ChatGPT Enterprise

ChatGPT was released to the public a couple of days before we created a policy for how engineers engage with the tool. I am 100% behind using any tools to solve problems and speed up delivery, but keeping possession of our Intellectual Property is also important. Even more important in the credit and collections industry it is critical that we safeguard personal information as much as humanly possible. Having data in Europe or UK the important, keeping on our own cloud services is ideal. There are some real challenges with compliance with channels like WhatsApp when they want to move data to the US.

It will be interesting to see how far Enterprise ChatGPT goes with addressing these concerns.

Read More →

Aligning Business with Technology

Aligning business with technology seems like it should be the simplest task in the world. Yet, it’s astonishing how challenging it can become. Often, it feels as if the tension arises from balancing cost against quality, or in defining UAT metrics that engineers can aim for. However, the issues can be much subtler than they appear.

Several years ago, before Webio, we worked on a webchat widget that utilized intent recognition to decipher customer inquiries and deliver appropriate responses. We dedicated considerable time and effort into training an IBM Watson data model to understand these questions and offer the answers. Concurrently, we developed an agent console to facilitate these conversations. All this was undertaken within an innovation lab, which aimed to revamp an outdated platform.

We pursued both tasks with enthusiasm. We located datasets, trained the AI, constructed workflows, and refined our strategies. We also determined contracts, calculated costs, and simultaneously, developed a state-of-the-art (for its era) extensible, restful API agent console, employing every conceivable framework and library to ensure its excellence.
Eventually, we secured a client. During our meetings, the client expressed significant excitement about becoming one of the pioneers in their industry to adopt this AI technology. Upon completing the product, we showcased it. The AI performed impressively; its intent recognition was almost flawless, save for the rare misunderstanding.

However, a pivotal moment came when one of the senior managers, who had been our staunchest supporter, assessed our work. She commended our achievements but then posed a simple question: “Wouldn’t it be easier for the customer if we just used buttons instead of intent recognition?” And she was unequivocally right. It simply streamlined the process. Consequently, we chose to retain the AI, but relegated it to a backup role, favoring a menu-driven interface. The overriding metric for success was usability, and this change aligned with that goal.

From that experience, I’ve consistently employed a balanced approach: using intent recognition for intricate queries and opting for buttons when only a few common questions are anticipated.

Read More →

Generative AI: The Little Things

I know everyone and their dog has been talking about generative AI and all the exciting opportunities it offers. However, sometimes it’s the simple things that have the most profound impact. For me, Alexa was a game-changer when I could ask it how to spell a specific word.

Now, with tools like ChatGPT and Bard, I’m finding their proofreading capabilities to be incredibly helpful! Like many software engineers, the neurodivergence that aids me in building software also makes proofreading exceptionally challenging.

When I’ve written white papers or crucial proposals in the past, I’ve often had to meticulously go through the content line by line. I’d start from the bottom, working my way up, reading each line out loud three times to even come close to producing something error-free. Now, with so many tools at my disposal (including for this post!), I can simply paste my text into ChatGPT, kindly request a proofread, and off we go!

Read More →

Building a Strong Foundation for AI and Innovation: How to Herd Genius Cats

Companies seeking to build ethical and robust AI systems face immense challenges. Key among them is establishing an effective framework to guide development and deployment. This requires laying a strong foundation that promotes responsible innovation from the ground up.

According to industry thought leaders, a critical first step is assembling diverse, multidisciplinary teams. Different perspectives are crucial for identifying potential pitfalls and unintended consequences early on. Fostering a collaborative “herd genius” environment allows teams to tap into collective intelligence. It enables them to approach AI with nuance, foresight and care.

Structuring teams and workflows in an intentional way establishes the right cultural tone. This empowers teams to build human-centric AI systems that earn public trust. With a robust foundation in place, companies can innovate responsibly and help realize AI’s immense potential for good.

Read More →