Artificial intelligence and automation have become the most significant forces behind modernizing record management, shifting it from a labor-intensive, error-prone activity to an engine for organizational insight and value creation. The story of Lutheran Social Service (LSS) of Minnesota offers an instructive, verifiable example of this transformation, illustrating both the promise and complexity of deploying AI workload automation. This case study not only demonstrates the tactical efficacy of recent advances in optical character recognition (OCR), large language models, and process automation—but also exposes the strategic opportunities and ongoing risks facing organizations wrestling with decades, or even centuries, of accumulated documentation.
Legacy records have long been an Achilles’ heel for organizations—storing invaluable historical data, regulatory proofs, and institutional memory in formats that are fragmented, unstructured, and, often, practically inaccessible at scale. Lutheran Social Service of Minnesota, a nonprofit founded in the 19th century with over 2,500 staff serving 91,000 individuals each year, exemplifies this challenge. With a mission centered around community, self-determination, and abundant living, LSS faced the immense task of accessing and making sense of a documentary record spanning more than seventy years, estimated at millions of pages in diverse states of preservation.
A critical stumbling block was the lack of searchability and structure within its massive SharePoint repository. The archive included handwritten letters, scanned forms, and microfilm-based documents—formats that digital tools historically struggled to parse and catalog. Staff members frequently found themselves opening file after file to chase down single facts, a process that was as time-consuming as it was unreliable. According to Microsoft Copilot’s own estimates, it would have taken one individual 28.5 years just to read through the documents—time no organization can afford. Employees felt frustrated, data requests took weeks to fulfill (if they could be completed at all), and even essential compliance reporting was an uphill battle.
The outcome was a bespoke document digitization framework—leveraging Microsoft’s AI Builder for advanced OCR, Azure OpenAI for data summarization, and Power Automate for robust, repeatable process workflows. The result? LSS could transform millions of paper-based, handwritten, and even faded or visually degraded documents into a standardized digital format. This solution enabled fundamentally new capabilities, including:
Heidi Leach, the organization’s senior director of information technology, underscored the practical impact of these changes: “Once we saw that AI could extract meaning from scanned documents—even handwritten or low-quality ones—it changed how we thought about our data. It was not just about managing an archive anymore. It was about reclaiming the stories, decisions and context that had been hidden for years.”
Rather than creating dependence on external consultants, the solution was deliberately designed as a platform for ongoing, internal innovation. LSS team members can now independently scan, summarize, and catalog archival records, building automated workflows that organize data from every page. This positions the nonprofit to continually add value to its operational data and extend the framework into new areas—such as compliance, service delivery tracking, and grant reporting.
Quantitative benchmarks cited by LSS executives and validated in pilot rollouts include:
Ben Vollmer, a director at RSM US, highlights a critical nuance: “Our solution augments the work of the association’s employees… Humans still have to go through the files and validate the information, but we have cleaned up the data so they can find it quicker, use it better and focus more of their attention on pressing issues in the community.”
The technology landscape is evolving rapidly, yet the LSS story reminds us that the value of these tools is always contextual: derived from how thoughtfully organizations align them with their unique missions, data landscapes, and cultures. Done right, AI and automation don’t just digitize history—they make history usable.
Source: RSM AI and automation transform record management | RSM US
Legacy records have long been an Achilles’ heel for organizations—storing invaluable historical data, regulatory proofs, and institutional memory in formats that are fragmented, unstructured, and, often, practically inaccessible at scale. Lutheran Social Service of Minnesota, a nonprofit founded in the 19th century with over 2,500 staff serving 91,000 individuals each year, exemplifies this challenge. With a mission centered around community, self-determination, and abundant living, LSS faced the immense task of accessing and making sense of a documentary record spanning more than seventy years, estimated at millions of pages in diverse states of preservation.A critical stumbling block was the lack of searchability and structure within its massive SharePoint repository. The archive included handwritten letters, scanned forms, and microfilm-based documents—formats that digital tools historically struggled to parse and catalog. Staff members frequently found themselves opening file after file to chase down single facts, a process that was as time-consuming as it was unreliable. According to Microsoft Copilot’s own estimates, it would have taken one individual 28.5 years just to read through the documents—time no organization can afford. Employees felt frustrated, data requests took weeks to fulfill (if they could be completed at all), and even essential compliance reporting was an uphill battle.
A Strategic AI-Driven Approach: RSM and Microsoft in Action
Recognizing that incremental improvement would not solve the scale and diversity of their document challenges, LSS of Minnesota turned to RSM US LLP, a leading provider of consulting services for nonprofits, for an enterprise-class solution. RSM’s engagement began not with a technological quick fix but with a carefully facilitated two-day workshop. The goal was not only to survey the technical landscape, but crucially, to reframe LSS’s archival dilemma as an opportunity for renewed operational insight and agility.The outcome was a bespoke document digitization framework—leveraging Microsoft’s AI Builder for advanced OCR, Azure OpenAI for data summarization, and Power Automate for robust, repeatable process workflows. The result? LSS could transform millions of paper-based, handwritten, and even faded or visually degraded documents into a standardized digital format. This solution enabled fundamentally new capabilities, including:
- Enhanced full-text and metadata search across formerly invisible content layers
- Structured and summarized records, replacing the need for manual “needle-in-a-haystack” searches
- Actionable results that could feed downstream reporting, compliance, and program management workflows
Designing for People: Conversational AI and Sustainable Innovation
Yet automation in record management is not solely a technical problem—it is equally a human one. Realizing this, RSM’s solution extended beyond back-end digitization to the front-end experience, deploying a conversational AI agent developed within Microsoft Copilot Studio. In under a week, RSM was able to prompt-engineer this agent to provide organization-specific answers to staff queries, referencing the newly accessible data corpus. The AI agent achieved an internal accuracy rating of over 85%, quickly earning the confidence of both IT leaders and front-line users at LSS.Heidi Leach, the organization’s senior director of information technology, underscored the practical impact of these changes: “Once we saw that AI could extract meaning from scanned documents—even handwritten or low-quality ones—it changed how we thought about our data. It was not just about managing an archive anymore. It was about reclaiming the stories, decisions and context that had been hidden for years.”
