The Wavelength: Artificial Intelligence

AI is one of the fastest-evolving areas in business today. What are experts in both hardware and software using it for, and what do they have to say about what’s next?

How are you using AI in your business or how can businesses make the best use of AI? 

Adam Field: Internally, each part of our organization is using AI to gain advantages in unique ways. Ultimately, though, it all comes down to augmenting workers so they are more proficient in their day-to-day jobs. For instance, our sales operations team uses AI to predict (with amazing accuracy) quarterly sales success. Our marketing team uses generative AI tools to enhance their content creation. Our engineering team uses AI to test code so that it’s bug-free out of the gate. As a company, we continue to inject AI across our product portfolio to help our customers transform unstructured data into formats that help them gain insights, allowing for faster and more accurate decision making. Personally, I’ve been using AI to summarize those all-too-lengthy email threads that I don’t have time to read!

Bob Lamendola: We are leveraging AI, specifically machine learning (ML) and natural language processing (NLP), in several ways within Ricoh. All our workflow automation solutions take advantage of intelligent scan and capture machine learning capabilities to actively improve automated workflows by making routing and document classification decisions. Our support processes also leverage AI capabilities to drive auto resolution and self-service chatbot capabilities through the consumption of product documentation and actual resolution knowledge created within our systems. We enhance our RFP responses by leveraging NLP capabilities as well, pulling together information from prior responses and related innovation content.

West McDonald: In my business strategy and AI-focused enterprise, AI is integrated across all operations. I employ it for editing, tailoring outreach to specific buyer personas, and crafting visuals and voiceovers for media content. Even my LinkedIn headshots and website photos have been AI-generated and refined. For clients, I offer both a traditional written AI readiness assessment and an interactive chat version for in-depth data analysis. I always advise businesses to commence their AI journey with a holistic strategy that encompasses every department and workflow.

Maxime Vermeir: In harnessing the power of AI, it’s imperative that businesses anchor their AI strategies around clear, purpose-driven objectives that map to positive business outcomes. Here are three key items to keep in mind to ensure the effective utilization of AI:

Establish clear business objectives you aim to achieve with AI. Whether it’s improving operational efficiency, enhancing customer experience, or driving innovation, having well-defined goals is crucial.

Instead of being swayed by the allure of generalized AI, opt for specialized, contextual AI solutions tailored to address specific business challenges. These specialized solutions are more likely to deliver accurate and actionable insights.

High-quality, relevant data is the cornerstone of successful AI implementations. Ensure your data is clean, well-organized, and reflective of the real-world scenarios your AI solutions will encounter.

Are you using any variation of a large learning model like ChatGPT?

 Field: Yes. We have integrated LLM capabilities into many of our products but at this nascent stage, it’s challenging for our customers to know how to achieve the best results based on their particular industry and use cases. So, as a strategy organization, we test against many different large language models across many permutations of parameters across many different use cases so that we can advise on how to get the best results, empowering our customers to make informed decisions, fast. You get very different results between gpt-3.5, gpt-4, Llama 2, and many others that are emerging on the scene almost daily. To augment my own work, I have subscriptions to ChatGPT Plus and Midjourney; the latter has helped up my creative game on many presentations.

Lamendola: We’ve started to leverage ChatGPT functionality within our support operations and are seeing significant improvement in resolution times. Enabling our support teams to interact using unstructured NLP-based inquiries helps them access the right information quickly, improving efficiency as well as customer satisfaction. Recently, we’ve modeled using ChatGPT as a tool to mine sales opportunities to create more efficient and consistent responses to new customer leads.  

