Vahid Nickmanesh
Founder & CEO AI systems builder · Trader · 12+ years running businessesFounder-led product architecture, implementation, integration and execution.
Operational AI for clinics
Models generate.
Models generate.
OpsBond operates.
OpsBond answers clinic enquiries in text and voice, turns each conversation into a complete request and hands it to the right person on your team. AI understands the customer; your rules decide what happens next.
Text + voice messages
Multilingual
Human-reviewed handoffs
Working Clinic productRecorded demo in English and Arabic
One conversation coreText and voice notes share context and state
Governed foundationsData and task foundations accepted, Sept 2026
Academic technical reviewsThree external computer-engineering academics
Why clinics need it
Every enquiry needs an owner and a next step.
Clinics work hard to bring in enquiries. Too many of them stall between the first message and the appointment.
Slow or missed repliesAfter hours, busy lines and too many channels to watch.
Lost contextDetails scattered across chats, calls and staff shifts.
Incomplete requestsNo clear service, time or contact, so staff have nothing to act on.
Follow-up by memoryWarm enquiries cool down when no one owns the next step.
With OpsBond
- Replies in the customer’s language, by text or voice
- Keeps context and corrections across the conversation
- Builds a complete request: person, service, preferred time, contact
- Names a person on your team for the next step
- Never tells a customer something is confirmed when it is not
How OpsBond works
AI understands the customer. OpsBond follows through.
01
Understand
Multilingual text and voice-message input. Keeps context and corrections: “not Tuesday, Wednesday.”
02
Ground
Answers come from the clinic’s approved facts. Unknown prices or availability stay unknown.
03
Structure
Turns the conversation into a draft request: person, service, preferred time and contact. The customer confirms it.
04
Follow through
Hands the request to an accountable person, with the context and a tracked status.
How a request moves through OpsBond
Customer · textHi, I’d like a skin consultation next Tuesday afternoon.
OpsBond ClinicWe offer skin consultations. I can’t see live availability, so I’ll prepare a draft request for you to confirm.
Customer · voice note
0:05
Actually, not Tuesday. Wednesday is better.
OpsBond ClinicNoted: Wednesday afternoon. What’s the best number to reach you?
Customer · text+971 50 ••• ••17
OpsBond ClinicDraft: skin consultation · Wednesday afternoon · +971 50 ••• ••17. Nothing is registered yet. Send this exact request to reception? (Yes / No)
Customer · textYes
OpsBond ClinicSent to reception. This is not a confirmed appointment yet; the team will confirm the time with you.
RequestHuman review
- Service
- Skin consultation
- Preferred time
- Wednesday afternoon
- Contact
- +971 50 ••• ••17
- Next step owner
- Front desk · review
A request is not a confirmed appointment. It is confirmed only when clinic staff or the clinic’s own system confirms it.
See it running in the real product: watch the demo →
Use cases
Where OpsBond helps first.
We start with the front-desk workflows that lose the most enquiries, and label each one honestly.
Working
After-hours and busy-hour enquiries
Replies when the team can’t, and leaves a complete request ready for the next shift.
Working
Voice-note enquiries
Customers can speak instead of type. Replies come back by voice, in Persian today.
Working
Multilingual conversations
Text conversations in Persian, English and Arabic at tested scope, with context kept across turns.
Working
Booking requests and callbacks
Captures service, preferred time and contact, then hands the request to staff to confirm.
Next
Follow-up on unfinished enquiries
Policy-based follow-up when a customer stops replying before the request is complete.
Next
No-shows, recall and waitlist
Filling empty slots and bringing customers back, under rules the clinic approves.
New vertical
Satellite methane watch for landfills.
Satellite plume detections from Carbon Mapper, UNEP MARS and NASA EMIT become cases with an owner, a field-check deadline, an audit trail and a signed Site Brief. Every number carries its source, capture time and uncertainty.
Satellite watch
Public plume detections near your sites arrive with their source, capture time and uncertainty attached.
Case workflow & approvals
Each detection becomes a case with an owner and a field-check deadline. Approvals and changes stay in the audit trail.
Tender & lead radar
Watches public tenders and leads for landfill gas and methane monitoring work, so the team sees them early.
Voice
Two voice capabilities. One business core.
Voice changes how people reach the clinic. It should not change the business rules, so voice is designed to reach the same governed core as text.
Clinic voice messages
Working · privateA customer sends a voice note. OpsBond transcribes it, continues the same conversation and replies by voice.
Voice note → local speech recognition → Clinic core → context kept → spoken reply
- Spoken reply
- Persian today; more languages next
- Mode
- Asynchronous voice messages, not a phone call
Realtime voice
Built · executedLive microphone conversation with spoken playback, interruption handling, and language selection and locking.
Microphone → turn detection → realtime model → spoken playback
- Runtime
- Persian, English, Urdu, Hindi, Bengali
- Human-tested
- Persian, Urdu, English
- Next
- Phone-line deployment
Control
The smarter AI gets, the more control matters.
AI understands. Your rules decide. Models provide intelligence; OpsBond governs what can become operational truth or a controlled action, so the clinic stays in charge.
AI models & agents
Probabilistic- Understanding
- Reasoning
- Generating
- Proposing
OpsBond control layer
Governed- Facts & state
- Controlled actions
- Policy & permissions
- Human accountability
- Authority & approvals
- Evidence & history
Soft understanding.
