Finding what the company knows
- Today
- Answers sit in manuals, contracts and old email threads. Whoever needs one asks a colleague.
- With HKode
- An assistant that searches your documents and answers with a source reference.
AI integration & automation
AI that actuallyruns in your business.
We bring AI assistants and automation into the systems you already work with. GDPR-compliant, built in Germany, at a fixed price per project.
GDPR & EU hosting
Your data stays in the EU or on your own servers.
Fixed price per project
No open-ended day rates, no surprise invoice.
Straight to the engineer
No sales layer, no hand-off to juniors.
What we connect
Use cases
Where AI saves time in everyday office work
Nothing futuristic. These are tasks that cost hours every week in almost every company.
Sorting the inbox
- Today
- Every enquiry is read, categorised and forwarded by hand.
- With HKode
- Email is classified, assigned and paired with a suggested reply automatically.
Capturing invoices and delivery notes
- Today
- Documents are retyped. Transposed digits only surface at closing.
- With HKode
- Line items are extracted, checked and passed into your system automatically.
Producing quotes faster
- Today
- Every quote is copied together from older files.
- With HKode
- A draft is generated from your CRM and ERP data. You review and approve.
Answering customer questions
- Today
- The same twenty questions take up support time every day.
- With HKode
- A chat on your website that answers only from content you have approved.
Connecting your systems
- Today
- Data is copied between tools because there is no interface.
- With HKode
- Automated workflows connect your programs, traceably and fully logged.
Your case not listed? Describe it in the consultation. We will tell you honestly whether AI is worth it.
Services
Three ways to make AI usable
We do not build models. We bring proven AI to the places in your processes where it measurably takes work off your team.
AI knowledge assistant
An assistant that knows your own documents and answers questions from them, with a source reference, so every answer stays verifiable.
- Chat across contracts, manuals, minutes
- Answers with source and page reference
- Internal assistant for your team
- Customer chat on your website
- Per-department access control
- Connects to your file storage and line-of-business systems
A fit if
Your knowledge lives in documents and nobody finds it fast enough.
Process automation
Recurring work runs by itself. We build on open automation platforms, hosted on your servers or on ours inside the EU.
- Email triage and routing
- Document and invoice processing
- Data hand-off between CRM, ERP and accounting
- Recurring reports and summaries
- Human approval at the points that matter
- Every step logged
A fit if
Your team copies data back and forth between programs.
AI integration & custom software
AI built in where your software already runs, as an interface, an extension, or an application of its own.
- AI features inside your existing software
- AI agents for multi-step tasks
- Interfaces and APIs
- Runs in the EU cloud or on-premise
- Built to EU AI Act requirements
- Web apps and portals around your AI solution
A fit if
You have a platform and want AI usable inside it.
The comparison
Hire an AI engineer or contract the project?
Experienced AI engineers are expensive, hard to find and rarely busy all year. For most companies, contracting the project is the better route.
| Kriterium | Permanent hire | Project with HKode |
|---|---|---|
| Cost | approx. €6,000 to €7,000 gross salary per month¹, ongoing | Fixed price per project, agreed up front |
| Time to start | 3 to 6 months to hire and onboard | Consultation this week, start in two weeks |
| Commitment | Open-ended employment contract | The project ends when the result is delivered |
| Focus | A full-time role that has to stay occupied | Focused solely on AI and automation |
| Utilisation | Full-time, AI project or not | You pay for the project, not for attendance |
| Afterwards | The knowledge leaves when they do | Documentation, training, and the code stays yours |
¹ Typical gross salary for experienced AI engineers in Germany (2026), before employer contributions.
How it works
Four steps and you decide after each one
No annual contract, no consulting package bought on faith. Every step produces a result, and you decide whether the next one happens.
Step 1
Free consultation
30 minutes
We listen, ask questions and tell you honestly whether AI is worth it for your case, including when the answer is no.
You get
- An assessment of feasibility
- A rough sense of the effort involved
- A clear recommendation for the next step
Step 2
AI potential check
1 to 2 weeks
We look at your processes and data properly and prioritise what delivers the most value.
