Like 23
We thoroughly enjoy engaging in partnerships with Independent Software Vendors (ISVs) at Cybage. Our proficiency lies in in-depth Product Engineering across diverse technology stacks, enabling our clients to deliver optimal and efficient software to their end customers. Given our extensive engagement with ISVs, both as consumers of software products throughout the engineering lifecycle and as collaborative partners in modern software development, we are particularly intrigued by the integration of Generative AI with existing software products.
Our interactions span various sectors, including Media/Marketing, Fintech, Retail, Travel, Healthcare, Edtech, and more. All our clients have distinguished themselves in their respective domains by creating thoughtful and feature-rich software products. It is evident to us that Generative AI does not directly replace the necessity for vertical-specific software; however, it holds the potential to enhance existing solutions. Leveraging the capabilities of Retrieval Augmented Generation and function-calling, combined with intelligent product design and user experience, ISVs can embark on organically incorporating “Copilots” into their software products.
This blog will showcase real examples of features powered by Generative AI that have been successfully integrated into current software products. We’ll also explore the intricacies behind these implementations, providing insights into what occurs behind the scenes and offering inspiration on how you can integrate similar experiences into your products.
As Large Language Models (LLMs) advance in their ability to:
- Comprehend more context in a prompt with increased token limits, and
- Generate responses in structured formats like JSON calls to an API; the excitement and predictability of these features grow.
Feature Spotlight: Brex Assistant
Brex is a fintech platform specializing in spending and expense management, issuing cards, and catering primarily to startups and smaller businesses. Consequently, the platform incorporates various invoicing and employee expense management functionalities. Recently, Brex introduced ‘Brex Assistant’, an AI-powered agent leveraging Generative AI technology, significantly enhancing the user experience. Here are some notable features of this assistant.
Image
Feature 1 — Inferring invoice content from calendars.
Image
Feature 2 — Responding to user queries about company policies.
Through intelligent utilization of Language Models (LLMs) on customer-specific data, Brex introduces compelling functionalities, such as:
- Extracting relevant fields from invoices/calendars by Brex Assistant, resembling a form of prompt engineering/RAG contextualized to meet the needs of Brex users.
- Responding to employee inquiries about company expense policies, resembling a customized version of RAG on specific data.
- Addressing employee queries regarding flight options,resembling a form of function/API calling by LLMs, akin to GPT plugins and related features.
Implementing these features at scale, conducting thorough testing of functionality and edge cases, and designing solutions to ensure predictability and return on investment pose considerable challenges. At Cybage, we concentrate on these aspects to assist our clients in incorporating such features into their products effectively.
Feature Spotlight: Shopify Magic (Sidekick: AI for Commerce — Powered by Shopify Magic)
Revolutionizing user experiences, Shopify, a stalwart in the B2B e-commerce domain, elevates its capabilities with the seamless integration of LLMs through ‘Sidekick/Magic’. This empowers users to:
- Generate content like discount codes for your products
- Take actions based on inventory status
- Generate insightful reports on sales performance
Now, let’s explore two noteworthy examples that showcase the impact of these enhancements.
Image
Feature 1 — Creating discount codes for your products.
Image
Feature 2 — Reporting on sales performance.
Other notable features include initiating inventory sales through API calls triggered by natural language prompts. Additionally, for creative marketing, it provides users with a platform to generate product descriptions, blogs, etc. all while staying in the Shopify product.
These features, applicable across various software products, tap into existing APIs and unstructured data, completing user workflows. It’s all about unlocking the value adds that Gen AI brings while staying within the same software product, using custom context and data, and building an experience intuitive to users. Implementing such features is challenging, involving troubleshooting LLM limitations such as high API counts, unpredictable responses, RBAC, token limits, and intuitive UX. However, our hands-on experience positions us as ideal partners to guide you through these challenges, turning them into opportunities for innovation.
Ready to brainstorm and enhance your software platform with a powerful “Copilot” experience? Reach out to us at Cybage, and let’s embark on this transformative journey together!
Originally published on Medium: https://medium.com/@genai_cybage_software/building-a-copilot-for-your-product-your-isv-guide-to-gen-ai-integration-8ad3ed68051a
