The opportunity
Artificial intelligence is developing rapidly. Of particular interest to OSASI is the emergence of increasingly capable open-weight AI models—models whose underlying weights can be obtained and operated independently rather than accessed exclusively through a commercial AI provider.
Rather than every practitioner depending entirely upon separate commercial AI subscriptions, OSASI is investigating whether a shared computing environment could provide members with access to capable open-weight models while allowing the appraisal community to develop its own tools, knowledge resources, evaluation methods and professional safeguards around them.
What we are exploring
OSASI is not proposing to develop or train a foundation AI model from scratch. We are investigating the more achievable possibility of acquiring or renting the computing infrastructure needed to operate existing open-weight models through a secure shared platform.
That infrastructure could potentially support:
- appraisal research and analytical assistance;
- analysis of market and comparable-sales data;
- statistical, geospatial and economic analysis;
- examination and quality review of appraisal reports;
- document and data extraction;
- integration with appraisal software and analytical applications;
- retrieval from professionally curated appraisal knowledge bases;
- development and testing of AI-assisted appraisal methodologies; and
- experimentation with specialized appraisal models, agents and analytical tools.
The objective would not be to replace professional judgement. It would be to investigate how powerful computational tools can be placed directly in the hands of valuation professionals and developed within a culture that understands appraisal evidence, methodology, standards, uncertainty and professional responsibility.
Why a shared approach?
The computing resources required to operate larger AI models can be expensive for an individual practitioner but become considerably more practical when infrastructure is shared across a professional community. A cooperative approach could also provide benefits beyond cost.
OSASI would not need to be tied permanently to one AI company or model.
Models could be tested against appraisal-specific problems, not only general benchmarks.
Infrastructure could be designed around professional confidentiality and data handling.
Model versions, analytical methods, limitations and performance could be documented.
Members and developers could build applications against common infrastructure.
The profession could build evidence about what these systems can—and cannot—reliably do.
An appraisal-specific AI benchmark
OSASI is particularly interested in exploring an independent appraisal AI evaluation benchmark. Instead of asking only which model performs best on general-purpose benchmarks, we could ask: Which models perform best on actual valuation and asset-analysis problems?
A benchmark might include anonymized problems involving comparable selection, market analysis, adjustment support, highest and best use, statistical analysis, report review, data interpretation, reconciliation and identification of unsupported conclusions.
This would allow OSASI to evaluate competing open and commercial models using evidence relevant to our profession. It may prove that a smaller, substantially less expensive model performs extremely well for appraisal work. That should be tested rather than assumed.
A measured first step
Our first objective is simple: determine whether there is sufficient interest within the appraisal community to investigate this seriously.
Depending on the response, an initial project could be modest: rent GPU computing capacity, evaluate several leading open-weight models, develop an appraisal benchmark and conduct a limited pilot with participating professionals.
Actual usage, performance, costs, privacy and security requirements, and member feedback could then inform whether OSASI should continue renting computing capacity, purchase shared hardware, seek institutional partners or pursue another approach.
We would like to hear from you
OSASI welcomes interest from individuals and organizations who may wish to:
- participate in an initial pilot;
- contribute appraisal knowledge or test cases;
- help develop an appraisal AI benchmark;
- contribute technical or software-development expertise;
- advise on privacy, security, governance or professional standards;
- provide access to computing infrastructure;
- sponsor or financially support exploratory work;
- collaborate as an educational, professional, research or industry partner; or
- indicate interest in using such a resource if it becomes available.
Is there a community of valuation professionals interested in collectively exploring open, professionally governed AI infrastructure for our profession? If the answer is yes, OSASI would like to bring that community together and determine what can realistically be built.