The Ascent Begins: – The Sherpa’s 5 Steps for Success in the Enterprise AI Journey
AI Climbing Gear: Identifying the Starting Point for AI in Enterprise SaaS
The question of where to begin the AI journey for an enterprise SaaS organization is paramount. While AI has demonstrated remarkable successes in diverse fields, from healthcare to environmental science, translating these achievements into concrete business value remains a complex challenge. Here are some recommendations to enable a solid start.
There is a surge in AI-powered solutions focused on classification, summarization, Semantic knowledge-based search, prediction, and image recognition, often delivering impressive results. However, the actual value lies in applying these capabilities to specific industry problems. Identifying these "Type 1" use cases is crucial.
Before implementing AI, organizations must address foundational challenges. Existing technical and functional debt, data quality issues, or organizational complexities can hinder AI adoption. Overcoming these hurdles is essential to prevent future complications or a chaotic environment.
Once these foundational issues are addressed, the focus shifts to identifying high-impact business use cases. These should align with strategic objectives and offer a clear path to ROI. If suitable use cases are not readily apparent, iterative exploration with the business is necessary.
Selecting the appropriate AI models is equally critical. While large pre-trained models offer versatility, specialized task models can excel in specific domains. A pragmatic approach can yield quicker results, starting with predictive AI on existing data and gradually progressing to generative AI.
Building a solid data foundation is indispensable for AI success. Establishing data access, defining data profiles, and enabling efficient data retrieval through APIs or flows are essential prerequisites. Using existing platforms like Salesforce, with its predictive segmentation capabilities to determine the likelihood of closing a lead based on historical data, can serve as a strong starting point.
Focusing on simple AI applications with high business impact is more strategic than pursuing many complex projects that can potentially be unsustainable.
Ensuring basic permissions and CRM data profiles are in place is vital for achieving accurate outcomes. If the foundation is set correctly, dynamic grounding of prompts through Flows or API calls can effectively retrieve relevant customer data and business scenarios. Additionally, there are OpenAI Python libraries for language translation, sentiment analysis, question-answering, summarization, and code generation. These libraries offer features like embeddings for search, clustering, recommendation, and classification, which should be thoroughly explored.
These steps guide companies on where to begin their AI journey, maximizing the potential of this transformative technology while mitigating risks.
Infosys’ Salesforce practice has developed an "Image to Text" Generative AI application that helps prevent potential infrastructure damage or community impact across industries. The application identifies the user and context by uploading an image in the Salesforce Experience Cloud, enriching it with instructions and grounded CRM data, and sending a request to a base model via an API call. The response can be parsed in a flow or OmniStudio to trigger notifications, actions, or work assignments to service technicians.
This use case can be adapted to other verticals, enhancing response accuracy with additional text, descriptions, and probabilities.
Organizations should first identify simple AI applications with high business impact rather than getting overwhelmed by evaluating numerous models or use cases. An AI roadmap should be created and integrated into the overall planning for new transformation programs to achieve a high return on investment.
