Your officers wear a badge. It scores each interaction against your own service charter, counts real footfall in the hall over Bluetooth, and cancels the visitor's voice on the device before anything is stored. Managers open one screen in the morning instead of waiting for the next assessment.
Nothing is installed at the counter, nothing changes in the queue system, and the officer has no new button to remember. Three parts, and the first one is a piece of hardware you can hold.
Worn on the officer's lanyard. Two microphones: one aimed at the officer, one serving as a reference channel that cancels everyone else. A Bluetooth module counts devices in the hall and the officer's approaches. One shift on a charge, one switch, no app for the officer to operate.
Speech recognition tuned for a busy hall, personal identifiers masked before anything is written to storage, and scoring against the charter your own quality team wrote. It runs inside your perimeter: on-premise or a private cloud hosted in the UAE.
The officer sees their own interactions and where to improve. The centre manager sees the hall. The authority sees every centre measured on one standard, which is the first time those numbers can honestly be compared.
You already have a written standard for what happens at the counter. The platform turns it into a scorecard and applies it to all of them, not to the handful a supervisor happens to observe.
Items and wording are yours. Scoring is calibrated side by side with your quality team until the system agrees with them.
Which step of the standard is dropped most often, in which centre, at which counter, and at what hour of the day.
A dissatisfied tap is joined to the interaction that produced it. Dissatisfaction becomes a ranked list of causes instead of a percentage.
Devices present in the hall and officer approaches, counted anonymously. Peak hours come from people who were there, including those who never took a ticket.
Which language each interaction ran in, where comprehension broke down, and which of those breakdowns produced a second visit.
The documents residents most often arrive without, and the step in the explanation where that information failed to reach them.
How long a service type actually takes at the counter, and how often a visitor is sent to a second counter to finish it.
Three roles, three views, one data set underneath. Nobody is asked to read a report that somebody else assembled by hand.
The officer. Their own interactions with the charter items marked, two growth areas, and a fragment where a colleague in the same hall handled the same case well. Visible to them and their direct supervisor only.
The centre manager. Real load against counters open, handling time by service type, and the language mix the hall is actually serving — which is where an accessibility gap shows up first.
The authority. Centres become comparable because the standard is identical everywhere. The weakest charter item across the network is usually a training decision, not a staffing one.
Everyone. Most avoidable second visits are caused by something written on a portal page, not by the person at the counter. The platform is usually the first thing to say so with evidence.
The officer clips it on at the start of the shift. Directional cancellation removes the visitor's voice on the device. Bluetooth counts presence in the hall anonymously.
Speech recognition tuned for hall noise, across the languages your centre serves. Names, ID numbers and case references are masked before anything reaches the database.
The interaction is matched against your charter, then joined to the rating the visitor gave, the ticket, and the load in the hall at that moment.
Three role views, alerts on the items that matter, and an export into the reporting the entity already produces — including the evidence an assessment asks for.
Two microphones. One faces the officer; the second is a reference channel for everything else in the hall. The visitor's voice is cancelled on the device itself, before anything is stored or transmitted. What reaches the platform is one side of the conversation.
This is usually the first question and the last objection. It holds because it is a property of the hardware, not a clause in a policy document.
A service centre already has a queue system, a rating device, a case system and an assessment cycle. Each of them holds one column of the picture. The platform supplies the column nobody has.
Tickets tell you when someone was called. The badge tells you how many people were in the hall, how many never took a ticket, and how many left before being served.
The tap stays the resident's own verdict. We attach the interaction behind it, so a dissatisfied rating arrives with a cause instead of a guess.
Service type, outcome and repeat visits join to the conversation, which is what makes "why do they come back" answerable at all.
Evidence on service consistency and staff behaviour accumulates continuously from real interactions, instead of being assembled in the weeks before an assessment visit.
Every metric is available through an API and scheduled exports, so the numbers land in the reports the entity already publishes.
Coaching material comes from real fragments in your own halls. At launch nothing is written to a personnel record — that stays the entity's decision, later.
