Tides
Highlights and conversations. Work that draws focus, signals movement, or holds a particular charge.
Tides• Autumn-Winter 2025 • Water
In Conversation with Ascanio Salvidio: Exploring the Latest Trends in the Italian Business Valuation
By Magdalena Zawadzka
M.Z.: Could you share some insights into the emerging trends or methodologies that are gaining traction within the business valuation industry in Italy? How are these new approaches enhancing the accuracy and reliability of valuation outcomes?
A.S.: Rather than the emergence of specific methodological trends, the main novelty has been the publication, in 2015, of the Italian Valuation Standards (Principi Italiani di Valutazione or, in brief “PIV”). PIV have been issued by our national standard setter, which is the OIV (Fondazione Organismo Italiano di Valutazione, www.fondazioneoiv.it). OIV has been established, in 2011, by foremost national entities concerned with the quality of valuations: Assirevi (society of major audit firms in Italy), Aiaf (society of Italian financial analysts), Andaf (society of Italian companies’ CFOs/CAOs), Consiglio Nazionale dei Dottori Commercialisti ed Esperti Contabili (institute of Italian chartered accountants) and by Bocconi University. PIV are largely inspired by the standards issued by the IVSC. But they are enriched by references to our national practice in business administration and finance, as well by the deepening of valuations that are required by certain Italian corporate laws. Since their issuance, PIV have gained a high recognition in the Italian valuers’ community. This, in turn, has resulted and is still resulting in a general improvement not only in the quality of valuations in terms of accuracy and reliability, but also in the transparency of valuation process and communication of results.
M.Z.: Are there any specific regulatory or legislative updates in Italy that have had a significant impact on the business valuation profession? How have professionals like yourself adapted to these changes to ensure compliance and maintain best practices?
A.S.: In 2005, adoption by many Italian companies of International Financial Reporting Standard has, for the first time, brought out a demand for recurrent valuations of goodwill, intangibles and business interests. Regular scrutiny of these valuations by managers, auditors, and investors has, in turn, led to a growth in valuation specialism. In the following years, various legislative and regulatory updates have also directly or indirectly increased demand for valuation services. I am referring to the reforms in the area of bankruptcy proceedings, which, already since 2007 and up to the present, have given increasing importance to non-liquidation solutions for the resolution of business crises. Now, it is clear that the preference of the restructuring of the company over its liquidation depends mainly on its suitability to preserve and create value. It is, therefore, a matter of being able to reliably measure whether this can occur. The technical tools are those of business valuations, although used in the context of more complex solvency and planning analyses. The culture of reliable, accurate and traceable business valuation has also found its way into insolvency proceedings. A further boost to the development of the demand of valuation services of companies and intangible assets has been given in recent years by various tax measures, such as: the recurrent provisions on the alignment of the historical cost of equity investments to their market value and the so-called "patent box" benefits.
Increase of demand of valuation services has also boosted competition and the way valuation professional that would like to keep pace with new developments is, basically, the same in Italy as everywhere: investing time in study and money in information sources.
M.Z.: Italy has a diverse business landscape, ranging from traditional industries to innovative startups. How does the business valuation profession in Italy address the unique challenges and intricacies associated with valuing companies operating in different sectors or stages of development?
A.S.: According to World Bank, Italy ranked, in 2022, at n. 10 in terms of GDP size. While this puts our country among the wealthiest ones, the average size of businesses is smaller than that of most other strong economies. Most Italian businesses are closely held by their founders and/or their heirs and families. This means, in other words, that whether in traditional industries or not, most Italian companies bear a significant risk of management and generational change and this phenomenon represents a true challenge for business valuers. The key to a reliable valuation would be, in this case, the attentive scrutiny of a company’s prospective financial information, rather than trying to increase cost of capital to off-set company specific risk on the basis of additional premia. It is, however, common experience of Italian (and I guess also of many non-Italian) valuers that planning and forecasting is one of the weakest traits of most of our medium and of not few of our large companies. So the true intricacy is to find the way to deliver in an acceptable time a valuation that tries to minimize the recurse to standardizes cost of capital adjustments, when company prospective financial information is poor and your client will not wait weeks or months for you to review and generate the “perfect business plan” (not considering the responsibility that a valuer will take on its shoulders when venturing in the field of corporate planning, which should be the “hortus conclusus” of managers only).
