What does AI consultancy from Sirap cover?
We help companies in Malta put AI to work on their own documents and processes, starting with a free 30-minute call by phone or video. After the call we tailor the work to what you need: training your staff to use AI safely, automating one business process with a person approving each step, building a private knowledge assistant on your own documents, or a combination.
Each piece of work starts small, with one team, one process or one document set, and grows once it works.
To get started, book a free call through the form at the foot of this page or call +356 2138 5911. The call is free. Training, automation and the knowledge assistant are price on request, quoted as separate lines after the call.
Free call
We start with a free 30-minute call by phone or video about what you want AI to do, which documents and systems are involved and what data must stay inside the company. If you are not sure where to start, we can follow it with a half-day session at your premises to rank your use cases by effort and value and draft a one-page AI use policy.
Staff AI training
Half-day sessions per team on using AI safely at work: what may be pasted into public AI tools, how to check an answer before relying on it, and how your company's AI policy applies. The examples come from your own documents.
Process automation
We pick one process with you, map it and automate the repetitive steps. A person on your staff approves each output or change before it takes effect, and each step is logged. We widen what the automation may do on its own only when you are ready, within limits we agree.
Private knowledge assistant
An assistant that answers staff questions from your own documents and cites the page behind each answer. It starts as a pilot on one document set, then rolls out in the cloud or on your own premises. On-premise, we supply and set up the hardware, such as the NVIDIA DGX Spark or a Supermicro GPU server with an Eaton UPS.
How does a knowledge-assistant pilot work?
A knowledge-assistant pilot runs for four to six weeks with a small group of your staff, on one set of your documents such as service manuals or operating procedures. Only your staff use it, and we agree before it starts how it will be judged, so the decision to roll out rests on a result.
A company knowledge assistant is an AI tool that answers staff questions from your own manuals, procedures and records, and cites the page behind each answer so the person asking can check it. We built one for our own staff first, and wrote up how we built our company knowledge assistant.
One document set to start
The pilot works from one set of documents your staff search often, such as service manuals, operating procedures or product data. Starting narrow keeps the work measurable and shows within weeks whether the assistant answers your people's real questions.
Documents compiled before they are searched
One fault can be spread across an alert list, a table of causes and a repair procedure three chapters later, so we join those pieces into one page per topic before the assistant uses them. Each page keeps a link back to the manual it came from.
Internal staff only
During the pilot, a small named group of your staff use the assistant and no customer talks to it. Which documents it may read and who may use it are agreed in writing before we start.
A success measure agreed up front
Before the pilot starts we agree how it will be judged. A typical measure is how many real questions from the past month it answers correctly with the right source, as checked by your most senior engineer or manager.
Four to six weeks, mostly spent on documents
Much of the early work goes into getting documents in order: finding current versions, retiring old ones and writing down decisions that so far exist only in email. Your company keeps that order even if the assistant goes no further.
What does staff AI training from Sirap cover?
Our staff AI training is for teams already using, or about to use, AI tools at work. Each team gets a half-day session on what may be pasted into public chat assistants, how to check an answer before relying on it, and how your company’s AI policy applies to them. The examples come from your own documents.
Since 2 February 2025, the EU AI Act has required companies that use AI systems to take measures for their staff’s AI literacy (Article 4 of the AI Act). It asks for no certificate or set course. A session on your own tools and policy, with a record of who attended, is one way to show those measures.
How do you automate a business process with AI, safely?
We pick one process with you and map how it runs today. Then we automate the repetitive steps, and a person on your staff approves each output or change before it takes effect. Each step is logged, so you can see what was done and who approved it. Once that process works, we add the next.
We start with a person approving each step and widen what an AI agent may do on its own only when you are ready, within limits we agree. Our own website runs the same way, and nothing on it goes live until a person approves it. A process automated this way can also be combined with staff training or a knowledge assistant on the same documents.
Why work with Sirap on an AI project?
We run our own staff knowledge assistant, built with the method we propose to clients, and we supply the servers, storage and power an on-premise assistant needs. We have been part of Maltese IT and industry since 1981, and the hardware, its installation and its support come from our team in San Gwann.
Our own assistant is the reference
The compile-and-cite method we use in a pilot has been tried on our own staff assistant first: compiling manuals, citing the source page behind each answer and queueing contradictions for a person to decide. Our team builds knowledge-augmented assistants (KAG) and AI agent workflows for our own business, always with a person approving each change.
We work alongside your IT provider
If you already have an IT provider or an in-house IT team, we work with them. They keep running your systems. We bring the AI work, the hardware and the training.
We supply and support the hardware
We are an authorised Supermicro partner in Malta and the distributor for Eaton in Malta. For an on-premise assistant we specify the GPU server, the storage and the UPS, install them and support them locally.
A person approves every change
Changes to our own website go through a weekly approval batch, and nothing goes live until a person approves it. Changes are logged and checked with before and after screenshots. The process automation we build for clients follows the same rule.
Your documents stay where you decide
Whether the assistant runs on a server in your building or with a cloud provider is settled with you before the pilot starts, from where your documents are allowed to go. We set out both options with their trade-offs before you choose.
AI guides from Sirap
- How we built a company knowledge assistant: what it took to compile our own documents, what the assistant still cannot do, and how a pilot is run.
- Secure AI across the enterprise: the AI risks to brief your staff on, the governance basics of inventory and change tracking, and when cloud or on-premise suits the work.
