According to A McKinsey reportGenerative AI could have an economic impact of $2.6 trillion to $4.4 trillion annually. For context, the UK’s entire GDP in 2021 was $3.1 trillion. About 75% of this value will come from productivity gains in customer support, sales and marketing, software engineering and research and development.
The impressive potential of AI has led to the emergence of various enterprise generative AI startups focused on applying multi-language model technology in the enterprise context. Standard products include worker co-pilots, content generation for marketing, back-end automation and enterprise knowledge mining.
These AI-powered solutions are starting to realize the anticipated global economic impact by significantly reducing operating costs, generating sales and increasing worker productivity.
Driven by enormous opportunities, the variety of enterprise generative AI startups has expanded rapidly over the past two years. Just searching for AI customer support chatbots returns lots of of startups. Two aspects help explain the significant variety of startups using AI. First, LLM technology is available through the APIs of huge AI research firms akin to OpenAI. Second, while most generative AI startups are recent, many startups founded before 2022 are also integrating generative AI into their product suites.
Given the confusion in the market and the large variety of startups pursuing generative AI for enterprises, it will possibly be difficult for enterprise clients and investors to evaluate and differentiate options.
Below are five things price paying attention to.
1. LLM Customization
Does the startup mainly use a ready-made LLM – e.g. OpenAI‘S ChatGPT — Is a significantly adapted LLM? Various ways to customize LLM include tuning an off-the-shelf model or building a custom one using open source LLM, e.g. Metais Llama. Greater personalization generally increases AI accuracy and reflects the startup’s technical expertise in AI.
2. Industry-specific training data
Does the startup have access to a great amount of proprietary, industry-specific data to train its LLM? For example, AI second pilot for customer support centers might be improved if the AI model is trained on large amounts of existing customer interaction data. The more industry-specific the training data, the higher.
3. Speech-to-text/text-to-speech power
Many enterprise AI products support voice audio for understanding or generation. This technology is different from the LLM and is normally offered by large technology firms akin to Google, Amazon AND Microsoft.
Understanding and testing STT/TTS offerings, akin to support for jargon or industry dialects, is vital if your AI solution has voice applications.
4. Width and depth of API integration
The distinguishing feature of generative AI for enterprises is that the AI system can directly interact with enterprise systems, e.g. Sales power 1 , SAPetc. This is mainly done through APIs, so wider and deeper API integration will enable the AI system to provide greater process automation. Ease of integration is also something to look out for.
5. Business user friendly
A generative AI solution might be higher used in enterprises if it offers wealthy features that are accessible to non-technical and non-data science users.
Look for solutions that provide a low-code/no-code development and operational environment, in addition to robust analytics and A/B testing capabilities available to business users.
The UI/UX needs to be intuitive enough to require no training for business users, and even if implemented, developer training shouldn’t take greater than a few days.
Effective AI might be versatile
The world continues to be captivated by the potential of generative AI to revolutionize the way knowledge work is conducted across the economy. Artificial intelligence has the ability to transform tasks and increase the efficiency of key enterprise functions akin to customer support and sales and marketing.
Looking ahead to the next few years, generative AI technology for enterprises will grow to be increasingly multimodal with recent solutions able to concurrently handle multiple forms of input and output data, akin to images, audio, video and text. This will unlock more sophisticated and versatile AI applications for enterprises, especially beyond traditional use cases, and help drive further growth in this exciting space.