We identify where AI will create value in your organisation and prepare the implementation roadmap.
We find the use cases and rank them by impact, data sensitivity and difficulty. We prove what works with a small pilot, then roll it out.
The work starts with the AI Index.
First we measure your organisation’s AI maturity with the AI Index and identify priority areas through an R&D needs analysis. The result of these two pieces of work is an AI Transformation Map specific to your organisation.
- 01AI Index
- 02R&D needs analysis
- 03AI Transformation Map
Our process
Each step ends with something concrete in your hands. You decide whether to take the next one.
- 01
Discovery call
We listen to your needs, teams and data rules, and tell you plainly whether there is a sensible first step.
- 02
Assessment
We review use cases, data and existing systems, and present a prioritised proposal.
- 03
Pilot
We try the chosen work in a small, measurable application. Success criteria are written at the start.
- 04
Rollout and training
We extend what works to other teams, train the people who will use it and document the rules.
Scope of services
AI action plan
We talk with your teams to find the work where AI will make a real difference, and assess each item by impact, data sensitivity and difficulty. The result is a plan showing what to do first, what to do later and what not to do at all.
WHAT YOU RECEIVE- A prioritised list of use cases
- A data and risk assessment for each use case
- A step-by-step plan with success criteria
Proof of concept and demo application
We turn a chosen idea into a proof of concept (PoC: a small trial that shows whether the idea works), an application running on your organisation's sample data. You see the result yourself before committing to a larger investment.
WHAT YOU RECEIVE- A working demo application
- An evaluation report against criteria set at the start
- A requirements and cost note for moving to production
Agent setup
An agent is an AI system that takes the steps itself once given a goal, and can connect to your systems such as email, calendar, CRM or files. We set up an agent for a recurring workflow, limit its permissions and add human approval at critical steps.
WHAT YOU RECEIVE- A working agent and a list of connected systems
- Permission and human approval rules
- A user guide and team training
RAG and company knowledge assistant
RAG means AI searches your organisation's documents before answering and bases its answer on them. We build an assistant that shows its source when asked about your procedures, contracts or product information.
WHAT YOU RECEIVE- A knowledge assistant connected to your documents, citing its sources
- A structure for updating documents and managing access
- Accuracy test results and a test question set
Model fine-tuning
Fine-tuning means further training a ready-made AI model on your organisation's examples so it performs a specific task more consistently. We first assess together whether it is really needed, and if so, run the process from data preparation to measurement.
WHAT YOU RECEIVE- A needs assessment compared with alternatives
- The fine-tuned model and documentation of the training dataset
- A before-and-after measurement report
AI governance and KVKK
We put into written rules which AI tools can be used in your organisation, with which data, by whom and for what purpose. We make data classes, transfers and responsibilities visible from the perspective of Türkiye's personal data protection law (KVKK).
WHAT YOU RECEIVE- A corporate AI usage policy
- A table matching data classes to tools
- A short rule card and briefing session for employees
Where a legal opinion is needed, we work with your legal team or lawyer; we do not provide legal advice.
When fine-tuning is needed
- When the same narrow task repeats many times in the same format (for example document classification or fixed-format output).
- When your organisation's style or terminology cannot be achieved with well-written instructions.
- When a smaller model must be specialised for one task, for cost or speed.
- When you have enough clean, labelled examples.
When it is not
- When the goal is to answer from up-to-date company knowledge. RAG fits better; you do not need to retrain the model when the information changes.
- When good instructions and a few examples already give the result you need.
- When there is no labelled data, or preparing it costs more than the task itself.
For most organisations, fine-tuning is not the right first step. We say so plainly.
Frequently asked questions
Where does our data stay?
Before building anything, we classify the data and decide with you where sensitive data will stay: we compare options such as your own servers, your corporate cloud account or hosting in Türkiye, together with their costs. Which data goes to which model is recorded in writing.
Which models do you work with?
We are not tied to any provider. We work with the business versions of commercial models such as Claude, ChatGPT and Gemini, and with open models that can run on your own infrastructure when data must not leave the organisation. We choose based on the task, data sensitivity and budget.
A ready-made model or our own?
For most tasks, a ready-made model with well-written instructions, plus RAG where needed, is enough. Fine-tuning or running your own model makes sense under specific conditions; we try the simpler path first.
How soon will we see results?
It depends on scope, so we do not give a fixed timeline. We start the pilot with a small, measurable task and write the success criteria at the start, so you can see progress at every step.
How do solutions connect with training?
Most projects start with training: the team learns the core concepts and data rules, and pilot ideas come out of the training. We also train the team that will use the finished solution.
Is it compliant with KVKK?
In every project, data inventory, privacy notices, cross-border transfers and access rights are addressed from the start. Where a legal opinion is needed, we work with your legal team.