McDonald: I utilize specialized tools that incorporate ChatGPT 4 and ChatGPT 4V (vision). These include inVideo for video editing and narration, and CustomGPT for creating containerized, secure chatbot presentations of assessment data. Additionally, I use the SMTP (Sell Me This Pen) platform for AI-driven sales coaching and funnel integration. Given my role in business AI strategy development and as a fractional chief artificial intelligence officer (CAIO), I leverage AI more extensively than the typical user. I always aim to be at the forefront, evaluating and recommending solutions for business owners

Vermeir: Indeed, at ABBYY, we have a longstanding tradition of embracing cutting-edge AI technologies to fuel our solutions. For several years, we have utilized our in-house built large learning model (LLM), which plays a pivotal role in recognizing text and handwritten information on documents, alongside performing classification and data extraction tasks. Unlike adopting off-the-shelf models, our proprietary LLM is trained on our own training data. This bespoke approach not only ensures the high accuracy and effectiveness of our solutions but also rigorously upholds compliance and data privacy standards, which are paramount in the domains we operate. Our tailored LLM is a testament to our commitment to delivering high-caliber, trustworthy AI solutions to our customers.

What are some of the issues you’ve encountered while using AI in your business?

Field: Our compliance team. I say that in a very positive way. Right away, they put forth guidelines on how we should use these tools. We never share any customer data with the likes of ChatGPT; everything is anonymized first. Even once Microsoft rolled out Azure OpenAI for the enterprise, our security team created separate instances for our research, development, and technical sales teams to ensure no comingling of any data whatsoever. This is why we have also taken a pragmatic approach to rolling these new capabilities onto the market. Our customers are looking to us for our expertise in how they should think about data privacy, ethics, and security when using this powerful, exciting, but new technology.

Lamendola: The most significant issue that we’ve encountered while using AI is to ensure that the data sets to which we expose the various tools are properly secured. There is a real potential for data leakage, so strong data governance is required for any successful AI project.   While there is absolutely a need to be innovative and creative, there is an equal responsibility to be good stewards of company data, assets, and intellectual property.

McDonald: In my business, staying updated with the swift evolution of generative AI has been challenging. Its pace surpasses any other innovation in my experience. In a short span, I observed LLMs updating from mid-2021 benchmarks to real-time web integration and source checks. Soon after, genAI introduced multimodal support for images and voice, swiftly followed by another significant leap in innovation. Just when you believe you understand genAI’s scope, it broadens, further revolutionizing workflows. Moreover, the daily emergence of new AI tools tailored for specific workflows emphasizes the importance of a continuous learning mindset. It’s a journey that’s simultaneously exhilarating and overwhelming.

Bill Melo: AI is really only as good as the training data. ChatGPT uses essentially the entire internet as its training data. In our case, our service applications utilize data from a few hundred thousand MFPs in the field. The current results are good and will get better over time.

What is the greatest innovation you (or your business) have discovered in the last year since ChatGPT became publicly available?

Field: ChatGPT’s Advanced Data Analysis (formerly Code Interpreter) has been a game changer. As I mentioned earlier, we do not share sensitive corporate data with ChatGPT, but using this feature with anonymized data has provided us with amazing ideas on how to interpret and present complex data in simple ways. For an ex-developer like me who has not coded anything significant in some time, various LLM-powered code writing tools have helped shake off the rust and allowed me to prototype some amazing things in a brief period without spending hours learning new languages or coding frameworks. Within a few days we went from an old-coder prototype to beginning to build real features into our products.

Lamendola: We’ve been able to transform a segment of our business from a labor-based handling operation to a digitally enabled profit center.  By implementing AI technologies to help our customers capture, connect and secure millions of physical documents per year, we’ve been able to not only improve throughput and productivity but also uncover additional revenue opportunities by deriving value from that data.

What is the next project you plan to work on using AI?

Lamendola: We’ve begun to explore the possibility of expanding our AI capabilities into our managed security services. Data aggregation and event correlation offer the greatest potential for AI-related technologies, as the time to respond to any security incident is often the most critical action. We see potential both for Ricoh as an organization as well as integration within our security services.

Melo: AI has a great application for device diagnostics and self-healing. By understanding customer usage patterns and anticipating possible stress on the device, an intelligent MFP can adjust to prolong the useful life of consumables, optimize output quality and schedule remote or on-premises service with the required parts identified.