Hard authority.
Hard authority.
Language understanding can stay flexible. Business authority must stay governed. Models can improve or be replaced; the rules, approvals and records stay under the clinic’s control.
Trust
Built for high-trust operations.
Clinics cannot afford an assistant that sounds confident and gets the facts wrong. A request is not a result, and a sent message is not a confirmation. These safeguards are part of the product.
Requests stay requests
No appointment is presented as confirmed until staff or the clinic’s system confirms it.
Medical-safety routing
Reported symptoms pause routine booking for medical review. Urgent signs trigger emergency guidance and an urgent case for staff. No diagnosis or prescribing.
Unknown stays unknown
No invented prices, availability or provider details.
A person owns the next step
Every handoff has a named, accountable owner on the clinic’s team.
Revision history, not overwrites
Every change is a new revision that staff can see next to the previous one.
Data scope agreed first
Before any pilot we agree what data is used, which AI and speech providers process it and how long it is kept.
Built & tested
Operational data foundation
Canonical state, history and decisions live outside the model. Accepted and closed in September 2026.
Built & tested
Task foundation
Tasks are created and moved with full history and duplicate protection. Production functional check passed at its current scope.
External academic reviews
Reviewed by computer-engineering academics.
Three academics outside the OpsBond team reviewed selected architecture materials and controlled demonstrations in 2026.
OpsBond represents a serious applied AI project with a research-oriented systems direction.
OpsBond represents a substantive software and AI systems engineering effort.
Maintaining canonical business state outside the language model and orchestration layer is a significant architectural decision.
Quoted with permission. Read the reviews in full →
Working with us
How a pilot works.
One clinic, one workflow, one owner on your team, and a decision you can measure.
Step 1
Discovery call
We look at one workflow you want to improve and who owns it today.
Step 2
Diagnostic
We map the workflow, your approved facts, the data scope and today’s baseline.
Step 3
14-day pilot
One channel, one owner, agreed limits and stop rules from day one.
Step 4
Decision
Continue, adjust or stop, based on the measures agreed at the start.
Stop rules come first. A pilot stops immediately on a false confirmed booking, an unsupported medical recommendation, wrong-recipient data or data use outside the agreed scope.
Questions
What clinics ask us.
Straight answers to the questions we hear most.
Is OpsBond a chatbot?
The conversation is the interface. The product is what happens after it: a structured request, kept context, corrections, and a handoff to an accountable person.
Does it book appointments?
It creates booking requests that the customer confirms and your team finalizes, so no one is told an appointment is booked before it is. Direct calendar integration is next.
Does it give medical advice?
No. It can share general information within a scope you approve. Reported symptoms pause routine booking for medical review, and urgent signs prompt emergency guidance and an urgent case for your team. Diagnosis, prescribing and clinical decisions stay with your clinicians.
Which languages does it support?
Text conversations have been tested in Persian, English and Arabic. Spoken replies to voice messages are in Persian today. The separate realtime voice prototype runs in five languages.
Which channels can customers use?
Telegram today, with text and voice messages. WhatsApp, Instagram and web are next, on the same governed core.
Won’t better AI models make this unnecessary?
No. Better models make OpsBond better, because its understanding layer uses them. But a model’s confidence cannot establish who is allowed to act, whether an approval is still valid or whether an outside system really accepted an action. Those answers come from governed state, policy and receipts. Why OpsBond →
How is this different from AI guardrails?
Guardrails constrain model behaviour, and prompt-level safety matters. OpsBond governs business authority, state and real-world effects: who may do what, for whom, in which state, and what proves it happened. The two are complementary.
Which AI models do you use?
We use leading language and speech models through replaceable adapters. The rules, approvals and records do not depend on any single provider, so models can be upgraded or swapped without changing how your clinic works.
What if a system or person doesn’t respond?
Nothing is reported as done. A request is confirmed only by your staff or your clinic’s own system; until then it stays open with a named owner. The architecture treats “unknown” as a real status to be reconciled, and a retry keeps the same action identity to prevent duplicate execution.
Does it replace our front desk?
No. It is designed to give your team complete requests and a clear next step, so staff spend their time on the conversations that need a person.
Where is our data processed?
Before a pilot we agree in writing what data is used, which AI and speech providers process it, where, and how long it is kept. Please do not send patient records through this website.
Where we’re going
Clinics are where we start.
The same governed core is built to carry revenue recovery, multilingual voice and, over time, a control layer for any AI that acts inside a real business.
Starting now
Revenue Recovery
Every enquiry gets an owner and an outcome.
Built · executed
Multilingual Voice
Natural spoken conversations on the same business core.
Platform
Trust & Control Layer
The governed layer AI models and agents pass through before they act.
Team
The people behind OpsBond.
Select a person to see their profile.
Dr. Mohammad Amin Shayegan
Head of AI Research & Evaluation PhD, University of Malaya · Associate Professor of Computer EngineeringEvaluation methodology, benchmark design, reliability analysis and technical review.
Behnam Shayegan
Technical Product EngineerOperational analytics, KPI modelling, commercial and unit-economics support.
Request a pilot
Start with one workflow worth measuring.
One clinic, one workflow, one operational owner and one measurable decision.