Digital Public Services: Smart-DLWD Artificial Intelligence (AI) Project

Digital public services connect data infrastructures, services and goods that facilitate social participation. Smart-DLWD investigates how public service providers can gain low-barrier access to artificial intelligence.

The consortium analyzes processes at EDG Entsorgung Dortmund GmbH, develops AI services and tests them in real-world operations. A taxonomy, a maturity model and a series of white papers make the results transferable to other municipal utilities. The central research question is: How can AI move from a pilot project into reliable operational practice?

Smart-DLWD aims to make artificial intelligence practical for use in public services. To achieve this goal, the consortium identifies operational use cases, develops data-driven AI services, integrates them into existing IT landscapes and tests their benefits under real-world conditions.

The project team has identified several key areas for improvement. According to IW Consult, the public sector could face a shortage of approximately 840,000 skilled workers by 2030. At the same time, modernizing IT structures, building technical expertise and securing adequate budgets could expand organizations’ capacity to act.

An internal project survey of more than 100 participants from the public services sector confirms this picture: 90 percent view legal uncertainty regarding sensitive data and the General Data Protection Regulation (GDPR) as a key challenge; approximately 80 percent cite data quality and system integration; and three-quarters highlight the need for additional staff. More than 70 percent seek guidance on identifying suitable use cases. Smart-DLWD addresses these needs by combining technological development with practical guidance for public service providers.

These results are available for reuse by public service providers:

• A taxonomy with nine characteristics for self-assessment and comparison of service offerings
• A maturity model for assessing an organization’s own artificial intelligence (AI) readiness
• A workshop format for identifying AI use cases
• Tested prototypes from five areas of day-to-day municipal operations
• The fifth result is a three-part white paper series that provides practical guidance. White Paper One, Fundamentals and Context, can be downloaded here.

Download the white paper

“The Smart-DLWD project demonstrates how artificial intelligence (AI) influences digital public services: In our use cases—which cover chatbots and AI agents, image processing in vehicles, street cleaning and waste management planning as well as digital credentials—we not only demonstrate general functionality but also measure the positive impact of AI solutions.”
Daniel Hefft conducts research on AI in public services at the Fraunhofer Institute for Material Flow and Logistics IML.

Smart-DLWD operates according to an impact logic that can be applied to other organizations:

1. Describe the service offering—nine characteristics of the taxonomy enable comparison of the type of service, access requirements, reach and competing objectives.
2. Examine processes and data—current workflows, interfaces and directly and indirectly available data are recorded and evaluated.
3. Select use cases—a workshop format and the maturity model show which projects align with the organization’s AI readiness.
4. Build target processes and data pipelines—a reliable data pipeline enables every model to become operational.
5. Test prototypes in operation—users evaluate utility, comprehensibility and usability and provide feedback that is incorporated directly.
6. Share results—White papers, conference papers and open-source components provide access for municipal enterprises.

This approach provides a field-tested foundation that other municipal companies can use to develop resource-efficient AI applications for digital public services.

You're wondering whether AI can make a valuable contribution to your municipal operations. With Fraunhofer IML, you can:

• Use nine characteristics of the taxonomy to describe service offerings in a comparable way.
• Assess their AI maturity before allocating budgets to use cases that require further development.
• Review proven use cases: cleaning scheduling, vehicle cameras, chatbots and digital documentation.
• Contribute requirements to the white paper series and ongoing validation activities.

The exchange takes place through project formats such as brief coordination sessions, validation discussions and gallery walks focusing on individual use cases. Participation is voluntary.

Request the maturity model now

EDG Entsorgung Dortmund GmbH previously planned street-cleaning routes manually, which required considerable time and effort whenever individual parameters changed. A data-driven planning tool now supports this work. It assigns road sections to teams and weekdays, proposes cleaning schedules and assists with planning new sections. The tool uses scheduling data, administrative guidelines and geodata.

Based on observations of cleaning crews, the project team developed a metric for the effort required to clean a road section. The metric accounts for machine use, trash cans, waste collection containers and travel time between locations, as well as seasonal factors such as falling leaves.

Fairness is a key consideration, particularly the equitable distribution of shifts and rest periods. During the workshop, users reported improved planning for distributed teams, greater transparency and more reliable decision-making. They requested consideration of special cases, a Global Positioning System (GPS) interface and a clear visualization of optimization effects. The approach can be transferred to municipal companies and other organizations that jointly plan teams and capacities.