Vermeir: AI is changing the way we interact and work with technology daily, much like how smartphones introduced the low-code/no-code era. At ABBYY, we are making our technology accessible to many, helping improve their work-life balance. Our next project focuses on enhancing our products with a new user experience powered by AI, making them even easier to use. This way, we will put powerful tools in people’s hands, simplifying their tasks and enabling more intuitive interactions with technology. By doing so, we aspire to bring the futuristic allure of seamless human-machine interaction, reminiscent of the camaraderie between Michael Knight and K.I.T.T. in Knight Rider, closer to reality. It’s all about making the futuristic idea of effortlessly communicating with technology a part of our everyday reality.

How do you ensure the AI systems you use are ethical and unbiased?

Field: I’m sure some would answer this question with a technical or legal lens — and they wouldn’t be wrong in doing so — but there is also another way to look at this challenge. As costs drop and this technology gets more democratized, outside pressures will force providers of these technologies to offer visibility into how they operate, or at least to ensure that they are consistently providing unbiased responses. Traditional predictive models, for instance, naturally have the ability to tell you the exact data points used to make a determination. It’s then incumbent upon vendors like us to provide that insight to our customers and, eventually, their end users. In another example, shortly after ChatGPT was released, there were many online communities illuminating the fact that its responses in many cases had certain political or ethical leanings. Maybe this was malicious or maybe the software was just responding based on training data, but nevertheless, OpenAI very quickly responded and fixed the problem. In parallel, many open-source alternatives popped up, and continue to regularly. In a free market, this is a forcing mechanism to keep biases in check.

How do you measure the success of your AI initiatives/what metrics do you use?

Field: For us it’s about doing more with the same, not about elimination of people. If our customers can automate their processes end to end without intervention because our AI accurately cut through all of their data, that’s a win. We are constantly polling our customers on the straight-through-processing metrics. Internally, if we can create marketing campaigns faster, generate cleaner code resulting in fewer trouble tickets, or focus our sales teams appropriately quarter after quarter, that is success. But what I have found really inspiring (but perhaps difficult to measure) is how the democratization of this technology is allowing people to uplevel themselves in areas beyond their core strengths. I mentioned earlier that a techie like me uses image generation tools to make my presentations far more impactful. I’ve seen colleagues who don’t have degrees in data science formulate opinions on financial data because AI tools were able to break it down in a way that was easy to understand. Technology has promised, and provided, much in the way of making our world better for so many years, but we are now getting to the point where the possibilities are even more encouraging than ever.

Vermeir: The crux of measuring the success of our AI initiatives lies in the tangible impact they have on business processes, rather than just the technical metrics. While technical metrics like F-scores can provide useful insights into the performance of AI models, they don’t necessarily translate to real-world effectiveness. Our focal point is on how AI can significantly enhance business operations.

The metrics we prioritize are those that reflect direct business value. These include:

Straight-Through Processing Rate (STPR): An increase in STPR denotes that more transactions or processes are being completed without manual intervention, thanks to AI, which is a primary indicator of success.

Time saved: Measuring the time saved by implementing AI solutions provides a clear picture of efficiency gains.

Return on investment (ROI): Ultimately, the ROI encapsulates the financial value derived from our AI initiatives, showcasing the cost-effectiveness and value addition to the business.

Do the benefits of AI outweigh the risks?

Field: The optimist in me says, “Yes!” Most innovations are not risk-free; however, over time, the ones that provide value to society have staying power. Consider people with disabilities, for instance, and how AI-powered solutions have positively impacted their lives. Whether that be allowing the visually impaired to use their voice to accomplish tasks, or the physically impaired to live more independently through the automation of tasks, AI technologies have empowered many to thrive in ways that might’ve seemed impossible only a few years ago. For these reasons, I believe we’ll figure out how to mitigate the risks for the better.

McDonald: Undoubtedly! Humans, by nature, face inherent risks, yet we continue to collaborate and work with them. We address human vulnerabilities, like failing to fact-check or clicking on malicious emails, but we also celebrate their strengths and capitalize on their talents. AI is no different. While it has risks and limitations, its profound effect on streamlining workflows and granting more time to individuals is undeniable. Choosing to avoid AI because of perceived risks is actually a riskier stance than harnessing its capabilities.