You get
- Workshop with your team
- Three assessed use cases
- An implementation plan
- A fixed-price quote for the pilot
Step 3
Fixed-price pilot
4 to 8 weeks
One use case goes live, not as a demo, but in real everyday operation.
You get
- A working solution in use
- Data protection documentation
- Training for your team
- Hand-over of code and access
Step 4
Operation & further work
monthly, cancellable
We keep the solution running and add further processes once the first one has proven itself.
You get
- Monitoring and maintenance
- Model and security updates
- Additional use cases
- A named contact
We quote prices once we understand your case. A list price would only be a guess. After the consultation you receive a concrete fixed price offer.
Data protection & regulation
Use AI without risking data protection
Data protection is the most common reason AI projects never start. So we settle where your data lives and who can see it before the first line of code.
A person decides
Anything that leaves the company or costs money is approved by a human. No unsupervised automation.
What the EU AI Act has required of you since August 2026
The transparency obligations of the AI Act have applied since 2 August 2026. For most companies the effort is manageable, you just have to know about it.
- Users must be able to tell they are talking to an AI. We build the notice in directly.
- AI-generated content has to be labelled as such.
- Your staff need a basic understanding of AI. Training is part of every pilot.
- Most administrative use cases count as limited risk. High-risk obligations only start in December 2027.
- We document what your solution does, which data it uses and who approves it.
We are engineers and not lawyers. We implement and document the technical requirements. The legal assessment stays with your legal advisors.
Who is behind this
You talk directly to the founders
HKode Solutions is a small owner-run company from Saarbrücken. No sales team, no hand-off to juniors.
Harishanth Sivakumaran
Founder
Engineering & delivery
There is a lot of noise around AI right now, and most of it will quieten down again. What stays are systems that quietly do their work day to day. I build those, right where the work in your company already happens.View on LinkedIn
Keerthana Kirupananthan
Co-founder
Operations & client relations
Many companies know they should be doing something about AI, and not where to begin. You do not need to understand the whole technology, only to name one process that costs you time. We work out the rest in a first conversation.View on LinkedIn
Common questions
What companies ask us first
What does a project like this cost?
It depends on the use case, which is why we publish no list prices. The sequence is always the same. After the free consultation you get a rough order of magnitude, after the potential check a binding fixed price offer. You never sign anything whose price you did not know in advance.
Where does our data live?
In European data centres or on your own infrastructure, and that is your choice. We sign a data processing agreement and disclose which service sees which data. Your content is not used to train AI models.
Can this work without any cloud at all?
Yes. For sensitive data we run the whole solution on your premises, with models on your own hardware. It takes more effort and is usually somewhat slower, but it is technically feasible. We will tell you in the consultation whether it is worth it for your case.
How long until we see something?
The potential check takes one to two weeks. A pilot then goes live in four to eight weeks. So within roughly two months you have something running in real daily operation.
What data do you need from us?
As little as possible. For a knowledge assistant, the documents it should answer from. For an automation, access to the systems involved with the lowest possible permissions. What we do not need, we do not want to hold.
We have no IT department. Does this still work?
Yes, and that is the normal case for our clients. We need one contact who knows how the work actually flows, not someone with technical skills. Setup, operation and training are on us.
What if we want to carry on alone after the project?
Then you can. Code, configuration and documentation belong to you and are handed over. We work with open, widely used tools so you are not dependent on us.
What if the AI makes things up?
That is why we do not build free talking assistants. Ours answer only from documents you have approved and cite a source for every statement. If the system finds no basis, it says so instead of guessing.
Contact
Book a free consultation
Tell us briefly what it is about. We reply within one working day, with an honest assessment and not a sales pitch.
How to reach us
The form is fastest. If you would rather write or call, everything you need is here.
info@hkode.de
Address
Gartenweg 6, 66130 Saarbrücken
Languages
German and English
In-person meetings
Possible on site in Saarland
Just mention it in your message.
Free consultation
30 minutes, no obligation. We look at your case and tell you whether and how AI pays off for it.
FeasibilityOrder of magnitudeData protectionNext step