Kapaciti is a Swedish AI automation company in Skellefteå that builds ALEX, an AI employee, and trains staff in AI. These are the questions we get most often, with straight answers. If yours is missing, email hej@kapaciti.se.
ALEX is Kapaciti's AI employee. It reads mail, source material and cases in your systems, prepares replies, reports and registrations, and puts them forward for approval. Nothing is sent, booked or changed until the right person has read it and said yes, and everything is kept in a signed log that your auditor can verify independently. How an AI employee works is explained on the AI employee page.
As of 30 September 2026, ALEX runs in shadow mode on Kapaciti's own server against our own inbox. No customer has ALEX in production yet. The workflows ALEX is meant to take on are described one by one in the module catalogue.
Microsoft 365 Copilot is an assistant in Word, Outlook, Teams and the other Microsoft 365 apps, and it helps whoever asks it in the moment. ALEX holds a role of its own in the business: it reads the inbox and the source material, prepares the work across several systems and puts it forward for approval by the person you have named. The two can be used side by side, and ALEX is built for Microsoft 365. The comparison is on the AI employee page.
Recurring work with clear input and a clear result: the monthly report from the books, replies to routine mail, suggested entries in the case register, deadline tracking and first-line customer service. It is a poorer fit where the judgement is the whole job. The workflows are described one by one in the module catalogue.
No. ALEX connects to the systems you already use, such as Microsoft 365 and Teams, and the connection is set up as part of the rollout. If there is no ready connection to your system, we build it. The connections available today are listed on the AI employee page.
Yes, and in the public sector we start with the case register. ALEX proposes registrations, finds duplicates, tracks deadlines and shows what may be disposed of under your own retention plan, and the registrar approves each proposal before it is entered. The case register exists as a demo today, and building permits, citizen dialogue and zoning plans are the next steps. More on the flow is on the case register page (in Swedish).
An AI employee is an AI system with its own role in the business. It prepares recurring office work in your systems and sends, books or changes nothing until the right person has said yes. The full answer is on the AI employee page.
A chatbot answers questions, and an AI agent carries out tasks in several steps. An AI employee holds a role, prepares the work across several systems and waits for approval from the right person before anything happens. The full comparison is on the AI employee page.
RPA follows a recorded path and stops when something looks different, while an AI employee reads the material and works across several systems. The comparison with RPA and with assistants inside single tools is on the AI employee page.
A thirty-minute call about your week. Then a shadow week where ALEX reads and suggests without touching anything, so you see the result before you decide. You can book the thirty-minute call on the contact page.
Half an hour to begin with, and one person on your side who knows the work and can say where the limits go. After the shadow week ALEX works inside those limits, and every message, booking or change waits for your approval. You book the half hour on the contact page.
The shadow week is the first step. How long the rest takes depends on which workflows you start with and which systems need connecting, and we go through that on the first call.
It depends on what is to be done, at what scale and in which systems. You get the price in writing after the mapping, before anything is booked. The mapping starts with a half-hour call.
You decide who approves what. Anything that is to be sent, booked or changed is prepared, put in front of you and waits until the right person has approved it. Whoever approves sees the whole proposal and can change it or say no. How the approval and the log fit together is described in the Trust Center (in Swedish).
Accountability stays with the business, exactly as when an employee gets it wrong. That is why the approval and the log matter most in the whole solution: every approval is tied to one specific text and one specific person. Everything ALEX does is written to a signed log with an external timestamp that your auditor can verify independently. How the check works is described on the signed log page (in Swedish).
Each customer gets a dedicated server in the EU, and your data is never shared with another customer. If you require it, ALEX runs in your own environment with your own model key. Every sub-processor, including the AI provider and where it operates, is listed openly in the privacy policy (in Swedish). The same list is in the Trust Center (in Swedish).
No. Your data is used only for what you have ordered, and we never train models on it without separate written consent. How we handle your data is described in the Trust Center (in Swedish).
When Kapaciti runs it, ALEX calls the Claude language model from Anthropic. In your own environment an open language model runs locally instead, or Claude with your own key if your agreement allows it. Which model fits depends on the task and your hardware, and the sub-processors in use are listed in the Trust Center (in Swedish).
Your data is kept as the agreement sets out, normally 24 months after it ends, and is deleted when you ask. We confirm the deletion within 30 days. How long each kind of data is kept is listed in the Trust Center (in Swedish).
That depends on which personal data is processed, the data processing agreement and how the use is classified under the EU AI Act. We sign a data processing agreement under Article 28 GDPR with every customer, your data is used only for what you have ordered and every decision requires a human yes. How your use is classified depends on what ALEX is to do for you, and we go through that against the legal text before you decide. The agreement is described in the Trust Center (in Swedish), and the EU AI Act checklist (in Swedish) shows which requirements may apply to you.
In three formats: half a day with the management team, training for staff on their own cases, and open seminars. It takes place at your offices or online, in Swedish or English. Each format is described on the AI training page.
Yes. We teach the working method and practise in the tools you already license, and if you are going to work with ALEX we practise in that too. Which roles the training suits is set out on the AI training page.
Yes, in the form of measures. Article 4 of the EU AI Act requires organisations that provide or use AI systems to take measures to support their staff's AI literacy, taking into account their knowledge, experience and the context the systems are used in. Since the amendment by Regulation (EU) 2026/1744, in force from 27 July 2026, no specific level is required of each individual. The full answer is on the AI training page.
Book half an hour with Alexander. We go through your week and answer what is specific to you.
Newsletter
The EU AI Act and what we learn building ALEX.
We email when something changes that affects Swedish organisations. Reply STOP to any email and you are removed.