M.Z.: Collaboration and knowledge sharing play a crucial role in advancing any profession. In Italy, are there any notable initiatives, organisations, or networks that foster collaboration among business valuation professionals? How do these platforms contribute to the professional growth and development of practitioners in the field?
A.S.: Business valuers in Italy operate from/through various entities and organizations: international consulting firms, investment banks and funds, chartered accounting firms of various sizes. There is no unique “platform” (a society, a body, etc.) of reference for valuers. In terms of collaboration and knowledge sharing, Italian chartered accountants are, in my view, very prolific in terms of courses and seminars in the field of business valuation. And the only Italian BVPO known to me, “Acova” (www.acova.it), has been created in 2018 by a group of chartered accountants. Finally, there is no Italian valuer designation. Not yet, at least. I believe that the reason is very simple: there are many authoritative valuers credentials world wide and Italians are very practical and do not like to re-invent the wheel.
M.Z.: Lastly, what do you envision as the future direction of the business valuation profession in Italy? Are there any specific areas or aspects that you believe will undergo significant transformations or require further attention in the coming years?
A.S.: The limit of every valuer today is that we live into a world that has never been so rich of information, but when we are instructed to value a company or an intangible asset, we barely have the time to collect what is (supposedly) relevant and to process it. AI developments will likely meet these time and process capacity constraints. Some of my fellow valuers I talk with tend to think that AI will force many out of this industry. I believe, instead, that the demand of valuation services will significantly grow thanks to AI and will attract more people than today in this industry.
M.Z.: Thank you, Ascanio!
This interview was originally published on BVIUK's Resources Blog on 11th July 2023
Tides • Autumn-Winter 2025 • Water
The BVIUK AI Policy Council Update
By Hafiz Imtiaz Ahmad and Ascanio Salvidio
The BVIUK AI Policy Council, also known as the AI Work Group, was established to address the integration and governance of Artificial Intelligence (AI) in business valuation. The Council's primary focus is on the practical application of AI in valuation, rather than its technical development, with an overarching goal to produce a guidance document that offers practical, globally relevant policy perspectives. The group recognises that AI is becoming essential for valuation professionals, and resistance to its adoption is expected to diminish as it becomes a standard tool.
Key participants in the Council meetings include Hafiz Imtiaz Ahmad, Graham Antrobus, Drew Dorweiler, Ascanio Salvidio and Omar Zaman. None of the participants are AI specialists, but they possess varying degrees of exposure to AI applications.
Here's an overview of the Council's key initiatives and the broader context of AI in finance and accounting, drawing from the sources provided and discussed in AI policy council meetings:
BVIUK AI Policy Council's Initiatives
The Council has undertaken several key initiatives as part of its efforts to guide the integration of AI in business valuation:
• Document Repository for AI in Business Valuation A shared folder was created to centralise and organise AI-related materials, aiming to house documents, articles, and research on AI regulation, guidelines, best practices, ethical considerations, examples of AI tools, and discussions on AI transparency and disclosure. A deep search using GPT was conducted to compile a large volume of AI regulations and best practices, which are now being reviewed for accuracy.
• Survey on AI in Business Valuation A survey is being finalized to assess industry perspectives on AI usage, its integration, and challenges among BVIUK members, gathering insights on adoption, reliance on AI-generated calculations, attitudes towards transparency, and ethical concerns. It was suggested that this could become an annual initiative to track AI adoption trends.
• Developing Guidelines for AI Use in Business Valuation The group agreed to draft guidelines on fair use and transparency in AI-driven valuations, disclosure expectations when AI is used, and how to account for the limitations of AI tools in reports. A significant discussion revolved around the extent of disclosure, with concerns that revealing too much could lead valuers to lose their competitive advantage. There was a consensus to avoid strict regulation and instead issue best-practice recommendations to prevent stifling innovation and flexibility. The Council's remit is specifically to produce a guidance document, not general commentary. The evolving structure of the guidance document includes sections on valuation integrity, auditability of AI-assisted valuations, responsibility for AI-generated outputs, and frameworks for transparency and explainability. The goal is to preserve professional judgment while embracing technology, addressing challenges like algorithmic opacity, bias, and versioning.