- On-premise AI compute for regulated iGaming and finance operators: the GPU servers, storage and power an on-premise setup needs, and when the cloud is still the right call.
- Supermicro GPU servers for heavier AI workloads: the air-cooled and liquid-cooled servers for training and larger models, up to a full rack.
For sector detail, see IT infrastructure for iGaming operators, IT for banks and financial services and IT and automation for manufacturers. An on-premise assistant also needs backup and data storage and UPS and power protection.
Should an AI assistant run on your own servers or in the cloud?
An AI assistant can run on your own servers or in the cloud, and the deciding question is where your documents are allowed to go. On-premise keeps the documents on your network and limits you to the models your hardware can run. A cloud provider can offer larger models. For GPU counts, storage and power design, read our guide to on-premise AI servers for iGaming and finance.
Either way, the documents the assistant reads need the same cyber protection and IT security as the rest of your systems.
For a pilot, a compact desk-side AI computer such as the NVIDIA DGX Spark, which we supply and set up, can be enough; we size it from the document set and the number of users. One unit has 128 GB of unified memory and runs models of up to about 200 billion parameters, and two linked units up to about 400 billion.
When the whole company uses the assistant at once, or the same hardware runs other AI work, the next step is a rack server such as the Supermicro X13 4U Universal GPU server. A smaller site can use the Supermicro AS-3015TR-i4 edge GPU tower. Either belongs on a UPS like the rest of the server room, and we fit Eaton power protection.
What can a knowledge assistant not do?
A knowledge assistant cannot answer from experience nobody wrote down, and it does not replace the engineer, accountant or manager who acts on its answer. It is only as current as the documents behind it, so it needs an owner who keeps them up to date and settles conflicting sources.
The assistant can also be wrong. Because each answer cites its source, a wrong one can be traced and the page behind it corrected, but checking answers stays part of the job, and the staff training covers how to do it. For a short handbook that staff can search in a minute, a well-kept shared folder does the job for less, and we will say so on the free call.
Can an AI project in Malta be part-funded by a grant?
It can be, in part. Micro, small and medium-sized companies can apply under Digitalise your SME (Call 2), which covers AI set-up services, software, hardware, cloud and training on the AI solution from the same supplier; consultancy fees are not eligible. MDIA adds a top-up on the AI costs. Each scheme sets its own rates, caps and dates, so check the call before you commit to a purchase.
Staff AI training can be part-funded under the Investing in Skills scheme run by Jobsplus; your company applies at least a week before the session. Companies of any size outside gambling, financial services, wholesale, retail and real estate sales can look at Malta Enterprise’s Business Development scheme for digital transformation, and companies with up to 50 full-time staff at the Micro Invest tax credit for hardware and software. Gambling companies are excluded from Digitalise your SME as well.
Under Digitalise your SME, nothing may be ordered before your grant agreement is signed, and the scheme asks for three quotations or an investment proposal. The MDIA top-up follows a year after the project ends, once you have reported on the results. We supply a grant-ready quote with each eligible cost on its own line, so each line can be matched to the scheme’s categories, and your company makes the application.
Updated October 2026.
AI consultancy in Malta, FAQs
What happens on the free call?
It is a 30-minute call by phone or video. We ask what you want AI to help with, which documents and systems are involved and what data must stay inside the company, so bring someone who can decide that. Afterwards we propose training, a process to automate, a knowledge assistant or a combination, each quoted as a separate line. Book it through the form below or on +356 2138 5911.
Is staff AI training worth it if our people already use public chat assistants?
Yes. That is when company documents start being pasted where they should not. Start with the teams that handle customer, staff or financial data; each team gets a half-day session built on your own policy and documents.
Which documents work best for a knowledge-assistant pilot?
A set your staff search every week and that changes slowly, such as service manuals or operating procedures. PDF, Word, spreadsheets, scanned paper, email threads and web pages all work. Avoid a set with several versions in circulation until one version is agreed, and make sure one senior person who knows the material can judge the answers.
Do we need new hardware for a private knowledge assistant?
Only if it runs on-premise. If your documents may go to a cloud provider, no new hardware is needed. On-premise, a pilot can run on a compact desk-side AI computer such as the NVIDIA DGX Spark, which we supply and set up. A company-wide rollout may need a rack GPU server and a UPS; we size it from the document set and the number of users.
Do you build AI agents that act on their own?
Not at first. We automate one business process at a time, and a person on your staff approves each output or change before it takes effect. We widen what the automation may do on its own only when you are ready, within limits we agree. Automation can be the whole project or sit beside training or a knowledge assistant.
What size of company is AI consultancy from Sirap suited to?
Companies where a few document sets and a handful of experienced staff hold most of the answers. A larger group can start in one department and extend the work once it has a result.
How do you start an AI project with Sirap?
You start by booking a free call, using the form below or by calling +356 2138 5911, Monday to Friday 8:30 to 16:30. Tell us what you would like AI to do, such as training your staff, automating a process or answering questions from your documents. We call you back to set a time for the 30-minute call and send a short list of what to have ready. The call is free; the work after it is price on request, quoted as separate lines.
Which AI models and hardware do we work with?
We work with models from OpenAI and Anthropic, open-weight models that run on your own hardware, and NVIDIA hardware such as the DGX Spark. We choose the model for each job on accuracy, cost and where your data may go, and we can review the choice later.