AI automation for enterprises, designed for the EU AI Act
We build AI into enterprise processes: document processing, internal knowledge assistants and workflow automation on your own data. We start with a narrow, measurable pilot and design compliance in from day one – risk classification, human oversight and audit logging.
Where projects usually start
The best AI projects start with a repetitive task that is done by hand today – not with the technology.
Manual data entry from documents
Delivery notes, receipts, technical documents and contracts are re-typed into enterprise systems by hand.
Knowledge lives in heads and folders
People spend minutes searching technical manuals, procedures and maintenance history, and onboarding new staff is slow.
Uncontrolled AI use
Employees paste company data into public AI tools, with no policy, permissions or logging.
Compliance uncertainty
It is unclear which AI use case falls into which EU AI Act risk category, and what has to be documented for it.
What we build
We connect large language models (LLMs) to your data and systems, so that the result shows up inside the existing workflow.
Document processing
Data extracted and validated from receipts, contracts, emails and technical documents, loaded straight into your enterprise system.
Internal knowledge assistant
An assistant that answers from your own documents, cites its sources and only sees what the user is allowed to see.
Workflow automation
Incoming requests classified, tasks assigned, summaries and reports drafted – with human approval.
Customer and partner communication
Incoming emails and requests classified, replies suggested and handed over to the right person when the case is complex.
AI inside existing software
Summarisation, search, classification or recommendations built into the enterprise application you already use.
Assessment and AI Act classification
We find where AI pays off, classify the use cases under the AI Act and start with a narrow, measurable pilot.
Clients
Companies we have worked for.
- HELL Energy
- FANUC
- iSTYLE
- ELTEC Holding
- AXIÁL
- Melde
- 4iG
Case studies are shared in a call, with the client’s consent.
How we work
An enterprise approach with low risk: at the end of every stage, you decide whether to continue.
- 01
Assessment
We map the process, your existing systems and your security requirements – under NDA if you wish. You get a written proposal, a schedule and a fixed price for the next step.
- 02
Proof of concept
In a short, fixed-price phase we prove the critical point on real data – an integration, an AI model’s accuracy or a workflow – before you commit a larger budget.
- 03
Pilot in one plant or department
We go live with one plant, department or team. We work in two-week cycles, refine based on user feedback and measure the goal agreed up front.
- 04
Rollout and operations
We roll the proven solution out to further sites and teams, with documentation and training. Defects are fixed free of charge for 12 months after handover.
Technology
We are not tied to a single AI provider: we choose the model by task, cost and data handling requirements.
Language models
- OpenAI
- Anthropic Claude
- Google Gemini
Data and search
- PostgreSQL and pgvector
- document indexing (RAG)
- permission-aware search
Integration
- Node.js
- TypeScript
- REST APIs
- webhooks and scheduled jobs
Compliance: AI Act, GDPR, human oversight
In an enterprise, AI is only usable when it is transparent and controllable. That is why every project follows the same principles.
Risk classification and documentation
Every use case is classified against the AI Act risk categories, and the matching documentation is produced.
Your data does not train the model
We use business APIs whose providers do not use your data for model training.
Human in the loop
Important steps – a payment, or a reply sent to a partner – are approved by a person.
Permissions and audit logging
The assistant only reaches what the user is entitled to; AI operations are logged and personal data is handled under GDPR.
About us
Pandai Tech Ltd has been building custom software since 2022. We are deliberately small: the senior engineer you meet in the first call is the one who builds your system – no chain of account managers and project managers in between, one point of decision and accountability.
We use AI-assisted engineering: machines speed up the routine work, while design, code review and accountability stay with people. That gets you to a working pilot faster than a large integrator would, without cutting corners on documentation.
- A direct, senior point of contact
- From the first call to operations, you talk to the person who builds the system.
- Fixed-price stages
- Assessment and proof of concept are fixed-price; pilot and rollout come with a scheduled quote and no hidden costs.
- 12-month warranty
- Defects in delivered features are fixed free of charge for 12 months after handover.
- English and Hungarian
- You choose the language of specifications, documentation and communication.
We also build and run our own products: SafetyPro for maintenance and health-and-safety records, and Anonimer for whistleblowing.
AI automation: FAQ
- Which tasks are worth automating with AI?
- Tasks that repeat often, are text- or document-based and are done by hand today: extracting data from documents, classification, summarisation, searching internal knowledge and answering recurring questions.
- How does the solution comply with the EU AI Act?
- During the assessment we classify every use case against the AI Act risk categories and shape documentation, transparency and human oversight accordingly. Users know when they are working with AI, and decisions are approved by a person.
- Is our company data safe?
- We use business APIs whose providers do not train models on your data. Access is tied to permissions, operations are logged, and on request we sign a non-disclosure agreement before the first call.
- What if the AI gets it wrong?
- That is why important steps have human review and the assistant shows the source of its answer. During the pilot we measure accuracy and only expand when the results are reliable.
- Can you connect it to our existing systems?
- Yes – AI is useful where the data is. We connect via APIs to your ERP, document management, email or in-house systems.
- How is an AI project priced?
- Assessment and pilot are fixed-price, so you learn what automation delivers at low risk. For the rollout we provide a scheduled quote based on the pilot results.
Request a discovery call
Tell us briefly which process or system the project is about. You will get a personal reply.
What happens next?
- 01
We reply personally and schedule a short call about your goal, existing systems and timeline.
- 02
On request, we sign a non-disclosure agreement before you share any details.
- 03
You receive a written proposal for the assessment – scope, schedule and fixed price – so you can decide on the next step.
Encrypted message
You can send confidential information encrypted with PGP. Download our public key and send the encrypted message to: info@pandaitech.com
Key fingerprint:
971D 4DBE 7197 1FB0 889B 83F0 4081 3C11 0699 9AC0Download PGP public key (.asc)Prefer email? info@pandaitech.com