10 most used gen AI tools in the enterprise

Gen AI has entered the enterprise in a big way since OpenAI first launched ChatGPT in 2022. According to Precedence Research, the global gen AI market was over $25 billion in 2024 and is forecast to reach a staggering $803 billion by 2033. And AI at Wharton, part of the Wharton AI and Analytics Initiative at the UPenn’s Wharton School, together with consultancy GBK Collective, also found in a study of senior decision-makers that enterprises with 1,000 or more employees invested on average more than double in gen AI in 2024 than 2023.

But a recent survey by Deloitte, found that even though a majority of director- to C-suite-level respondents experienced with piloting or implementing gen AI plan to increase their overall AI spend in the next fiscal year, they’re tempered in their approach, expecting only 30% or fewer of their gen AI experiments to be fully scaled in the next several months — a reflection of the tools at their disposal. So given the current climate of access and adoption, here are the 10 most-used gen AI tools in the enterprise right now.

ChatGPT

ChatGPT, by OpenAI, is a chatbot application built on top of a generative pre-trained transformer (GPT) model. Launched in 2022, it’s the most-used gen AI tool in the enterprise, with 62% of respondents to the recent Wharton survey saying they currently use it and 28% saying they don’t currently use it but are evaluating or testing it.

ChatGPT’s strengths are natural language understanding and generation, making it an excellent tool to deal with complex linguistic challenges including automating and enhancing customer service, streamlining operations by handling repetitive inquiries, generating reports, drafting emails, and creating product descriptions.

Like all gen AI, ChatGPT’s accuracy can vary and its inability to understand emotions can lead to issues in some situations. Human oversight and intervention may be necessary.

Microsoft Copilot

Microsoft Copilot is a conversational chat interface embedded in Microsoft 365 to enhance productivity in applications like Word, Excel, PowerPoint, Outlook, and Teams. Launched in 2023, it leverages OpenAI’s GPT-4 foundational LLM and is the second most used gen AI tool. According to Wharton’s survey, 52% of respondents say their enterprise currently uses Copilot and 35% say they are evaluating or testing it.

Copilot’s strength lies in its integration with the productivity apps employees in Microsoft 365 environments use all day. It adheres to the security, compliance, and privacy policies the enterprise already has in place, and is available as an add-on to existing Microsoft licenses. Uses include:

• Drafting documents and summarizing text in Word
• Generating charts from existing data in Excel
• Creating outlines and designing slides in PowerPoint
• Managing your Outlook inbox, summarizing email threads, and drafting replies
• Summarizing key points from Teams meetings and creating action items
• Leveraging Power BI to create reports and visuals using natural language queries

As with ChatGPT, Copilot’s accuracy is not perfect. For instance, when summarizing content, it may miss the main point or overlook key details. Likewise, even when drawing on branded materials, copy, and images, its output won’t necessarily align with brand standards. Users need to keep an eye on the output.

Google Gemini

Google Gemini (formerly Bard) is a conversational chatbot built on Google’s family of Gemini multimodal AI models. Wharton found 40% of respondents to its survey are currently using Gemini, and 39% are evaluating or testing it.

Compared with ChatGPT and Microsoft Copilot, Gemini, launched in 2023, took an early lead in multimodal features — the ability to process and generate images, audio, and video in addition to text. Gemini is integrated with Google Workspace tools like Gmail, Docs, and Slides.

Potential uses of Gemini include:

• Drafting emails based on prompts from Smart Email Composer
• Creating slide decks in Google Slides based on prompts
• Identifying trends, outliers, and correlations in Google Sheets via Spreadsheet Analyzer
• Generating background illustrations and simulating professional lighting for Google Meet video conferences, and even improving the audio experience

Gemini integration with Google Workspace makes it an excellent tool in environments that use Workspace but has limited utility outside such environments as it operates within Google’s secure environment. And as with ChatGPT and Copilot, accuracy may be an issue and human oversight of its output is recommended.

Meta AI

Meta AI is Meta’s multimodal AI virtual assistant for the company’s messaging and social media applications, including Facebook, Instagram, WhatsApp, and Messenger. As with Gemini, 40% of respondents to Wharton’s survey said they currently use Meta AI, while 36% said they’re evaluating or testing it.

