6 Keys to AI Adoption in Accounting and Finance

Andrea Wright, CPA Partner, Johnson Lambert LLP
Trends in Accounting, Auditing, and Consulting

Ninety-four percent of executives believe artificial intelligence (AI) will transform their industries within the next five years, and it’s been predicted that AI could add $2.6 to $4.4 trillion in value annually. These findings from Deloitte and McKinsey, respectively, make it clear that AI is rapidly becoming a transformative force with immense potential—if implemented strategically.

Here are six steps growth-minded organizations must consider before adopting an AI strategy.

1. DEFINE CLEAR USE CASES

To prevent your AI initiative from becoming fragmented and not producing the results you expect, you should first start by creating a strategic vision for using AI. This involves a thorough analysis of your business processes to find the areas where AI can be most effective. To start, assemble a cross-functional team to pinpoint these areas and their best use cases and create a detailed roadmap with specific goals, timelines, and key performance indicators that’ll monitor your progress.

Importantly, you’ll need to reimagine your existing business processes with the use of these new tools as the foundation for the business objectives they’ll support. While augmenting your human-centric business processes with AI can be valuable, it’s far more beneficial to truly rethink your business in the context of humans and AI working together in order to leverage AI’s full potential.

2. ENSURE DATA QUALITY

Many organizations face challenges with poor data quality, including inaccuracies and inaccessibility, which can significantly impact AI models regardless of their sophistication. When implementing AI across your organization, it’s important to keep in mind that the effectiveness of AI is heavily reliant on it being able to access high-quality data.

One crucial step to ensure data quality is to establish a comprehensive data governance strategy. This will help maintain the integrity of your data and processes over time. A data governance strategy should include a comprehensive system of policies, processes, standards, and roles that ensure the availability, usability, integrity, security, and compliance of the data used and generated in the course of business. This means having clear answers and controls around:

• What data is collected. For example, client financial records, tax information, internal firm data, and market data.
• Where data is stored. For example, cloud servers, on-premises systems, and third-party platforms.
• How data is processed and used. For example, tax preparation, audit procedures, financial analysis, client communication, fraud detection, and internal operations.
• Who has data access (and under what conditions). For example, specific firm personnel, AI algorithms, and third-party providers.
• How data accuracy, completeness, and consistency are maintained. This ensures “garbage in” doesn’t lead to “garbage out” from your AI tools.
• How sensitive and confidential client data is protected. This ensures adherence to professional ethics and data privacy regulations.
• How long data is retained and securely disposed of. This details how you’re meeting regulatory requirements and managing risks.
• How compliance is ensured. This includes Generally Accepted Accounting Principles, IRS regulations, and data protection laws, among others.

Additionally, data processing (or data cleansing) is a crucial step for improving data quality, particularly when inaccuracies, errors, and omissions will hamper the machine learning (ML) capabilities of AI. Clean data is fundamental to accurate AI-assisted analysis and informed decision making; without it, AI models can produce misleading and/or incorrect results. Therefore, organizations that invest time and resources into cleaning their data—including through standardizing, validating, deduplicating, and analysis—can unlock the full potential of their AI initiatives and drive better business outcomes.

3. LEVERAGE APPLICATIONS AND MIDDLEWARE

Legacy systems pose a significant challenge to AI adoption due to their reliance on older business processes and modalities. While it’s possible to begin AI integration without modernizing these systems, it can be time consuming and expensive. If your organization is in a position where this is the only route forward, consider leveraging custom application programming interfaces (APIs) and middleware to integrate existing systems with AI technology whenever possible—think of it as a bridge that allows each of your different applications to communicate with each other.

Importantly, as each organization’s integration needs are unique, it’s recommended to partner with a digital transformation expert that can help design the right integrations for your specific needs.

4. PRIORITIZE CYBERSECURITY

Remember, AI systems introduce cybersecurity risks—they’re vulnerable to adversarial attacks, data poisoning, and hacking of sensitive algorithms. Organizations should prioritize protecting data from malware and misuse by implementing a robust cybersecurity defense program integrated with their data governance approach.

Additionally, companies that plan on using third-party AI models should take extra precautions when working with vendors. Make sure your third-party management agreements clearly outline data ownership; what can and can’t be used by the vendor; and detail the security, confidentiality, and privacy controls that’ll be used by the vendor. These vendors should also be able to provide an independent audit as evidence that they have adequate procedures in place (at least yearly to ensure their product is meeting expectations).

5. REMEMBER ETHICS

Because AI can amplify human biases and compromise decision making, organizations should consider the ethics of AI use before implementing any new tools or models. For instance, organizations must proactively counter prejudiced data inputs to ensure unbiased AI programming and ML modeling. Annotators can play a crucial role in this by carefully analyzing training data before it’s fed into the algorithm to help prevent biased conclusions.

Further, the processing of sensitive data by AI systems raises ethical concerns about misuse. Organizations can mitigate these risks by implementing strict data governance frameworks, adhering to stringent AI ethics policies, and ensuring compliance with relevant regulations and laws. Designing AI systems with privacy in mind, using techniques like data anonymization and encryption, and maintaining open communication with stakeholders about data usage and protection measures are essential steps in responsible AI adoption.

6. COMPLY WITH AI REGULATIONS

The increasing prominence of AI in data-centric operations has led to a rise in legal regulations, particularly in highly regulated industries like finance and health care. Organizations must comply with these regulations to maintain high privacy and governance standards. As regulations increase, third-party auditors will likely be in greater demand to help ensure compliance. Ultimately, a flexible approach to compliance can help organizations operating in these and other regulated industries meet evolving standards.

Overall, organizations that proactively address potential challenges and risks associated with AI adoption beforehand will have an easier time ensuring their existing business processes can reap the rewards of productivity, efficiency, and growth.

Related Content:

• ‘Discover’ Where to Start With AI: When it comes to implementing artificial intelligence (AI) in your accounting practice, the practical first step should be an AI discovery process. Here’s how CPA firms can get started before the next tax season.
• 6 Strategies for Establishing Robotic Process Automation in Your Organization: It’s time for corporate finance leaders to embrace the automation opportunities that the digital workforce offers.