Quicker replies for residents and clients, and a real person always close by.
Our starting point is the officer at the counter, on the phone or in the inbox: case details, proposed replies and call notes prepared for them. Self-service comes later and only where it is safe, with firm boundaries and a handover that passes the whole conversation to staff.
For ministries, local councils and Maltese companies that answer residents and clients by phone, chat, email and letter, in English and Maltese.
A resident or client gets in touch by phone, chat, email or letter. The system works out the purpose, the language and any delicate subject, and opens the case. Delicate or unsupported requests pass to a specialist officer with everything attached. For the rest, the assistant prepares a reply and suggests what to do next, and an officer amends and sends it. Each contact is logged in the CRM and a sample is checked for quality.
Before and after
How the working day changes
Today
With the system
Today
Today: Hunting for information mid-call
To answer one question, an officer flicks between the CRM, the case system, the knowledge base and a thread of old emails.
With the system
With the system: The file is already open
Before the officer greets the caller, the screen shows where the case stands, earlier contacts and the knowledge articles that fit.
Today
Today: Notes and codes after every call
Each contact closes with several minutes spent writing up what happened and picking reason codes, mostly from memory.
With the system
With the system: Proposals, never verdicts
The assistant writes a proposed reply and the call notes; the officer amends them, sends them and remains accountable.
Today
Today: Chatbots that go round in circles
Scripted bots repeat themselves on questions they were never built for, and the resident reaches an officer annoyed and has to explain everything again.
With the system
With the system: A way out at every point
When self-service runs out of road, or the resident simply asks, a member of staff takes over with the full conversation and its purpose.
Today
Today: Language decides the queue
Callers wait longer or shorter depending on whether someone who speaks their language is on duty.
With the system
With the system: English and Maltese in mind
Proposed replies follow the language the resident used, in plain wording, and the officer checks each one before it goes.
Workflow
The workflow, step by step
Help for officers comes first: the assistant prepares, a member of staff sends. Self-service follows later for clearly bounded questions, with the same controls and the same handover route.
Illustrative workflow
7 steps · 2 human checkpoints
Legend
System
AI agent
Person
Human checkpoint
Exception path
Service assistance with its handover route (illustrative)
Read the diagram as text
The main route starts when a resident or client gets in touch, moves on to working out purpose and background and to the assistant’s proposed reply, and reaches the officer who checks and sends it. The contact is then logged in the CRM and may be picked for a quality check.
Nothing is sent before an officer has read the proposal and approved any suggested step: this is where a person decides on the main route.
Requests marked delicate, complicated or unsupported leave the main route right after the purpose is identified and pass to a specialist officer with everything attached. They come back to the main route at the logging step.
01
Step 1: Resident or client gets in touchSystem
Calls (transcribed as they happen), chat, email, portal messages and scanned letters land in one queue. Where a request calls for it, the person is identified through the portal sign-in or Malta e-ID.
02
Step 2: Purpose and backgroundSystem
The system reads the purpose, language and urgency, and marks subjects that must always go to staff, such as complaints, vulnerable callers or legal disputes. The case status is fetched through authenticated APIs.
03
Step 3: Hand to a specialist officerPerson
Delicate, complicated or unsupported requests join a specialist queue together with the transcript, the purpose identified and the case details, so the resident never has to start over.
Exception path: Delicate, complicated or unsupported· Returns to step 6
Human checkpoint
The assistant does not give legal advice and does not decide on entitlements. The service owner, not the model, sets the rules for handing over.
04
Step 4: Proposed reply and next stepAI agent
From cleared knowledge articles, the assistant writes a proposed reply in plain wording and in the resident’s language, and suggests steps it is permitted to prepare: booking an appointment, recording a new address, giving a status update.
05
Step 5: Officer checks and sendsPerson
The officer amends the proposal, accepts or turns down the suggested steps and sends the reply on the channel the resident used.
Human checkpoint
No reply leaves and no step is carried out until an officer has approved it. The officer’s changes are kept to make future proposals better.
06
Step 6: Log and closeSystem
Call notes, the reason for contact and the outcome are saved in the CRM or case system, and the conversation is kept for the retention period agreed.
07
Step 7: Quality checksPerson
Supervisors read a sample of conversations and proposals, note problems and revise knowledge articles. Every change passes the test set before it goes live.
Components
The building blocks
01
Channel and telephony links
Connections to the contact-centre platform, web chat, email and portal, live transcription of calls and a single conversation history whatever the channel.
Case details are fetched only once the person has signed in, and only for that person. The reply assistant can prepare a short, fixed list of steps, each with its own permission, and has no means of editing records itself.
Human review
When the assistant supports officers, a member of staff approves every reply and every step. In self-service the handover triggers are written down: a delicate subject, low confidence, the same question asked again, or simply a request to speak to someone.
The assistant gives no legal advice and takes no decision on entitlements, benefits or claims.
Auditability
We keep the conversation, the proposals, the officer’s changes and every suggested or completed step, with the model and prompt version, so any complaint can be traced back one step at a time.
