This post was adapted from the Odaseva Data Innovation Forum session, “How AI is Impacting the Way Solutions are Architected and Delivered” featuring:
Watch the session replay here and see a clip below.
It is evident that AI will change the way solutions are architected. It’s time to recognize that there is a fundamental shift in how one needs to approach architecture.
Let’s understand this better by taking a closer look at the following five topics where we’ll explore how AI is influencing the architecture and delivery of solutions:
In the traditional approach, solutions were built based on an algorithmic understanding of the problem, ensuring consistency in output with the same input. In a CRM system with a process around account segmentation, the traditional method involved defining fields on the account and applying business logic for segmentation, driving other automations in the system.
On the other hand, in the age of AI, models are generated based on large amounts of data, creating predictive models. Large language models can handle more detailed personalized segmentation and can take into account a much wider array of data, which brings about a change in the approach to how solutions are architected.
The current surge in popularity, particularly with gen AI, can be attributed to its transformative impact on the user experience. For the first time, users can engage with technology using natural language. This marks an evolution as it can receive requests that precisely matches the user’s intentions in a comprehensive manner.
There is a shift from the if-then-else mindset of platforms to an NLP (Natural Language Processing) focused experience. The user no longer has to navigate through a multitude of fields. Instead, when the user asks a question, they get an answer in English. This simplifies onboarding, making it faster and more efficient, without the need for extensive training for representatives.
AI facilitates achieving more with fewer customizations and as a result, enhances job efficiency.
While AI has its advantages, it also introduces new challenges, such as:
And so in an era where data privacy and trust have taken center stage, it is important to develop techniques for error handling and enhancing the predictability of AI’s output to ensure data security and privacy.
The traditional approach to delivery has largely remained unchanged for the past two to three decades, despite the evolution of methodologies and increased agility.
Until now.
In the following ways:
With the NLP, the types of roles and the nature of the skills that people need are going to change significantly. If the volume of data generated by the sales team during the discovery phase can be summarized into a handover, it would save the project team and the customer a lot of time, accelerating all the stages of a typical delivery and making the process more efficient.
AI, specifically in professional services engagements, gives the ability to create commodities for manual tasks. Often, when executing an engagement, whether it’s Salesforce delivery or anything else, there are many manual tasks needed to keep things coordinated and in sync. AI gives us the ability to eliminate that and turn it into a commodity, enabling the human brain to concentrate on more complex tasks, delivering more business value to customers.
For example, for complex CPQ (Configure, Price, Quote) projects, the user can focus on designing the right pricing rules and collaborate with their customers without worrying about billable hours for manual tasks.
AI is everywhere, it impacts architecture and professional services. Can it solve everything? Or is it just something fanciful, and its practical utility is questionable?
Let’s take a deeper look at this.
There are very high expectations from AI. Within this context, there’s a prevalent concern among individuals that they might lose their jobs to AI.
On the contrary, AI helps make jobs easier and get them done faster, freeing up their time to pursue other things of interest.
Considering the hype cycle and its continuous evolution, it’s crucial to acknowledge that the transformative journey of AI has just begun and the perspective should be one of curiosity and a commitment to improvement, viewing AI as a new frontier and approaching it with an open and inquisitive mind.
On the flip side, the energy impact of AI requires attention. AI’s widespread use could contribute to a significant carbon footprint. Addressing this challenge, encompassing data, data security, and environmental aspects, is crucial on a global scale, that would necessitate solutions sooner than later.
Until recently, language was considered a unique human capability. The ability to communicate complex ideas and emotions, once thought exclusive to humans, is now replicated by large language models, even if they lack a true understanding of the content they generate. This shift is inherently significant, challenging the notion that language generation is solely a human capability.
Despite its advancements, AI still requires explicit instructions on what we want to achieve. Regardless of the complexity of the task or implementation, human intelligence/minds are crucial in guiding AI to produce the desired output. While AI can enhance productivity and speed up tasks, it ultimately functions as a tool that relies on human direction.
The evolution of AI signifies a shift in problem-solving capabilities. AI can be utilized in the context of the current technology landscape, by identifying the low-hanging fruits, and determining what can easily be delivered to end-users.
In the exchange of information between teams, particularly in the context of testing, important side conversations often get missed. AI has the potential to maintain a constant understanding of intent, facilitating intent-based testing which can revolutionize the testing process, ensuring that user stories and requirements consistently drive testing efforts.
AI brings a set of tools to address long-standing problems. For persistent issues in various domains, such as pipeline predictability and routing, AI is a transformative force in changing how these problems are tackled and enhances the efficiency of tasks.
However, when it really gets to value, it’s about depth. There are generations of knowledge in certain industries, such as healthcare, where individuals have accumulated 40 or 50 years of experience, and they might be retiring. The challenge is how to archive that information, put it into a domain-specific large language model (LLM), and tap into centuries worth of healthcare-related knowledge at our fingertips, while at the same time being conscious of whether going back to the 1800s or looking at information from that time is really valid now.
Plan, develop, build, test, release, and deploy are the key stages of the DevOps process. The big impact is around testing. Exploratory testing allows an end user to just explore and find the edge case. AI has the potential to democratize DevOps at a very high speed allowing people that have never been able to participate in software delivery to participate.
With large language models, particularly, there emerge entirely new classes of security threats that the community doesn’t yet comprehend fully.
The extent of these threats remains unknown at this point. A certain level of caution is advised regarding the level of autonomy granted to AI-driven processes and how users can contribute input. Basic prompt injection attacks, where attempts are made to manipulate the large language model into deviating from its instructions, prove highly effective and can deceive existing safeguards. Traditional security has had an extensive history of the ongoing struggle between those attempting to compromise systems and those seeking to protect them. However, with the advent of generative AI, particularly the new kinds, we are still in the early stages of understanding and addressing potential risks and so when it comes to delineating rights, monitoring procedures, and the involvement of humans at critical junctures in the development and deployment of such systems, it advisable to tread with utmost caution.
Want to learn about more ideas, opportunities, and strategies to maximize the value of Salesforce data? Watch sessions from the 2023 Data Innovation Forum for Salesforce Architects here.
Data Protection
Data Security
Data Archiving
Data Compliance
Data Platform
