Joanna Maciejewska's words about AI have resonated widely. Frequently quoted and discussed in various contexts, they have struck a chord with audiences. The application of AI or automation should take place in specific areas to serve humanity, not to completely exclude people from the process. So, where does AI make sense? What is responsible automation?
You know what the biggest problem with pushing all-things-AI is? Wrong direction.
— Joanna Maciejewska—Snakebitten is here. Get it! (@AuthorJMac) March 29, 2024
I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.
Tweet Joanny Maciejewskiej
Undoubtedly, AI, especially in the form of large language models (LLMs), is being used more and more widely, speeding up work for some. However, in many cases and for various reasons, LLMs cannot replace human input. There are also many areas where they should not replace it.
AI vs Humans
Many companies believe in maximizing profits, not by increasing sales or expanding into new markets, but primarily by cutting costs. Hoping LLMs or other generative AI will support them in this effort, they decide to reduce departments like sales or customer service. But what do customers think about this?
According to CGS survey conducted with consumers in the United States a staggering 86% of respondents prefer to interact with a human during the sales process rather than a chatbot. Furthermore, 71% say that the likelihood of using products or services from brands without human customer service drops significantly. On the domestic market, 42.6% of people reported feeling frustrated after interacting with a chatbot, while only 17% experienced positive emotions.
Fot. kreska_ / Agata Krajewska
Whether this is the right direction for AI utilization is becoming increasingly relevant. Can consumer frustrations be alleviated with better AI? Or is this about something much more delicate and elusive, such as empathy? Perhaps we need to rethink how we approach AI, using it where it can provide real benefits and avoid negatively impacting our customers, employees, or communities.
Dehumanizing contact with an organization isn't the best path forward. We know that customer service is difficult, often happening when the end customer is most frustrated. This can, in turn, take a toll on the mental health of customer service staff. Perhaps we should explore how AI could help solve this problem without replacing humans with LLMs.
Where should AI be used?
Although LLMs are not as well-received as "customer service employees" and tend to hallucinate or invent sources, they can still help and bring value to organizations. However, it's worth focusing on other areas than those most visible in mainstream discussions. We can apply generative AI to support our employees internally within the organization—across various processes that are not necessarily related to direct customer service.
According to Velo Bank (ITwiz interview 3-4 (116)/2024, p. 42) applying generative AI in tools supporting analytical, anti-fraud, or risk departments can reduce costs by 10-15% (in the case of a bank, that's savings of around 80-90 million PLN annually).
Technology is becoming crucial in banking, where data volumes are increasing, and decisions must be made based on precise analyses. Large language models (LLMs) can significantly support employees in analytical, anti-fraud, and risk management departments by offering advanced tools for data analysis and information processing.
One of the most important applications of LLMs in banking is the ability to search massive databases using natural language queries. Employees can ask questions to a chatbot, formulating them naturally without needing knowledge of complex SQL queries or other programming languages. An analyst can ask, "Show me all transactions over $10,000 in the last three months," and immediately receive the necessary data. This functionality speeds up analytical processes, reduces the risk of errors, and increases work efficiency, partly because obtaining information in such a way is much more natural and pleasant for our minds.
Another useful application is automatic document summarization. Banks operate on vast amounts of documentation, from contracts to legal regulations. LLMs can automatically generate summaries of these documents, enabling employees to quickly understand their content and catch key information (or even errors). A document that is dozens of pages long can be condensed into a summary containing the most important data and conclusions. This significantly shortens the time needed to review documentation and allows employees to focus on more strategic tasks. For this purpose, it's worth using applications trained on our data sets to create organizational assistants—similar to Nest Bank's. Of course, setting the right temperature for generative AI[1] and catching and correcting hallucinations is crucial.
In the context of anti-fraud efforts, generative AI can detect irregularities and patterns indicative of fraud. These models analyze historical transaction data and behavioural patterns to identify unusual operations that suggest criminal activity. Suppose the system detects sudden and atypical large money transfers from a customer's account. In that case, it can automatically flag such transactions as suspicious and notify the relevant team for further analysis or to contact the account holder.
In risk management, generative AI supports employees in assessing credit, market, or operational risk by analyzing diverse data sources and generating forecasts and recommendations. It can facilitate a quick preliminary analysis of sentiment toward a brand in social media or the press and provide additional information on market moods influencing investment or credit decisions.
In summary, using large language models in banking holds great potential for increasing work efficiency and easing human workload, improving the quality of analyses, and enabling quicker responses to threats. Searching databases in natural language, automatic document summarization, and advanced pattern analysis are just some of the functions that can significantly relieve bank employees in their daily duties.
Unfortunately, these advantages also come with costs—such as energy consumption and increasing an organization's carbon footprint. So, should we adopt generative AI solutions driven by the trend? Should fear of being "left behind" by competitors already brandishing the Artificial Intelligence banner overshadow earlier sustainability commitments? Or can we combine the two? If so, how? Perhaps we should focus not on using AI at any cost but, for example, on sustainable automation supported by AI.
Automate Responsibly
Automating business processes with BPM platforms, like our Flowee, regardless of advanced generative AI solutions, can significantly boost a company's efficiency if applied correctly. According to McKinsey research, companies that successfully implement automation have seen productivity increases of 20-30% and operational cost reductions of 15-20%. Digitization of processes allows for quick and accurate execution of routine tasks, freeing employees to focus on more creative and strategic activities. For example, automating accounting processes can reduce invoice processing time by 60-70%, significantly speeding up document flow within the company.
Fot. kreska_ / Agata Krajewska
In contrast to advanced AI solutions, automation based on BPMs does not require as much computational power, making it less environmentally taxing. Solutions like Flowee are less energy-intensive, which aligns better with sustainability strategies. By consuming less energy, companies can better manage their carbon footprint, which is significant in the face of global climate challenges. For example, using software robots (RPA) to automate processes significantly reduces paper and electricity consumption, allowing for more environmentally friendly resource management.
Despite numerous benefits, automation also presents many challenges. Many companies lack full knowledge of their business processes, making it difficult to implement automation effectively. The absence of process inventory can lead to the automation of the wrong areas, which, in turn, doesn't yield the expected results. Some companies, caught up in the excitement over automation, implement expensive solutions for work performed by just two people, even though the break-even point is, for example, 100 employees.
Another challenge is the need to change organizational culture. Process digitization often requires a shift in mindset among employees and management and adaptation to new technologies. Here, concerns and resistance can always be expected. Listening to the team, identifying its needs, and responding with empathy may be crucial. Implementing automation solutions also requires the right technical skills, which involves additional training and investment in employee development or building an internal IT department.
Fot. kreska_ / Agata Krajewska
Finally, automation requires careful planning and project management. Companies must thoroughly analyze which processes can and should be automated to avoid mistakes and inefficiencies. This process often begins with an inventory and mapping of all processes in the organization,allowing the identification of areas that can provide the greatest benefits. Conducting a thorough analysis and implementation strategy is a step that cannot be skipped if we want to automate successfully.
In conclusion, business process automation can significantly improve companies' efficiency and sustainability but requires careful planning and management. Despite the challenges, the benefits of automation, such as increased productivity and cost reduction, are worth the effort. Automation can partially solve organizational problems without venturing into the uncertainty of generative AI while still leveraging its support.
[1] The "temperature" of generative AI refers, in short, to its level of creativity and specificity. The lower the temperature, the more concrete it is. The higher the temperature, the more creative it becomes.
Editing: Agata Krajewska