Powered by Meta’s Llama LLM, users can leverage Meta AI to offer suggestions, answer questions, edit images, and provide translations in the company’s apps. Enterprises can use Meta AI to set up virtual assistants that can:

• Automate customer interactions, including answering questions and giving personalized recommendations
• Facilitate commerce within Meta’s applications
• Analyze data to help businesses create targeted marketing campaigns

Meta AI is already integrated with the apps in which many customers spend much of their time online, but there are potential privacy and ethical implications of access to customer data.

Dall-E 3

Gen AI isn’t just about chatbots and virtual assistants. DALL-E 3, also from OpenAI, focuses on generating visuals from text descriptions, and 30% of respondents in the Wharton survey said they currently use DALL-E 3, and 35% said they’re evaluating or testing it. OpenAI launched the original DALL-E model in 2021, and the DALL-E 3 deep learning model leverages computer vision and natural language processing to create visuals. Potential business uses include product ideation, app mockups, logo design, creating images and videos for social media posts, and educational materials.

Among AI image generators, DALL-E 3’s strength lies in its integration with ChatGPT, yet many users say it struggles with photorealism, with a distinctive style that makes it easy to spot the model generated an image.

RunwayML Gen-1 and Gen-2

Runway uses text, images, and video inputs (including content generated by other gen AI tools) to generate video, and 25% of respondents to Wharton’s survey said they currently use Gen-1 and Gen-2, while 31% said they were evaluating or testing the models.

The Gen-2 model, released in 2023 (Gen-4 was released in April), extends the Gen-1 model’s capabilities, allowing videos up to 18 seconds to be created. Users can direct the movement of the camera to customize content and stitch clips together to create longer videos, too. Among other things, enterprises can use the tool to animate static assets, visual effects, and storyboard.

Runway has a steep learning curve for some features and can be resource intensive.

Stable Diffusion

Released in 2022, Stable Diffusion is a deep learning model from Stability AI that generates high-resolution, photorealistic images from text and image prompts. It can also create videos and animations, and 24% of respondents to Wharton’s survey said they currently use Stable Diffusion and 33% said they’re evaluating or testing it.

Based on a latent diffusion model that requires much less processing power than some competitors, Stable Diffusion is capable of running on desktops and laptops with consumer-grade GPUs rather than being restricted to cloud services.

Potential business uses of Stable Diffusion include medical imaging, generating visuals for catalogs, designing digital prototypes, generating concept art and storyboards, and image editing including color correction and noise reduction.

Midjourney

Midjourney is a tool to generate images from natural language prompts. Launched in open beta in 2022, 23% of respondents to Wharton’s survey said they currently use Midjourney and 32% said they’re evaluating or testing it.

Midjourney can be accessed via the Discord communication platform or the tool’s official website.

Images created via a Midjourney commercial license can be used commercially but can’t be copyrighted unless transformed significantly by a human. As such, images created by Midjourney and not transformed by a human artist are considered public domain.

Perplexity

Perplexity, launched by Perplexity AI in 2022, is an AI-powered search engine that leverages LLMs to process queries and synthesize responses, and 21% of respondents to Wharton’s survey said they currently use Perplexity and 33% said they’re evaluating or testing it.

Perplexity is built on Microsoft Azure. The free version leverages OpenAI’s GPT-3.5 LLM, but paid users can choose their model. Uses include:

• Internal search for searching content simultaneously across the web and internal documents
• Finance, including real-time stock quotes, price tracking, industry peer comparisons, and basic financial analysis tools
• Perplexity Assistant, an AI-powered search assistant for Android devices

Claude

Claude is a conversational AI chatbot released by Anthropic in 2023 and 19% of respondents to Wharton’s survey reported currently using Claude, while 34% said they were evaluating or testing it.

Claude relies on Constitutional AI (CAI) principles, developed by Anthropic, which is an approach for training AI systems that seeks to address concerns around transparency, safety, and decision-making without reliance on human feedback. CAI aims to give Claude ethical boundaries, giving it the ability to refuse to answer queries if it judges the response would cause harm.

An alternative to other generalist offerings like ChatGPT, businesses can use Claude to automate and enhance customer service, streamline operations by handling repetitive inquiries, generate reports, draft emails, and create product descriptions.

Claude is known for the clarity of its responses, but also a tendency to be verbose.