• Ethical and Practical Considerations of AI in Valuation Concerns were raised due to AI "hallucinations" (incorrect outputs or fabricated references), a lack of transparency in decision-making, and inconsistencies between AI tools (e.g., ChatGPT vs. Claude yielding different results). The group also noted AI's potential to reduce professionals’ cognitive engagement, comparing it to the loss of memorizing phone numbers due to smartphones. A repository will be created for members to share tested AI prompts, report experiences with various AI tools, and provide recommendations for AI-assisted valuation processes. Users must also be aware of security and privacy settings, especially regarding confidential data.
• Proposal for AI Tools and Workflow Integration A structured valuation process flowchart will be developed to identify where AI can assist at each stage and determine which AI tools are reliable for different tasks. A second repository focused on AI tools and real-world use cases will also be created. AI is anticipated to enhance efficiency by speeding up data extraction and preliminary analysis, increase accuracy through standardized methodologies, and lead to better resource allocation by automating routine tasks.
• AI Training and Future Plans The Council aims to help BVIUK members use AI effectively and responsibly, with proposed training topics including AI prompt engineering, AI data verification techniques, and AI best practices for valuation professionals. It is believed that AI is becoming essential, and valuation training programmes must evolve to include AI methodologies.
• Promotion Visual and Keyword Strategy An AI-generated promotional image was created using GPT-4 in real-time for the survey promotion on LinkedIn, described as "clever" and "effective”. A focused list of keywords was also agreed upon to improve visibility across platforms.
• AI Tools Blog or Knowledge Section A proposal was made to establish a dedicated blog or knowledge section for AI tool reviews, either as a public feature in The Valuer or as an internal, member-exclusive newsletter. A public blog could position BVIUK as a thought leader, while an internal newsletter would offer safety from legal challenges if vendors are negatively reviewed. A rotating subgroup system was proposed for weekly or biweekly testing of AI tools.
• Benchmarking External AI Tools An external prompt library and a live GPT tool performance dashboard was introduced for benchmarking emerging GPTs, developing a tailored BVIUK prompt library, and contributing working prompts to a collective database.
AI's Role in Finance and Accounting
AI, particularly machine learning (ML) and large language models (LLMs), are profoundly impacting economic sectors, often referred to as the "electricity of the 21st century" in finance and accounting (Koklev, 2022).
• Applications of AI in Finance and Accounting
o Business Valuation: at academic research level, ML has been used to predict market capitalization of quoted companies, based on financial statements, offering an alternative to traditional methods like Discounted Cash Flow (DCF) and Multiples (Koklev, 2022). ML can also help in identifying most relevant features impacting market capitalization, such as Comprehensive Income from the Income Statement (Koklev, 2022). ML algorithms select explanatory variables and can assist in peer selection (Grbenic & Jagrič, 2023; Koklev, 2022). ML has been used to predict value of unquoted companies, exploiting databases containing prices generated by M&A deals (Grbenic & Jagrič, 2024). At commercial level, a notable example of ML technology applied to predict a company’s value is represented today by the platform launched in 2024 by M2M Capital Inc. (USA).
o Financial Forecasting and Analysis: researchers found that LLMs can process corporate disclosures for earnings forecasts, extracting critical information based on factors like sentence position, clarity, numerical content, and sentiment (Li et al., 2024). And that they can also perform financial statement analysis purely from numerical data, predicting future earnings changes with accuracy comparable to specialized ML models and outperforming human analysts (Kim et al., 2024). AI can also predict stock price and market dynamics (Koklev, 2022; Li et al., 2024; Ming et al., 2024).
o Textual Analysis: LLMs can analyse annual reports for sentiment and complexity scores, which correlate with price reactions and future profitability (Bilinski, 2024emulating equity analysts' skills by learning from earnings call transcripts to develop an "Analyst Insight Score" (Ming et al., 2024).
o Risk Assessment and Fraud Detection: ML algorithms provide superior forecasts for equity betas of private firms or non-traded assets compared to traditional methods (Alanis et al., 2024). AI and ML are used in fraud prevention and detection, although AI can also be used in AI-assisted fraud schemes, such as deep-fake voice technology (AICPA FLS Fraud Task Force, 2024).
o Professional Certification and Knowledge Work: LLMs like ChatGPT-4 and Claude Opus have demonstrated the ability to pass the CPA exam's multiple-choice questions (Zacher & Kuppannagari, 2024).