Data protection
Transcripts and logs are kept only as long as agreed with your data protection officer. Personal data is masked in reporting and is never used to train outside models.
AI transparency
Residents and clients are told when they are dealing with an AI system, in line with the EU AI Act, and may ask for a member of staff whenever they like. A reply an officer sends remains the organisation’s own communication.
Measures
Indicators we would track
We settle these indicators with you in discovery and compare them with the same queues as they run before the assistant arrives.
Indicators we would track
Metric
Why it matters
How we would measure it
01Resolved at first contact
Why it mattersThe plainest evidence that the resident got what they needed and did not have to call back.
How we would measure itFurther contacts on the same case within an agreed period, taken from the CRM.
02Time spent after each contact
Why it mattersNotes and reason codes are where help saves effort without interfering with the conversation.
How we would measure itWrap-up time per contact as recorded by the contact-centre platform, by queue and by channel.
03Handover quality
Why it mattersA handover only works if the specialist officer can carry on without starting again.
How we would measure itThe specialist’s score for the information passed over, and how often a call-back is needed because something was missing.
04Resident and client satisfaction
Why it mattersA fast reply counts for little if the person feels nobody listened.
How we would measure itShort surveys after contact and recurring complaint themes, compared by channel and language.
No targets are set before a baseline exists.
Rollout
A phased rollout
Phase 01
Discovery
Why people get in touch, through which channels, in which languages and with which systems behind. We select frequent, low-risk requests and list the subjects that must always go to staff.
Exit criteria
Request types and handover rules settled
Current performance recorded per queue
Data protection review under way
Phase 02
Pilot of officer support
Background panels, proposals and notes for a single team on one or two channels. Officers send every reply themselves.
Exit criteria
Proposals reach the agreed quality
Officers trained and consulted
Logging and retention working
Phase 03
Supervised self-service
A few request types answered directly in chat or on the portal, with handover to the same team and the full conversation.
Exit criteria
Handover tested from start to finish
AI notice shown to users
Self-service results compared with complaints
Phase 04
Extend
Further request types, channels and languages, each clearing the same test gate.
Questions to a ministry, public entity or local council about applications, permits and benefits, with Malta e-ID sign-in for status checks and plain replies in the official language the resident chose.
Bookings, refunds and administrative questions at private clinics and care homes, with any medical question passed to qualified staff.
Illustrative example
A resident assistant for a local council
Situation
A local council answers residents by phone, email and at the counter about permits, bulky refuse collection, complaints and local services. A small team spends much of the day on the same questions and on writing up notes.
System
An assistant on the council’s website and in the officers’ console prepares replies from the council’s own knowledge articles, in English or Maltese, and drafts the call notes. Status questions are answered after Malta e-ID sign-in.
Human control
Officers send every assisted reply. Complaints, disputes and vulnerable residents always go to a member of staff, with the conversation attached, and residents can ask for an officer at any time.
What we would measure
Time spent after each contact and resolution at first contact per channel, and officers’ scores for the quality of handovers.
Malta
Two official languages, one service standard
In Malta, residents may deal with the public administration in Maltese or in English, and many contacts start on servizz.gov or behind a Malta e-ID sign-in. We design one assistance layer with knowledge articles and handover rules that cover both languages and the sign-in your service already uses, so every resident gets the same quality of reply and the same route to an officer.
Under the EU AI Act people must know when an AI system is talking to them, and the European Accessibility Act, transposed in Malta by S.L. 627.03, sets requirements for consumer-facing digital services. Both are part of the design from discovery onwards, along with GDPR retention rules for transcripts and logs, which the IDPC supervises.
No. The first release supports officers, so they lose less time searching and typing and give more attention to the person in front of them. Self-service is introduced only for tightly defined questions, and a member of staff is never more than one step away.
02How do you weigh self-service against handover?
The handover is designed before anything else. Each self-service request type has written exit conditions and passes the full conversation on. We track how good handovers are and how often people come back, not only how many conversations the assistant closes.
03Can it work in both English and Maltese?
Under Article 5 of the Constitution, anyone may address the administration in Maltese or English and be answered in that language. Proposed replies follow the resident’s language and use plain wording. English works well from the start; for Maltese we test quality on your own conversations first, and officers review every reply before it is sent.
04Must residents be told that an AI system is involved?
Yes. The EU AI Act requires people to be informed when they deal directly with an AI system, and in Malta the MDIA leads market surveillance under that Act. We build in the notice, the route to a member of staff and the logging from the first release.
05Will it fit our current CRM and contact-centre software?
That is the aim. Wherever we can, we build inside the console your officers already use, for instance as components in Salesforce Service Cloud, and connect to your telephony and case systems through their APIs.
06What is the first step?
Look at why people contact you. The most frequent, low-risk requests are a sound first scope for officer support. Read how we handle AI Agents & Automation.
Show us your service channels
Share the most common reasons people contact you, your channels and your languages. We will point out where officer support helps first and where a member of staff must stay up front.