Deniz Bektas AI Agents & Assistants
Multilingual assistants that answer questions, pull information from internal systems and hand over to a human with full context when it matters.
AI & Automation · Data Science
If it repeats, it can run itself.
I'm Deniz — a data scientist who spent almost a decade automating processes where being wrong is expensive: credit risk at Deloitte, data analytics in risk management. These days my work is mostly AI agents and the kind of automation that still behaves on a bad day.
- Ex-Deloitte · Credit Risk & ML
- M.Sc. Business Mathematics
- Frankfurt am Main
- Lead handling
- Customer service
- Invoicing
- Onboarding
- Reporting
- Scheduling
- Follow-ups
- Data entry
- Approvals
- Document review
- Email triage
- Forecasting
- Bookings
- CRM updates
Focus
What I work on
The problems I keep coming back to — from a single assistant to a back office that runs without anyone retyping anything.
End-to-End Process Automation
Invoices, onboarding, approvals, follow-ups, reporting — repetitive workflows rebuilt to run from trigger to result without manual steps.
System Integration
Getting CRMs, ERPs, mailboxes, spreadsheets and internal tools to talk to each other, so data flows instead of being copied by hand.
Document & Data Intelligence
Reading, extracting, validating and classifying documents, emails and contracts automatically — with human review kept where it counts.
Analytics & Decision Support
Dashboards, forecasts and risk alerts that turn raw data into decisions, built with the rigour that regulated risk work demands.
Machine Learning & Risk Models
Credit risk and scoring models in Python, including explainable-AI methods — the field I spent most of my career in.
Why automate
Every hour spent copying data, answering the same question or chasing a follow-up is an hour the work itself never needed. Most of it can run on its own.
Examples
What this looks like in practice
Illustrative scenarios, not client work — to show the shape of the problems rather than claim results.
Sales & Real Estate
Instant response to every inbound lead
- Challenge
- Leads wait hours for a reply — and contact the competitor in the meantime.
- Solution
- An AI agent replies within seconds, qualifies budget and timeline, and books meetings straight into the sales calendar.
Response time from hours to seconds
Hospitality & Healthcare
24/7 booking and service assistant
- Challenge
- Front-desk teams spend their day answering the same questions.
- Solution
- A multilingual assistant handles bookings, pricing and FAQs, and hands complex cases to a human with full context.
Fewer repetitive calls, more bookings after hours
Finance & Back Office
Hands-free document processing
- Challenge
- Invoices and contracts are typed into systems by hand.
- Solution
- AI extracts and validates the key data, posts it to accounting and flags only the exceptions for review.
Manual data entry reduced to exception handling
Approach
How I work through a process
- 01
Understand the process
Map how the work actually happens today, not how the documentation says it does.
- 02
Find what is worth automating
Separate the steps that genuinely repeat from the ones that only look like it. Not everything should be automated.
- 03
Build and test
Implement in small, verifiable pieces, with the people who do the work checking each one.
- 04
Measure and adjust
Compare against the manual baseline. If it does not actually save time, it goes back on the bench.
About
Hi, I'm Deniz Bektas.
For almost a decade I have automated complex processes where precision is non-negotiable: as a Data Scientist at Deloitte for leading international banks, and in data analytics for risk management. Data must be right, processes must be auditable and mistakes are expensive.
That background shapes how I think about AI: a demo that works once is easy, a system that still behaves on a bad day is the actual work. I care about the second kind — and about being honest when a process is better left to a person.
-
M.Sc. Business Mathematics · Credit Risk & Data Science
Graduated with distinction (1.2); thesis on modelling market dynamics with neural networks.
-
Deloitte · Data Scientist · Credit Risk & Machine Learning
Automated and modelled credit risk processes in Python for leading international banks, including explainable-AI models.
-
Banking · Data Analytics in Risk Management
Builds the analytics and tooling behind risk reporting, and turns manual risk processes into fast, automated pipelines.
-
Today · AI & Automation
Building AI agents and automations, and writing about what actually holds up outside a demo.
- Precision over speed when the two conflict
- Privacy and data protection by design
- Plain explanations, no black boxes
- Saying when automation is the wrong answer
Contact
Get in touch
Questions, ideas or just want to say hello — write me a line and I'll usually get back to you within a day.
contact@deniz-bektas.com
Frankfurt am Main · Germany