IA em finanças: as aplicações de hoje e as possibilidades de amanhã

Technology   |   Jawwad Rasheed   |   Dec 8, 2023 TIME TO READ: 9 MINS
TIME TO READ: 9 MINS

CFOs, like all leaders, face anxiety around ‘FOMO’ (‘fear of missing out’ for the non-Gen Zs out there). This anxiety has materialized more with the advancements of artificial intelligence (AI) and the benefits that technological innovation promises to bring. The value CFOs are looking to realize from AI includes improving the ability to analyze, predict, and uncover important patterns from data and thereby automate work, make informed decisions, compute large quantities of information, and avoid risk.

This ability to perform tasks that traditionally necessitate human intelligence is familiar to finance professionals. The application of AI over the last decade has already helped, for example, augment productivity, improve forecasting, and accelerate decision-making. The evolution in AI capabilities, specifically deep learning and generative AI (gen AI), has reignited the future of finance vision and promoted healthy challenges to the status quo. Technology may not affect all businesses equally and not at the same time. However, gen AI could present significant further opportunities for value creation across all industries and geographies.

What is AI in Finance?

While machine learning is a subset of AI, ‘deep learning’ is a subset of machine learning underpinned by artificial neural networks. Gen AI models leverage the power of deep learning to create new content that exhibits characteristics learned from the training data. The interplay between the three fields allows for improvements and innovations, propelling AI forward.

More on gen AI later. Let’s start with a recap on ‘traditional’ AI applications for finance, which have been used to solve analytical tasks, including classifying, predicting, clustering, analyzing, and presenting data. Opportunities to realize the full benefits of traditional AI applications across finance are still widespread, often hindered by a lack of attention and investment in analytics automation and digitalization initiatives. The important characteristic that may differentiate successful CFOs from others is accepting the role as the pioneer of innovation – constantly learning more about new technologies and ensuring that businesses are prepared as applications rapidly evolve.

AI Use Cases in Finance

Here are three practical examples of AI integration across finance processes where Alteryx is already driving significant benefits:

  • Accounts Receivable: Substantial gains have been made to automate and digitize the invoicing and collection processes – for example, auto-generation and management of invoices based on the digitization of customer orders and payment schedules. Various machine-learning applications have been pioneered and adopted across the order-to-cash cycle. These include improving cash flow projections and working capital levels given the history of invoice payments, improving the predictability of customer payments and default risks, and estimating the probability of customer disputes based on prior interactions and payment data. This has helped to improve cash flow forecast granularity and accuracy and increase efficiency via collections automation. This has enabled more time for customer personalization and root cause analysis.
  • Account and Bank Reconciliations: Given the variety in data forms for bank statements, collecting and extracting data manually is prone to error. Businesses have adopted Optical Character Recognition (OCR) solutions, which automate converting printed documents into machine-encoded texts and highlight errors for further investigation. Transactions that match can be deleted from the reconciling inventory. Using pre-programmed algorithms, machine learning can then fine-tune the automation process by recognizing familiar patterns from previous manual interventions. Based on these discoveries, the system can resolve future unmatched items independently, significantly reducing the number of odd items reported. This minimizes reconciliation process effort and allows more attention to address anomalies and identify potential fraud.
  • Financial Planning & Analysis: The heightened need for agility in planning, forecasting, and impact analysis has accelerated AI integration into FP&A. Automating data extraction, aggregation, and consolidation has been the launch pad for enhanced AI applications. AI has enabled the automation of driver-based analysis and increased precision and certainty in explaining movements in reported numbers, such as month-on-month or plan vs actual variances. Scenario planning can consider a wider spectrum of variables and parameters, with more accurate correlation analysis across multiple scenarios. Businesses have harnessed AI to optimize revenue streams with better analysis and predictability of customer trends that can improve pricing decisions and resource management.

Future Applications of AI Technology

Now, let’s consider some of the gen AI applications pioneered across Finance. CFOs can work closely with their executive peers to shape the AI strategy for their teams as the world enters another ‘iPhone moment,’ a major turning point in our personal and professional lives. Some examples of how gen AI could further revolutionize finance include:

  • Drafting Reports: Consider creating highly reliable first-draft internal management reports or external financial reports with auto-generated narratives to explain the numbers – ready for review before publication. The bonus is prompting AI models to run queries against current standards to gain assurance on compliance for reporting.
  • Fraud Detection and Prevention: AI is key in analyzing vast transactional datasets and identifying patterns indicative of fraudulent behavior, alerting finance teams in real- time, and minimizing potential financial losses. Gen AI integration could continuously learn from new data and adapt its fraud detection capabilities, staying ahead of increasingly sophisticated fraud schemes.
  • Insight Generation: Consider the creation of customizable interactive charts through natural-language queries. Solutions could provide a Q&A chatbot, a chart creation tool that generates charts seconds after receiving a prompt, and a visualization tool that customizes charts using existing code and validates the accuracy of the code.
  • Policy Interpretation: Generative AI could help review extensive collections of existing financial policies, like travel and expense policies. This could help provide initial recommendations for how those policies could be applied, allowing finance teams to evaluate further and refine the recommendations.

