Igor Józefowicz / Web & AI
Practical AI implementations
Who this service is for
For businesses where employees search across many documents or repeatedly process similar text. We first check whether AI is needed and whether the available data can support a useful solution.
Where AI can help
- An assistant answering questions from company documents.
- Internal knowledge search with links to sources.
- Classifying requests and organising text information.
- Draft summaries and content for an employee to review.
- Connecting a language model to an existing application.
Answer quality and limitations
We start with representative questions or tasks and agree evaluation criteria. We test accuracy, source references and what happens when information is missing. Important decisions and content need appropriate human review.
Retrieval-augmented generation (RAG) finds relevant company material and supplies it to a model as context. This helps ground answers in documents but does not eliminate errors. A conventional search tool may be a better choice for simple retrieval.
Data, permissions and running costs
Before a prototype, we agree which data may be processed, who may access it and whether it can be sent to an external model provider. User permissions must also apply to retrieval and generated answers.
Storage, provider terms, operation logs and error handling depend on the implementation. We discuss model usage, hosting and maintenance costs separately.
Start with a small prototype
We choose one task and an agreed dataset. A prototype tests quality and costs before a broader integration. If it falls short of the criteria, we adjust the approach or leave AI out of that workflow.
How I work with you
Direct collaboration
You work directly with me, from understanding the problem through to delivery. We agree priorities and scope together.
Technology that fits the need
I start by looking at your current tools. An integration or a configured off-the-shelf product may be enough.
A manageable first step
We can start with one workflow or a prototype. We plan further development around your business needs and feedback from users.
Applications and AI together
I combine application development with machine learning, proposing AI where its usefulness and quality can be evaluated.
A clear goal. A considered process.
How we can work together
- 01
Understand your workflow — you describe the process, the difficulties and the tools you use.
- 02
Choose an approach — we consider existing tools and agree scope, pricing and how to evaluate the result.
- 03
First version — I build a selected part of the application or a prototype, and we gather feedback from users.
- 04
Launch and next steps — we put the solution into use, cover day-to-day operation and agree maintenance arrangements.
Contact
What takes up too much of your team’s time?
igor@jozefowicz.plTell me how your company works today and what you would like to improve. You do not need a technical specification. We can work out whether to connect existing tools, automate a task or build an application. Scope, pricing and maintenance arrangements are agreed before work starts.
A few sentences are enough to start
- What does your company do?
- What is getting in the way?
- Which tools do you use?
- What would you like to improve?