Generative AI (GenAI) tools have been rapidly gaining traction with their potential for software companies to improve existing processes and increase efficiency and scalability. I have picked two specific areas where Frog portfolio companies have been implementing GenAI in practical applications and explore their benefits as well as their limitations.
- Code automation tools are changing the game for developers. These AI assistants analyse code context and suggest relevant completions, functions, and even full code blocks, reducing the time spent on repetitive tasks. This frees developers to focus on complex logic and feature development; accelerating development cycles and improving sprint efficiency.
- LLMs like Gemini (Google) and LLaMA (Meta) are transforming the power and use of internal knowledge bases by understanding natural language and context. They can automatically curate content, extract key insights from unstructured documents and conversations, and even answer employee questions in a natural language format.
Code Automation
Where speed and innovation are critical demands of software development teams, code automation tools are increasingly explored for their potential to optimise developer workflows. Tools like CodeWhisperer (Amazon) and Copilot (GitHub) extend beyond simple auto-complete for code.
- Velocity – Code Automation tools are able to scan code, offering contextually relevant completions. Automatic alignment with coding standards, the auto-generation of repetitive functions and syntax correction are saving hours of keyboard time and letting developers focus on complex logic and business-critical features.
- Hyper Focussed On Demand Help – These AI companions are now acting as on demand encyclopaedias, offering explanations and contextually-aware suggestions. Whether understanding an obscure API call or interpreting an ambiguous error message these tools are even able to suggest potential fixes on the fly.
- Quality – Code Automation tools promote cleaner and more maintainable code. By suggesting best practices, existing code can be refactored for efficiency or to protect against newly identified security vulnerabilities. Technical debt and the need for debugging is reduced leading to better platform stability.
As GitHub’s name suggests these tools are your copilots and not replacements. While they may improve your capabilities; critical thinking, problem solving and the need to understand code remains essential. Code automation is not just about coding faster, it is about coding smarter.
Large Language Models (LLMs) – Powering Smarter Internal Knowledge Bases
The traditional use of static knowledge bases (e.g. Confluence or Wiki ) is undergoing a shift with the advent of LLM powered knowledge hubs. These hubs offer a novel approach using natural language processing to contextualise user inputs and producing summary results from disparate and unstructured data. Content aggregation is optimised, eliminating redundant data and identifying conflicting content.
- Semantic Search – LLMs understand the meaning behind your questions, delivering precise answers to even the most natural language queries. For example, asking “What are the key KPIs for a Sales Leader?”, LLMs can now reference, include and discount multiple sources of data before outputting a summary answer and links to its own reference material.
- Content Curation – Maintaining knowledge bases particularly for software companies who are continuously delivering always runs the risk of being outdated at least in part. LLMs can automate knowledge base updates by ingesting data sources such as release notes. This dynamic learning continuously adapts keeping content relevant.
- Tailored Knowledge – LLMs can learn your needs and preferences, personalising knowledge delivery. For example the breadth and depth of onboarding documentation for a software platform such as a company CRM can be tailored by job role, prioritising the features and functionality that are immediately relevant to the member of staff.
The integration of natural language processing and the optimisation of content aggregation is not only streamlining information, but also resolving conflicting data and improving accuracy. Beyond these efficiency gains, LLMs are also enabling wider access to content and information. Language barriers disappear making knowledge bases available to wider international audiences and markets through seamless translation.