Alteryx AI Tools

Alteryx has fully embraced the leap forward with Alteryx AiDIN, the AI engine that infuses machine learning and gen AI across the Alteryx Analytics Cloud platform. With Alteryx Auto Insights, users can create their own ‘Missions’ with automated root cause analysis so explanations can be provided around movement in numbers. This is extremely powerful, given the pain that management goes through when attempting to identify and prioritize drivers behind the numbers, with hours spent on data mining and trend analysis.

The newly released ‘Magic Documents’ in Alteryx Auto Insights goes one step further, leveraging ChatGPT to create an AI-generated analysis and findings summary. Magic Documents will draft communications for users as an email, message, or presentation, removing the burden of extensive management or performance reporting and freeing up the capacity to focus on the data insights.

Framework and Guiding Principles

Gen AI holds the potential to be a revolutionary technology. Still, it doesn’t change foundational principles of economics: a company must generate a return above its cost of capital with investment in technologies. Even with an AI business case defined and cost and resources consideration factored in, deploying AI in finance faces several challenges:

Training Data and Output Validation: The “garbage-in, garbage-out” principle holds for gen AI. If models are trained on incomplete, poorly governed, or even biased data, the output will not be reliable; hence, preparing and validating datasets remains critical. CFOs should institute a process to validate large language model (LLM) outputs to build trust in the models over time. For example, generating synthetic data that emulates existing real-world data is a good way of supercharging the AI learning process. Feeding the algorithms diverse data may help reduce the ‘hallucination’ effect whereby LLMs tend to confidently provide incorrect answers or reinforce stereotypes or exclusions based on the training data set.

AI Transparency and Maintenance: AI can be a black box, especially when using unsupervised learning models. A side effect of this situation is the model generating unpredictable results at unexpected times. Even supervised models can fail to account for unprecedented events. The inability to see behind the curtain and understand how calculations are performed with complete transparency and ‘explainability’ remains a challenge. Furthermore, a machine learning algorithm might be self-learning, but its assumptions should be checked periodically. Failing to maintain the validity of the algorithms may increase the risk of reaching conclusions that are highly incompatible with business strategy.

Security and Data Privacy: Cybersecurity remains a serious concern, as ChatGPT and the equivalents open new avenues for hackers to breach advanced cybersecurity software. In addition, AI’s threat to data privacy and compliance is a risk many organizations are grappling with. Natural language processing technology continually draws on input data to train and improve its output. For many businesses that use these solutions, that might put sensitive, confidential, or proprietary data at risk—making it accessible or searchable to others that use the tool, including competitors. If CFOs want to integrate AI into their day-to-day operations, their teams must understand the security and data privacy risks and find ways to mitigate them.

Finding Qualified Data Scientists: The need for data science skills doesn’t necessarily diminish with the advent of gen AI. ML algorithms can automate several processes in finance workflows, though they still rely on high-quality data and qualified data scientists to monitor everything from data governance to sourcing. Democratization data in your organization will generate insights. However, you cannot guarantee the lack of biases unqualified data analysis professionals may bring to their conclusions. The risk is pursuing the wrong goals and compounding opportunity costs. Data scientist capabilities may need to evolve rapidly to help organizations meet AI compliance and ethical standards.

In conclusion, AI offers CFOs extensive opportunities to revolutionize financial operations, enhance decision-making, and gain a competitive edge across the ever-evolving business landscape. However, CFOs must diligently address the associated risks and challenges of AI. By implementing robust data security measures, promoting ethical AI practices, ensuring compliance, safeguarding against cyber threats, and upskilling their workforce, CFOs can effectively manage the risks and unlock the full potential of AI.

With careful consideration and responsible implementation, AI can elevate the role of CFOs and their finance teams, empowering them to make data-driven decisions and lead their organizations confidently into the next age of innovation. The AI-powered Alteryx analytics platform is well-positioned to help CFOs transition to their future finance vision. What the future holds may only be limited by our imaginations.

Tags