.
Look something up, answer a question, log a note — it just does it. Money, a promise, someone's record, anything irreversible — that is designed to stop and hand over to a person.
Approval routing across a team and swipe-to-approve on mobile are in active development. Today the same boundary is enforced through the operation catalogue, credential scoping, spend caps, handoff branches in each workflow, and admin override.
It works where your people already are.
One AI employee, many places to reach it — same procedure, same knowledge, same permissions.
3D avatar
A face-to-face presence for interviews, front desks and kiosks.
Voice
Natural speech with true mid-sentence interruption, plus voice notes in and out.
Web chat
One tag on your own site or portal.
Telegram
Text, photos and voice notes, already in production.
Inside your existing platforms
The AI behind the conversation and service platforms you already run.
More channels
WhatsApp and other messaging apps through partner platforms today; phone calls are on the roadmap.
Already at work.
Real deployments against real systems. Names withheld at partner request; the technical details are unaltered.
AI Knowledge Officer
A university library
Grounded Q&A, room booking and loan renewal, across two cloud regions behind a white-labelled gateway.
AI Recruiter
An HR screening product
Conversational candidate screening with a swappable digital-human avatar.
Platform
A customer-support platform
A third-party service product running its AI features on Wetel's infrastructure.
AI Claims Coordinator
Insurance claims on WhatsApp
A real-time, multi-party conversation state machine, not a single-user chatbot script.
Or try one yourself — every demo on the docs site is a live application built against the same API.
For software companies
Building AI into your own product? Wetel is the infrastructure underneath.
The same runtime that powers Wetel's AI employees also powers partner products in customer conversations, education and recruitment — under their brands, with no Wetel branding anywhere in them. Think of the payments processor, not the checkout page.
Software companies and platforms
Agent capability inside your product, for your customers: a stable, versioned SDK contract, programmatic tenant provisioning, per-customer credential isolation and contract-drift detection. We don't build inbox or contact-centre products, and we won't compete with you for your customers.
Agencies and implementation partners
Reusable workflow patterns across clients, multi-tenant management from one console, white-label branding, and a certified-operation catalogue so you're not re-proving the safety of every integration on every project.
Enterprise platform teams
Standardise AI employees across business units: dedicated-database isolation, single sign-on, audit trails, per-unit spend controls, and a clean answer to “what did it do, when, on whose authority.”
Nothing gets thrown away. The role piloted with one team is the same role serving the whole company two years later — same runtime, same graph.
Where this goes: people and AI employees, working side by side.
We label every capability by tense, and we mean it. If it says today, you can go and check.
Today
- AI employees that answer, look up and act inside your systems through reviewed operations
- Chat, voice, voice notes, 3D avatar, web and Telegram
- Every run traced; drafts tested before they go live
- A supervising agent that delegates background work to worker agents
Building
- Work that starts on a schedule or a trigger
- A supervision view for each AI employee
- Approval routing across a team
- Phone calls, on the same procedures
Next
- Ready-to-hire role templates
- Memory that carries across conversations
- AI employees in different departments sharing one picture of the business
Everything is documented. Including what isn't built yet.
A full integration guide, a complete GraphQL API reference generated from the live schema, a per-step workflow reference with worked examples, and an honest roadmap. Plus a drop-in SKILL.md if your team codes with an AI assistant.
Talk to us
Tell us which work you'd hand over first. Whether you're hiring your first AI employee, building AI into your own product, or standardising across an enterprise — we'll reply with a real answer, not a drip sequence.