Case studies

What changed for our clients

Nine systems running for clients in six industries. Each one shows the situation before, what we built and the result.

One you can check for yourself: caremyphone.de.

Document processing · Logistics

From 40 hours of data entry to zero

Before: Three people keyed more than 10,000 invoices a month into the ERP by hand, about 40 hours a week, and the backlog kept growing. Two earlier automation attempts had failed.
What we built: A pipeline that reads each invoice, classifies it, checks the fields and routes it into the ERP. Anything that fails a check goes to a person.
Result: Invoices now flow into the ERP at 99% field accuracy, and only the exceptions reach a person. The three people who keyed invoices now work on vendor negotiations.

Saved per year€155K

Manual processing85% less

Show the calculation

Counted in full, the manual process cost about €12.9K a month: processing time, fixing errors, missed early-payment discounts, supplier disputes, oversight and time taken from other work. That is about €155K a year.

Payback: about 10 weeks, on a build of about €28K. See the full breakdown

Built with: Python, Azure Document Intelligence, FastAPI, PostgreSQL

5,000 conversations a day, without hiring

Before: Tickets arrived faster than the team could answer them. Replies took 8 hours on average, two senior agents had left, and a key enterprise client was asking about service levels.
What we built: A support assistant that finds the answer in a live knowledge base and replies in under 2 minutes. Complex cases go to a person with the conversation attached.
Result: Most conversations now close without a person, and the company absorbed its growth without adding seats. The team picks which problems deserve their time.

Saved per year€210K

Resolved automatically78% of tickets

Show the calculation

Avoided hiring 5 additional support agents to handle volume growth. At ~€42K loaded cost per seat: 5 × €42K = €210K/yr in headcount not added.

Payback: about 10 weeks.

Built with: LangChain, GPT-4o, Pinecone, FastAPI, Redis

The workflow engine that replaced 6 spreadsheets

Before: Requests lived in spreadsheets and three email chains, and approvals took days. When something slipped, nobody could trace where.
What we built: One platform that routes each request to the right person, approves automatically where the rules allow, and shows the status of every item live.
Result: The team spends its time deciding instead of chasing, and every request can be traced from start to finish.

Saved per year€105K

Approval timeDays → same day

Show the calculation

6 team members reclaiming ~2h/day previously spent on manual routing and approval chasing. 6 × 2h × 250 days × €35/h (fully loaded) = €105K.

Built with: Python, Celery, React, Docker

Websites and web apps · Analytics

Ask your data a question, get an answer

Before: A business question went to the data team in Slack, someone wrote a database query, and the report arrived the next morning. By then the decision had often been made without it.
What we built: A web app where you type a question in plain English. It turns the question into a database query and shows the answer as a chart.
Result: People get their numbers while the decision is still open, and the data team gets its time back.

Saved per year€78K

Time to an answerNext day → instant

Show the calculation

Recurring BI contractor reports eliminated (€39K/yr) + 12h/week of senior analyst time freed at €65/h × 50 weeks (€39K).

Built with: Next.js, OpenAI, D3.js, Supabase

Websites and web apps · Local business

A phone repair shop that looks like it belongs in 2026

Before: No website, only a Google Maps listing. People who searched found nothing they could trust and went elsewhere.
What we built: caremyphone.de, a bilingual, mobile-first site that answers what customers ask before they call: services, prices, how a repair works and real Google reviews.
Result: Searches now turn into walk-ins and bookings. The site is live, so you can check it yourself.

PageSpeed score98 / 100

Built in3 weeks

Built with: HTML/CSS, JavaScript, responsive layout, SEO

Document processing · Healthcare

From a 3-day backlog to same-day patient intake

Before: Receptionists spent hours a day typing paper forms into the patient record system, field by field. Errors crept into records, and paper files sat around the clinic without a proper GDPR review.
What we built: Forms, handwritten or digital, are read and checked, then sent to the clinic’s patient record system (EHR) through the HL7 FHIR standard. Originals are archived in an encrypted, GDPR-compliant store.
Result: Nobody retypes forms any more, fields are captured at 97%+ accuracy, and the loose paper files are gone.

Saved per year€175K

Intake backlog3 days → same day

Show the calculation

~7.5 FTE-hours/day of transcription eliminated across the clinic network (€47K) + reduced GDPR exposure (€40K risk premium) + same-day processing enables patient throughput gains (€88K).

Payback: about 9 weeks.

Built with: Python, Azure Document Intelligence, FHIR API, PostgreSQL

Month-end close in 4 days instead of 12

Before: Four analysts spent most of every month on the close, working through hundreds of manual reconciliation steps. Nearly a third of entries needed rework because they didn’t match the first time.
What we built: A reconciliation engine that takes in the transaction feeds, matches entries across accounts and puts mismatches in a queue for review. Audit-ready reports come out on schedule.
Result: Matching reaches 99% accuracy, and the analysts spend their time on the edge cases that need a person.

Saved per year€175K

Close cycle12 days → 4 days

Show the calculation

4 analysts × 8 working days recovered per monthly close × 12 months × €365/day = €140K direct labor + restatement and rework costs eliminated (est. €35K). Rework rate before: 30%.

Payback: about 9 weeks.

Built with: Python, pandas, Azure OpenAI, PostgreSQL, FastAPI

800 returns a week, handled in 6 hours instead of 4 days

Before: Every return was checked by hand: the customer’s email, product photos, reason code and policy. Decisions depended on who picked up the ticket, and customers waited days for an answer.
What we built: A returns assistant that reads each request, checks the product photos and applies the policy the same way every time, with a full refund, a partial refund or a rejection with a clear reason.
Result: Only genuine edge cases reach a person, and the returns team went from four people to one managing exceptions.

Saved per year€155K

Time to a decision4 days → 6 hours

Show the calculation

Returns team reduced from 4 FTEs to 1 (3 × €38K = €114K saved) + consistent policy enforcement cuts fraud and chargeback losses (est. €41K/yr).

Payback: about 10 weeks.

Built with: Python, LangChain, GPT-4 Vision, FastAPI, Redis

ISO audit prep, from 6 weeks to 3 days

Before: Across more than 40 production lines, every ISO audit meant weeks of analyst time pulling records, compiling reports and hunting for gaps. Documentation went stale between audits, so each one started from scratch.
What we built: Production data now feeds a compliance system that produces ISO-formatted reports on schedule, keeps a tamper-evident change log and flags documentation gaps before an auditor does.
Result: The next audit had zero findings on documentation completeness.

Saved per year€130K

Audit prep6 weeks → 3 days

Show the calculation

About 48 analyst-weeks of audit prep a year, across 4 analysts, reclaimed at €1,100/wk = €53K + external ISO consultancy eliminated (€65K) + avoided non-conformance penalties (€12K).

Built with: Python, Azure AI, PostgreSQL, React, Docker

Free 30-minute call

Seen a problem that looks like yours?

Tell us about the task. We reply within 24 hours with an honest view on whether automation is worth it for you.

See how we price

“You talk to me, the engineer who builds your system. If AI isn’t the right fix for your problem, I’ll tell you.” Ayman Santeer, Founder

Book a free call

Prefer email or phone?

contact@elanova.ai +31 6 40747312 WhatsApp