So far, I have dealt with how to safely and securely implement and leverage AI — in particular generative AI — technology. So, there shouldn’t be many questions open regarding how to get a trustworthy, efficient, and effective AI implemented in any organization.
This being my last column article of 2023, I think it is time to look at finding business scenarios that are valuable to implement. This is the natural time for a deep gaze into the crystal ball.
Because, until now, there are only a few that are talked about. Yes, without AI we probably wouldn’t have found some vaccines as fast as we did; yes again, we do know that AI systems can be better at diagnosing medical conditions than doctors. And self-driving cars wouldn’t be as close to reality as they are without AI.
Still, the fact remains that so far, business software vendors have only done a decent job identifying and delivering what I would call low-hanging fruit. They delivered (or are about to deliver) solutions that give people back time to do “real” work instead of busying themselves with chores.
These solutions are built on what is core to a generative system: Generate “new” content like product descriptions, articles, and emails, summarize content, or provide answers to questions, as well as some coding. Especially low-code and no-code vendors are embracing generative AI. Don’t get me wrong, all this is helpful but it’s only a start.
While these scenarios improve work for employees and customers, they are largely focusing on gaining efficiencies.
From a business point of view, these scenarios and their implementations can only be described as tactical rather than strategic. Again, they have their value, but they are not the most valuable use cases that could be implemented.
Which begs the question: What are some more valuable use cases?
Glad you asked!
Let’s look at some of the nearer and not-so-near-term opportunities to create more value using generative AI without losing a view on ethics. While I can understand that studios see some value in having whole screenplays auto-generated or actors replaced by generated avatars, I do not see a strong ethic behind this — except for overly dangerous stunt scenes.
So, what are some possible scenarios?
More effective support
One of the more obvious scenarios lies in customer service. Customer service, be it self-service or assisted service, relies a lot on documentation. Still, documentation that answers an inquiry is often either not found or only after searching repeatedly. Or it doesn’t exist at all. In the latter case, new organizational knowledge gets created. In the former situation, existing knowledge needs to be analyzed and updated, rephrased, or attributed with changed metadata.
In the recent past, we have seen several announcements that are about improving existing documents. The next step would be the creation, review, and publication of new support documentation. The underlying knowledge is there, be it in product design documents or case notes written by agents or even in the chat flows of successfully resolved cases. The information about a case being solved, thereby creating organizational knowledge is available, too. It is part of the case resolutions. I call on vendors of customer service and knowledge management solutions to fix this gap.
As an alternative, vendors of cognitive search could address this, too, as it is not necessary to create and persist new documents if the underlying knowledge can be extracted from existing documentation and incorporated into an answer to an inquiry. In this case, the generative AI “only” covers the interface to the human customer or agent.
The value lies in streamlined customer service and knowledge management. Information is faster and more readily available to (internal and external) customers and agents, increasing productivity.
I would not be surprised if vendors like Coveo or Microsoft with its Cognitive Search are already actively looking at scenarios like this.
Conversational user interfaces
This one might get a boost in the advent (pun intended) of numerous co-pilots.
So far, we as humans are mainly forced to work with computers using mainly unnatural, acquired interaction methods: drag and drop — operating computers the computer way. This also forces us to log in to the applications and be bound to their user interface. There have been activities to change that for several years now, including major investments such as the acquisition of Slack by Salesforce. The main goal is twofold: First, make application services available outside the frame of the application, where users often operate. Second, enable users to interact with applications in a more natural way – natural for humans, not computers.
Which doesn’t mean that the application response is always text. Sometimes, a diagram or a table is the better means. Still, application interaction can be greatly optimized. I wrote about this already back in 2016 (although not accounting for LLMs). This is largely possible due to an API-driven way of developing applications and advances in NLU and NLG via LLMs.
The advantages are quite clear. The more effective and efficient user interaction model increases peoples’ efficiencies while contributing to a better UX. And then, the increased mobile ability helps reduce downtimes. The big lever is that this is possible throughout business applications.
The holy grail of generative AI
Let’s look at the convergence of process mining, low-code, no-code, robotic process automation RPA, and generative AI. Many a business has already implemented one or more of these technologies to identify process gaps and process inefficiencies and then to fix them. Tools like Signavio or Celonis are incredibly good at identifying process gaps and/or process inefficiencies. Companies like UIPath strive to automate business processes. Yes, Celonis does this, too. LeanIX and other companies map the enterprise application landscape along with the applications’ purposes. Companies like Creatio, Pega, ServiceNow, Ues.io, Outsystems, and tools like Mendix, Power Platform, Zoho Creator and many, many more, strive to facilitate and ease the implementation of these processes. All of these companies and tools have a varying scope and use various degrees of AI in their systems. What they do in their entirety is help to identify and solve process issues that range from simple to very complicated.
Which is good!
But not enough!
Why? Because these tools do not connect the dots. And because it needs too many of them to do the job. The job is to improve business processes to help the employees better serve customers. Or the other way round: Improve EX and CX.
What is the solution? The missing dots are the ability to automatically fix broken processes. All the data, information, and capabilities are there. The application landscape is known. The processes are known and measured, which surfaces its issues. Application development environments ranging from no-code to pro-code, in combination with a generative AI, are capable of fixing identified problems.
The advantages are obvious: The turnaround time between identifying an issue and the delivery of a solution is drastically reduced, therefore increasing process efficiency and contributing to employee satisfaction. As a side effect, the load on stretched IT departments is reduced and the number of applications developed by “citizen developers” is drastically reduced, lacking a need. Not to speak of this being an entry into something that one could call “intelligent applications” — applications that adapt their behaviour to evolving business needs.
The next step from there would be the creation of processes, based upon requirements and regulations.
I, for myself, am waiting for the first vendor to deliver a solution that addresses this opportunity holistically. I seriously think that some of the vendors mentioned above do have the capability to tackle it and should do so.
What do you think are mid- and long-term high-value business scenarios for the application of ethical, efficient, and trustworthy generative AI?