• Benefits and Advantages of AI AI models can produce more accurate predictions and analyses and perform tasks faster than traditional human analysis (Alanis et al., 2024; Kim et al., 2024; Bilinski, 2024). They are well-suited for handling complexity and large datasets and can uncover non-linear interactions (Koklev, 2022; Alanis et al., 2024; Bilinski, 2024; Ming et al., 2024). Unlike human analysts, AI models are not motivated to provide biased forecasts or exhibit behavioral biases (Li et al., 2024). Additionally, machine learning can be a less costly approach to valuing a company and processing complex disclosures (Koklev, 2022; Bilinski, 2024).
• Limitations, Challenges, and Concerns of AI
o Data security: submitting proprietary and/or restricted information to a cloud-based LLM bears the risk of accidental dispersal and may expose sensitive data to malicious activities (there is no guarantee that an AI cloud-based model will not be “hacked”). Due to the relative novelty of LLMs for commercial purposes and mass use, it is unclear if such a risk is comparable to analogous security problems that are affecting clould based data repositories and services. There is the possibility to install an LLM on premises, but at the cost of lower performance than cloud-based ones.
o Hallucinations and Inaccuracy: A primary concern is AI's tendency to generate factually incorrect or ungrounded information ("hallucinations"), which can be presented authoritatively and deceptively (Bilinski, 2024; CBV Institute, 2024; Magesh et al., 2024). Even legal AI tools produce incorrect information between 17% and 34% of the time (Magesh et al., 2024). Since full responsibility for a business valuation rests entirely with the valuer (who is liable for errors, negligence, and omissions, even when caused by the analytical tools employed), the issue of hallucinations and inaccuracies in LLMs used during the valuation process is particularly serious. At present, any professional intending to rely on generative AI—whether for the entire valuation process or only for part of it—must critically review all the information and its sources, as well as the analyses, summaries, and calculations produced by the LLM. This greatly reduces, at least for now, the anticipated advantages in terms of time and cost for valuations based wholly or partly on LLM-based procedures and models. Experience with existing LLM models shows that hallucinations and errors can be mitigated if the estimation process entrusted to the LLM is broken down into micro-phases or micro-tasks, making it more manageable and avoiding saturation of the model’s processing capacity. But at this stage of technology “perfection” is far from being reached.
o Bias: AI responses are based on patterns learned from large datasets, which can lead to biased or inaccurate information, amplifying human biases (Bilinski, 2024; CBV Institute, 2024).
o Interpretability ("Black Box"): Some AI models may not be easily interpretable, making it difficult to understand the specific factors driving their predictions (Grbenic & Jagrič, 2023).
o Limited Data and Processing Limitations: Neural networks can be ineffective with heterogeneous financial reporting data, and ML models can decrease in effectiveness with small datasets (Koklev, 2022). LLMs lack real-time market data or private access to management, and have knowledge cut-off dates (Li et al., 2024; Ming et al., 2024). They also have token length limits, leading to long processing times for large documents and operational costs (Bilinski, 2024; Li et al., 2024).
o Ethical and Professional Responsibility: Professionals must ensure AI outputs are documented, credible, and defensible, as using AI-generated content without proper disclosure may violate ethical standards (AICPA FLS Fraud Task Force, 2024; Kishel & Ramamoorti, 2024; ABA Standing Committee on Ethics and Professional Responsibility, 2024). Lawyers, for instance, have a duty of competence to understand the technology and ensure its output is accurate, requiring independent verification (ABA Standing Committee on Ethics and Professional Responsibility, 2024; CBV Institute, 2024). Confidentiality and privilege issues arise if sensitive client information is entered into public LLMs (ABA Standing Committee on Ethics and Professional Responsibility, 2024).
