Skip to content

AI for IT teams: integration, data access and maintenance

We integrate AI into the company's existing systems within agreed access limits, and the decision stays with the IT team and the designated operator.

The first conversation is free. We reply within 3 working days.

At a glance

  • In IT, AI connects to the company's existing systems (API, CRM, databases) and uses only the data covered by the agreed access.
  • AI Focus builds the integration through Implementation; AI Care maintains AI systems built by us or by others.
  • With External AI Department, AI Focus specialists work on assessment, implementation and maintenance every month.
  • Decisions on architecture, data access and stopping a system stay with the IT team and the designated operator.

Where AI helps in IT

Integration with existing systems

  • The model drafts answers and summaries from the data in the company's systems (API, CRM, databases), following the architecture described in Phase 1.
  • The company's IT team approves the architecture before the build.

Data access rights

  • The system gets access only to the data the chosen workflow needs, within the limits set in the specification.
  • The IT team grants the access rights and can revoke them.

Testing changes

  • AI drafts test cases from the approved requirements.
  • The IT team validates the test cases, the expected results and the conditions for accepting a change.

Incidents and changes

  • AI summarizes the incident and change log.
  • The designated operator checks the events and decides which ones need action.

A worked example: agreed access to data

  1. A services company wants an AI assistant to draft replies to customer requests recorded in its CRM.

  2. In Phase 1, the IT team and AI Focus agree on which CRM fields the model can read and what is logged; the model cannot write anything directly to the CRM.

  3. The pilot runs on a real use case, and the designated operator approves each reply before it is sent.

  4. The pilot is judged against the acceptance criteria; the company then decides whether to start the build.

Where the model connects to the company's systems

The model gets only the agreed access to the company's systems, and its actions are logged. The designated operator monitors the log and can stop access. The diagram describes a proposal, not an existing integration.

  1. Company systems: API, CRM, databases
    • Agreed access
  2. Agreed access boundary
    Agreed access
    • Proposed model placement
  3. Proposed model placement
    • Logging
  4. Logging
    • Designated operator
  5. Designated operator
    • can stop access: Agreed access

What stays with a person

  • The company's IT team approves the architecture, data access and acceptance criteria.
  • The specification sets what the system does on its own and what the designated operator must approve.
  • The decision to stop or change a system stays with the designated operator.

Sometimes the answer is no AI. If data moves between systems by fixed rules, a standard API integration or a tested script solves the problem without an AI model.

How to start

  • Implementation

    We build software with AI integrated into your systems (API, CRM, databases), in two phases: a pilot on a real use case, then the build.

    Integration with existing systems: the architecture, the pilot on a real use case and the acceptance criteria, in Phase 1.

    From 960 EUR excl. VAT, per phase

    The Implementation service

  • AI Care

    We maintain existing AI systems, built by us or by others: monitoring, fixes and a monthly report, within the agreed capacity.

    Maintenance of AI systems already in production, built by us or by others.

    From 640 EUR excl. VAT per month

    The AI Care service

  • External AI Department

    An outsourced AI department: AI Focus specialists for assessment, implementation and maintenance, every month, with one contact person.

    AI Focus specialists who work with the IT team every month, with one contact person.

    By offer, no public price

    The External AI Department service

  • AI Courses for Companies

    AI training for employees: a custom course built on your team's work, or our courses run only for your team.

    Each group is paid for separately, at the full price of the course.

    Shared methods for the team that will operate the systems.

    From 2,000 EUR excl. VAT

    AI Courses for Companies

Prices are estimates in euros, excluding VAT. The proforma invoice and the invoice are issued in lei, at the National Bank of Romania (BNR) rate on the day of issue. The agreed scope sets the final price.

For someone who wants to learn on their own: the public course "Getting started with AI: working with Claude".

Next course

Getting started with AI: working with Claude

In person · starts on 23 November

2,016 lei + VAT per ticket

The Getting started with AI course

Main questions

Do you work with the systems we already have?

Yes.

  • Through Implementation we build software with AI integrated into the company's systems (API, CRM, databases).
  • Phase 1 describes the architecture and ends with the specification, a pilot on a real use case and the acceptance criteria.
  • The company then decides whether we start the build.

The Implementation service

Where is the data stored, and who has access to it?

For each system, we document what data the model reads, where the generated output is stored and who has access to it.

  • All of this goes into the Phase 1 specification, which the IT team approves before the build.
  • Personal data is covered below.

Data and GDPR

Do you take over AI systems built by others?

Yes.

  • AI Care maintains existing AI systems, built by AI Focus or by others: monitoring, fixes and a status report every month, within the agreed capacity.
  • Before we send the written offer, we assess the system to establish what maintenance we take on.
  • Major new systems are not included; for those, the right service is Implementation.

The AI Care service

How do we avoid AI vendor lock-in?

The data and the workflow rules stay in the company's systems, and the model is only given access to them.

  • Switching models therefore does not mean moving the data.
  • The work needed to switch depends on the chosen architecture: we describe it in Phase 1, and the IT team checks it before the build.

Other questions

Working with AI in IT

What does the IT team need to prepare before the project?

Access to the systems and data involved and to the people who know them, plus someone who approves the architecture.

  • The exact list is in the written offer, before work starts.
  • This access also affects the duration, which the offer states for each phase.
What data is sent to the model, and which provider receives it?

Only the data agreed for the chosen workflow, as set out in the specification.

  • The model provider is chosen together with the IT team, and the written offer states which licenses are needed.
Who operates the system after handover?

The company's designated operator, who receives the documentation and the operating guide at handover.

  • Ongoing operation is not included in Implementation; AI Care can take it over.
Who approves the acceptance criteria?

The company.

  • The acceptance criteria are written in Phase 1, which ends with the specification, the pilot on a real use case and these criteria.
  • The company then decides whether to start the build.
How are the AI system's actions logged?

Implementation builds in the AI Act requirements that apply to the system from the start: labeling AI-generated content, logging and human oversight, where the law requires them.

  • The specification states what is logged.
  • AI Care then keeps an incident and change log.
  • Formal legal audit: not included.
What can an AI agent do on its own, and what needs approval?

The specification sets this for each workflow: which steps the agent carries out on its own and which the designated operator approves.

  • A person approves any step that has an effect outside the company.

Working with us

Can the External AI Department service replace an in-house IT team?

No.

  • External AI Department is not employment and not staff placement: AI Focus specialists work on assessment, implementation and maintenance every month, with one contact person.
  • Decisions about the systems stay with the company.
How do you handle personal data?

If we process personal data on your company's behalf, we first sign a data processing agreement.

  • Course exercises use examples that are anonymized or cut down to what is needed.

Which AI system would you like us to build, or to take over and maintain?

  1. ConversationThe first conversation is free.
  2. AcceptanceThe offer fixes the scope, the capacity and the responsibilities.
Let's talk

We reply within 3 working days.

Or e-mail us directly at office@aifocus.ro