Mineko and Data Reply Build AI Data Extraction Tool for Tenfold Acceleration in Throughput on AWS

Founded in 2014, Mineko is a German startup that offers analysis and audits of utility bills to check for legality and offer recommended actions for tenants. With a strong focus on growth and the need to meet increasing demands, the company explored solutions to make its systems more scalable and efficient. The company worked with AWS Partner Data Reply and Amazon Web Services (AWS) to use AI to accelerate data extraction by 10 times, cut data collection costs in half, scale faster, and open new lines of business. Now, Mineko’s systems can perform 66 percent more audits while staff can concentrate on customer experience.

Berlin-based Mineko provides a service for tenants to check whether they are paying the right amount for utilities and service charges. It also works with large corporations to ensure they are paying the correct amounts for commercial properties, as well as with property management companies and landlords to pre-audit their utility bills before they are sent to tenants, helping to reduce tenant objections and ensure accuracy.

The company’s rule-based decision engine, which also on AWS, required some manual intervention. Staff needed to extract data from lengthy tenancy agreements and utility contracts and manually enter that data into the system. Although the decision engine runs quickly and can scale as required, it was slowed down by the manual processes, which also distracted employees from higher-value activities.

Contracts come in various formats, often with handwritten additions, so locating the relevant information can be time consuming. Although this process was supported by an AWS machine learning service that extracts printed text and handwriting from documents, entering documents into the system was taking up too much of the data entry team’s time. Coupled with the increasing demand, this highlighted the need for more scalable systems. When a major property management firm requested an audit of thousands of utility bills, Mineko realized its existing infrastructure needed enhancement to efficiently support such large-scale projects.

Felix Jordan, chief technology officer at Mineko, met some AI experts from Data Reply at an AWS event and they offered a 30-minute consultation to see if AI could help speed Mineko’s business processes. “I was really impressed by Data Reply,” says Jordan. “They’re not just consultants—they get stuff built and on the road. They also really understood how we work and our DevOps culture.”

Collaborating with Data Reply and using AWS, Mineko built an AI system, based on Amazon Bedrock, that can make sense of the necessary data points and feed them straight into the decision engine. Amazon Bedrock provides a single API but can support many of the leading foundation models, so users can pick and choose the resources they need and switch models if required. Data Reply used Amazon Elastic Container Service and AWS Lambda to scale rapidly as required without incurring excessive costs.

The rules-based engine must check two questions relating to German property law before it can determine which set of analytical rules it should apply to that contract. No analysis can be done on the contracts until these two questions are answered. This crucial step is now 100 percent achieved by the AI tool, which removes a major bottleneck in the flow of data.

The project took about 5 months to complete, from proof of concept to production. “To be honest, we were probably the brake on this,” says Jordan. “The timeline could have been even shorter, but at times we had limited internal resources to support the implementation. ”

The AI-powered system is now capable of dealing with 90 percent of with 10–20 percent better accuracy than a Mineko human expert, and data collection times have been reduced by 80 percent. Although the system still requires human oversight, it has increased productivity by a factor of 10. For contracts that are similar to one another—such as those from a large landlord that contain common and repetitive elements in standard utility agreements—Mineko does not even require that human oversight.

The system frees staff to add personalization to reports, focus on complex cases, and, in some cases, override automated decisions. But it means their work is more creative and rewarding. And they can enter and analyze faster than before. Previously, each employee could analyze about 20 contracts or utility bills a day. That number has now increased to 30–40 daily.

The improved process is so good, and so economical—achieving a cost savings of 50 percent for data collection—that it has also given Mineko the technological foundation to pursue entirely new business opportunities. With the ability to reliably scale and handle large volumes of data, the company is now exploring new product offerings that were previously not possible. While details remain under wraps for now, Mineko is excited about the possibilities that lie ahead and is confident in its ability to meet future demand with ease.

The company is now looking at using AI in other areas of its business. It has a project underway to build intelligent search capabilities into its internal document database. Mineko also credits Data Reply for being transparent about how it arrived at solutions. “Our development guys learned so much and gained real confidence,” says Jordan. “That transfer of knowledge means we’re now looking at implementing AI right across the company. We still want to work with Data Reply, but, for the simple stuff, we can just do it ourselves now.”

Based in Berlin, Mineko is an innovative real estate technology company focused on making rental contract and utility bill assessments more accessible and reliable for tenants. Its platform enables private and commercial tenants to upload rental contracts and utility bills and receive comprehensive reports that address critical legal questions, ensure compliance, check accuracy, and offer actionable recommendations. It not only provides suggestions for savings but can also deal with landlord communications.