• Human Role and Oversight Mandatory verification and fact-checking of all AI-generated content is crucial before publication, especially in professional reports, including verifying citations (AICPA FLS Fraud Task Force, 2024; ABA Standing Committee on Ethics and Professional Responsibility, 2024; CBV Institute, 2024; Kishel & Ramamoorti, 2024; Magesh et al., 2024; Kishel, 2025). Experts must apply their professional judgment to interpret AI findings and ensure due care (CBV Institute, 2024). AI models are primarily seen as complementary tools to assist humans, not replacements for human expertise (Grbenic & Jagrič, 2023; Li et al., 2024; CBV Institute, 2024; Kishel, 2025; Ming et al., 2024). Firms and professional bodies should establish clear guidelines for AI tool use, and professionals need to stay updated on AI advancements (CBV Institute, 2024; ABA Standing Committee on Ethics and Professional Responsibility, 2024; Kishel, 2025).
Challenges in AI-Driven Valuation
Specific challenges highlighted in AI-driven valuation include:
• Data Extraction Limitations: AI tools currently have a 20% error margin in data extraction, which is problematic as financial statements must be 100% accurate in valuations.
• AI-Generated References (Hallucinations): AI often fabricates citations, posing significant credibility risks for valuers, akin to a lawyer facing legal trouble for presenting fake AI-generated references in court.
• Cognitive Engagement Concerns: Excessive reliance on AI tools might weaken critical thinking skills and reduce professionals’ cognitive engagement.
• Shared Folder Access: Operational issues like inability to access shared documentation have also been reported.
Future Outlook and Next Step
The Council continues its work with purpose and clarity, committed to providing robust, practitioner-led guidance on the ethical and practical integration of AI within business valuation practices. Immediate action items include finalizing AI regulation document review, drafting initial AI valuation guidelines, gathering AI prompts and case studies for a repository, and planning an AI-focused training programme for BVIUK members. Input for the guidance document will be refined through smaller subgroup discussions, and a preliminary outline will be circulated internally before the next full Council meeting.
References
ABA Standing Committee on Ethics and Professional Responsibility. (2024, July 29). Formal Opinion 512: The use of generative artificial intelligence in the practice of law. American Bar Association
AICPA FLS Fraud Task Force. (2024, March 29). AI and forensic accounting. AICPA
Alanis, E., Lesseig, V., Payne, J. D., & Quijano, M. (2024, January 25). Can machine learning methods predict beta?
Bilinski, P. (2024, February 10). The usefulness of ChatGPT for textual analysis of annual reports. Bayes Business School, City, University of London
CBV Institute. (2024, June 11). Primer on artificial intelligence: Balancing innovation with the ethical and responsible use of emerging technologies. Chartered Business Valuators Institute
Grbenic, S. O., & Jagrič, T. (2023, December 1). Can artificial intelligence algorithms (ChatGPT & Co) appraise companies? European Association of Certified Valuators and Analysts (EACVA)’s 16th Annual International Business Valuation Conference
Kim, A. G., Muhn, M., & Nikolaev, V. V. (2024, May 20). Financial statement analysis with large language models. The University of Chicago, Booth School of Business
Kishel, M. (2025, February 22). Detecting AI generated content. National Association of Certified Valuators and Analysts (NACVA)
Kishel, M., & Ramamoorti, S. (2024, April 20). Using artificial intelligence tools in professional report writing. NACVA QuickRead
Koklev, P. S. (2022). Business valuation with machine learning. Finance: Theory and Practice, 26(5), 132–148
Li, E., Tu, Z., & Zhou, D. (2024, April). Moneyball: GPT’s playbook for earnings forecasts. Zicklin School of Business, Baruch College
Magesh, V., Surani, F., Dahl, M., Suzgun, M., Manning, C. D., & Ho, D. E. (2024, May 23). AI on trial: Legal models hallucinate in 1 out of 6 (or more) benchmarking queries. Stanford Law School
Ming, J., Malloch, H., & Westerholm, P. J. (2024, May 30). Can ChatGPT replicate analyst recommendations?
Zacher, W. Jr., & Kuppannagari, S. (2024, April 16). Can LLMs pass the CPA exam? Evaluating large language model performance on the Certified Public Accountant test.