Waste collection and street-cleaning vehicles travel across nearly the entire road network. Cameras on these vehicles provide a continuous snapshot of road conditions. The system detects, classifies and localizes road signs, construction sites, litter hotspots and road damage. A server pipeline automatically blurs faces and license plates before the images are analyzed. EDG provided an initial dataset of 670 gigabytes for this purpose.

In the Streetwatch use case, Building Information Cloud GLWG GmbH links detected road damage with the City of Dortmund’s excavation and contract data and determines when the damage occurred based on the weekly sequence of images. The Civil Engineering Department processes around 500 cases of road damage per year. Instead of making on-site visits based on suspected damage, staff receive a prepared case containing information on the company, the acceptance date and the remaining warranty period.

Knowledge is stored in documents that are rarely searched: work instructions and guidelines, technical operating documents and maintenance manuals, sorting instructions and hazardous material information, as well as customer-related information, are scattered across various systems and not directly integrated into employees’ day-to-day work. Four assistants developed by IITS address this issue.

An internal knowledge bot for administrative processes consolidates internal files and work instructions, allowing employees to ask questions in their own words instead of having to guess keywords, including on mobile devices. The workshop assistant provides step-by-step guidance for maintaining a small street sweeper and documents the process hands-free. The sorting assistant identifies unknown materials at the recycling center based on sorting instructions and technical rules, now also using voice input and photos. The customer service assistant helps staff quickly find the right information to respond to customer inquiries.

The workshop evaluation produced mixed results. Participants confirmed benefits for onboarding, customer service and situations involving uncertainty, and responded positively to the inclusion of source references. Areas for improvement include error diagnosis, response speed, voice input in noisy environments, and a role and access rights concept. An integrated GenAI platform is intended to consolidate the individual solutions.

Until now, department heads have determined driving authorizations based on their own knowledge. Most certificates were on paper or remained in employees’ minds. Spherity GmbH is building an infrastructure for digital credentials to address this.

A dashboard solution consolidates qualifications based on roles: employees can view their own certificates while department heads can see the status of their teams. A traffic-light filter indicates validity and push notifications remind users of renewals—such as the five modules required every five years. The system supports uploading, verification and digital signing and also links each training or instruction to the corresponding vehicle type. Certificates remain tied to specific employees and can be exported.

The team has further developed the prototype and completed an initial round of feedback. The team will soon address data protection approval, verification via a digital wallet, encrypted transmission, authentication and integration with Systems, Applications and Products in Data Processing (SAP) as the central data source. The solution will also introduce digital product passports for components in the circular value creation process.

Anyone wishing to use artificial intelligence (AI) must first have a clear understanding of their own service offerings. The taxonomy developed in the project describes municipal public services based on nine characteristics:

• Type of service—economic, non-economic or social
• Accessibility barriers
• Regional scale
• Connectivity
• Contribution to territorial capital
• Effective reach
• Access requirements
• Demographic influence
• Priority business objectives

Each characteristic is accompanied by guiding questions. For example, a waste management company might describe itself as a fee-based, non-economic service in a major city with local reach and direct demographic influence. Experts within and outside the sector validate the taxonomy. In addition, a maturity model uses defined characteristics to assess the progress of a company’s AI implementation. Together, the taxonomy and maturity model support comparison and assessment of the current state.

Six partners support Smart-DLWD: two research institutions, three development companies and one end user. This public-private partnership brings together a municipal waste management company, private technology providers and public research institutions.
The Fraunhofer Institute for Material Flow and Logistics IML coordinates the project and contributes expertise in robotics, AutoID, transportation logistics, maintenance and generative artificial intelligence (AI).
The Chair of Business Logistics at TU Dortmund University provides scientific support and investigates platform-based value creation ecosystems.
EDG Entsorgung Dortmund GmbH, with 1,100 employees and more than 350 vehicles, makes its processes available for testing and evaluation.
iits-consulting GmbH is responsible for software development involving large language models (LLMs), retrieval-augmented generation, computer vision and optimization.
Spherity GmbH provides a trusted data infrastructure, while Building Information Cloud GLWG GmbH handles automated document verification.
Associated partners support the transfer of project results.

Fundamentals and Context

The first white paper in the series lays the groundwork: It contextualizes artificial intelligence and digital public services, describes current challenges facing municipal enterprises and presents nine characteristics for describing their range of services.

Initial empirical findings from a survey of more than 100 public service enterprises supplement the conceptual framework.

Download the White Paper