Private credit firms want more frequent valuations that give them a clearer view of liquidity and portfolio risk.
AI appears well suited to the task, particularly when the underlying information is dispersed across source documents and disconnected systems. Yet it remains largely absent from valuation.
Our survey of 105 senior private credit leaders found that only 4% currently use AI-based automated valuation models. Third-party pricing services account for 22%. Broker quotes and discounted cash-flow methods each account for 21%.
AI is already being applied elsewhere in the credit lifecycle, so why does valuation remain an exception?
Valuation asks more of the underlying data
AI can be used to extract terms from loan agreements or process information from financial statements, either independently or with manual input. Automated valuation places much greater weight on the output. A valuation can affect portfolio reporting, liquidity assessments, investor decisions, and discussions with counterparties. The information feeding the model must be current and consistent, but firms must also be able to trace each input back to its source and understand how it contributed to the result.
Many firms are still working with data processes that make this difficult. While 36% use AI to extract information from credit agreements and financial documents, 45% rely on semi-automated or hybrid methods involving manual input. Information may pass between documents, spreadsheets, and separate systems before reaching the valuation process. These methods can support periodic valuation, but they are less suited to producing dependable results on a continuous or near-daily basis.
A model cannot compensate for late or incomplete information, particularly when formats vary. Faster calculation does not solve weaknesses earlier in the process.
Confidence depends on being able to explain the result
Investment teams need to understand how a model reached its valuation and explain that result to investors, auditors, regulators, and counterparties. This helps account for the relationship between governance and adoption in our findings. Of those surveyed, 61% have formal policies covering the ethical use of AI in credit decisions, while 29% rely on informal principles. Firms with formal frameworks also expressed greater comfort with automated valuation and underwriting.
A policy does not make a model dependable by itself. Formal frameworks can, however, set expectations for responsibility, testing, version control, escalation, and human review. Helen Wang, our Chief AI and Data Science Officer, describes AI as a means of managing scale “without sacrificing accuracy, transparency, or governance”. Those conditions are especially relevant when a model contributes to the value assigned to an asset.
In valuation, explainability is practical rather than theoretical. If a firm cannot show how an output was produced, tested, and approved, it will be difficult to rely on that output in live portfolios.
Will planned investment close the gap?
Valuation and pricing systems rank third among planned areas for AI-related technology investment over the next three years, selected by 17% of respondents. This indicates that firms want to make progress, but the investment cannot stop at acquiring a model. Dependable data feeds, consistent valuation policies, connected systems, clear records, and responsibility for reviewing exceptions will all influence whether more frequent valuation becomes workable.
The 4% adoption rate does not suggest a lack of interest. The findings instead point to the operational preparation required before firms can rely on automated valuation for decisions that affect investors and counterparties.