On-premise, or a private cloud hosted in the UAE. No cross-border transfer. A local language model is available where nothing may leave the network at all.
AES-256 at rest and in transit, PII masking before storage, role-based access, full audit log, and retention periods you define per data type.
Arabic and English as standard, plus the languages your hall actually serves. Mid-conversation language switching is detected rather than lost.
No camera, no fixed microphone, no change to the counter itself. The badge and its charging dock are the entire hardware footprint in the hall.
Components are imported and assembled in-country, which is the practical route for the UAE. Employee safety certification is handled at the scale stage, not for a pilot.
99.9% SLA, model and dictionary updates inside the subscription, and a named engineer through the pilot and the first rollout wave.
The pilot is designed to end with a decision, not with a demo. Six to ten counters in one centre is enough to see whether the numbers hold.
Your written standard becomes scoring items. Service types, document lists and dictionaries are loaded. Nothing is invented by us.
Badges issued, officers briefed, and scoring calibrated against your quality team's own marks until the system agrees with them.
The hall runs normally. Weekly readouts on charter compliance, causes behind ratings, real load and language coverage.
What was found, what it is worth, and a rollout plan by centre. Three levers usually move: counters against real load, avoidable second visits, and supervision that scales.
We do not publish a guaranteed percentage before seeing your hall. The pilot measures the range on your own interactions and reports what it found.
A handful of staged visits and an auditor's judgement, with behaviour that changes because someone is watching. Between assessments the hall is unobserved again.
Every interaction, every day, against one written standard, with no staged behaviour and no external auditors in the hall.
Waiting time and headcount, with no idea what was said. Video in a public hall also carries the heaviest consent and retention burden of any capture method.
Load from anonymous device presence, quality from the officer's own speech. No image of a visitor is captured at any point.
Twelve to eighteen months to a first result, a hardware programme nobody planned for, and speech models the entity then maintains itself.
A working platform, a pilot in eight weeks, calibration with your own team, and model updates inside the subscription.
NUMU Group has been shipping since 2014: 11+ years in production, 16 platforms in operation, clients in 50+ countries.
Frontline employees work in the platform every day across 100,000+ points of service. The government configuration runs on a working core.
Specialists across four development centres from UTC+1 to UTC+9, behind an international group of venture and private investors.
NO NAME SOFTWARE FZCO, a freezone company in Dubai Silicon Oasis. Deployment, support and data stay inside the UAE.
No. The badge carries two microphones and cancels the second speaker on the device, before storage or transmission. What reaches the platform is the officer's side of the interaction. Because nothing of the resident is captured, there is no resident data to disclose, retain or delete.
What decides it is how the scores are used. At launch the system runs in coaching mode: visible to the officer and their direct supervisor, with no entry in a personnel record. Officers also gain an objective appeal route with timecodes, which is usually the first thing they ask about.
Arabic and English as standard, plus the languages your hall serves in practice. The language mix is itself a reported metric, a mid-conversation switch is visible rather than lost, and comprehension breakdowns are flagged instead of inferred.
Nothing physical. No camera, no fixed microphone, no rewiring, no change to the queue system. The officer clips on a badge at the start of the shift; the charging dock sits in the back office.
Inside your perimeter — on-premise or a private cloud hosted in the UAE, with AES-256, role-based access, an audit log and retention you define. A local language model is available where nothing may leave the network at all.
It replaces neither. The rating stays the resident's own verdict and we attach the cause behind it. For assessments, evidence on service consistency accumulates continuously from real interactions rather than being prepared ahead of a visit.
Eight weeks on one centre with six to ten counters, including charter configuration and calibration against your quality team's own scoring. It ends with a report on real interaction quality, causes behind ratings, hall load and language coverage, plus a rollout plan by centre.
A 30-minute walkthrough of the workspace on data from a comparable service centre, then a scoped pilot: one centre, six to ten counters, eight weeks, ending in a report and a rollout plan.
Your email client opens straight away: contact@numuhq.com · we reply within one business day.