Rather than creating dependence on external consultants, the solution was deliberately designed as a platform for ongoing, internal innovation. LSS team members can now independently scan, summarize, and catalog archival records, building automated workflows that organize data from every page. This positions the nonprofit to continually add value to its operational data and extend the framework into new areas—such as compliance, service delivery tracking, and grant reporting.
Profound and Measurable Results
The immediate efficiency improvements for LSS Minnesota have been substantial. Where information discovery and retrieval once took weeks of staff time, it now takes mere days. Regulatory and compliance reporting—often a dreaded exercise involving manual trawling through poorly indexed documents—has become largely automatic. More importantly, AI-driven automation has freed up limited staff time for mission-critical work: direct service, community engagement, and strategic planning.Quantitative benchmarks cited by LSS executives and validated in pilot rollouts include:
- Scanning and processing of approximately 100,000 legacy files, representing millions of pages
- Return-on-investment (ROI) “accuracy” rates above 85% for the AI-generated summaries and search, a level viewed as “astounding” by leadership
- Service request turnaround time slashed from weeks to days for internal queries and compliance checks
Ben Vollmer, a director at RSM US, highlights a critical nuance: “Our solution augments the work of the association’s employees… Humans still have to go through the files and validate the information, but we have cleaned up the data so they can find it quicker, use it better and focus more of their attention on pressing issues in the community.”
Strategic Advantages: Turning Archives Into Assets
A successful records digitization program driven by AI and automation is not just a tool for hindsight—it is a platform for foresight and proactive decision-making. For LSS of Minnesota, the tangible results go well beyond speed and compliance:- Organizational Agility: Employees now use natural-language search interfaces to surface stories, historical data, and program artifacts that drive fundraising, board decisions, and policy formation.
- Error Reduction: Automated data extraction and standardized recordkeeping reduce the risk of manual transcription errors, missing files, and document loss.
- Transparency: A more accessible archive provides verifiable documentation for audits, stakeholder reviews, and reporting to regulatory agencies—a critical concern for nonprofits reliant on public funding.
- Data-Driven Management: Summarized and structured information fuels business intelligence initiatives, helping LSS identify trends, spot repeat needs, and demonstrate program effectiveness more credibly.
Technical Realities and Cautions: Risks on the Road to Automation
Despite clear benefits, the journey from paper to Copilot is not without risk. Organizations considering similar initiatives should weigh several critical realities highlighted by the LSS experience and independent technical reviews:- OCR Limitations: Even state-of-the-art OCR struggles with severely degraded documents. Results for older handwriting (especially pre-1950) and cuneiform scripts can be inconsistent—even with AI “human-in-the-loop” review.
- Data Quality: If original data is incomplete, ambiguous, or duplicated, AI models can propagate errors or hallucinate summaries. This risk increases with highly sensitive records, such as medical or legal files.
- Privacy and Security: Converting analog records to digital format raises new challenges around data privacy, especially for regulated industries. Robust access controls, encryption, and audit logging are essential when handling sensitive information.
- Change Management: Organizational resistance and unfamiliarity with AI tools can blunt adoption. Success hinges on thorough user training, transparent communication, and clear documentation of AI’s decision-making process.
- Cost and Complexity: Though cloud-based AI tools like Microsoft Copilot and Azure OpenAI minimize upfront infrastructure costs, long-term expenses may include licensing, staff upskilling, and ongoing AI model retraining.
Independent Validation: Cross-Checking the Claims
A review of third-party research and technical documentation confirms the core features and claims made about the LSS transformation:- State-of-the-art OCR and AI text extraction algorithms, such as those available through Microsoft’s AI Builder and Azure OpenAI, have demonstrated higher than 80% accuracy rates on diverse handwritten and scanned documents in recent technical benchmarks.
- Conversational AI agents developed with Microsoft Copilot Studio are increasingly being adopted by nonprofits and enterprises alike, with documented ROI and efficiency gains.
- Industry case studies validate that the most successful AI/automation rollouts embed both technical guidance and staff enablement, allowing for sustainable improvement beyond the initial “pilot” phase.
Broader Implications: A Glimpse of the Future
The LSS of Minnesota case is more than an isolated success—it stands as a sign of a broader revolution in information governance. As AI continues to mature, similar frameworks will become indispensable across sectors: healthcare, education, government, and the corporate world. For organizations with deep “paper legacies,” AI-powered automation is now the difference between letting knowledge fade in forgotten archives and extracting strategic advantage from every document, letter, and form.The technology landscape is evolving rapidly, yet the LSS story reminds us that the value of these tools is always contextual: derived from how thoughtfully organizations align them with their unique missions, data landscapes, and cultures. Done right, AI and automation don’t just digitize history—they make history usable.
Final Analysis: Strengths, Risks, and What Comes Next
Strengths:- Dramatic efficiency improvements in managing archival records
- Enhanced searchability, compliance, and operational reporting
- Solution scalability and internal sustainability
- Empowerment of staff through user-friendly interfaces and workflows
- Potential for data quality errors without robust validation
- Privacy and regulatory compliance challenges
- Organizational inertia or skill gaps that slow adoption
- Ongoing costs of AI licensing and staff training
Source: RSM AI and automation transform record management | RSM US