Enterprise AI vs consumer AI: Why ChatGPT doesn’t work at work

Consumer AI tools don’t have context for the tasks you’re asking them to do at work.

Glenn Jaume

Product Manager at Coda

Enterprise AI vs consumer AI: Why ChatGPT doesn’t work at work

By Glenn Jaume

Share this article

Share on TwitterShare on LinkedInShare on Facebook
AI · 8 min read
Do you ever feel like you’re missing something in the AI craze? Sure, chatbots are funny, and it’s nice to have better predictive text and grammar checking for emails. I’ve even heard of friends or friends of friends using ChatGPT to plan meals, generate grocery lists, and plan their workouts. But if you’re working on sensitive information on cutting-edge enterprise teams, the chances are good that consumer AI can’t actually do all that much for your workflow. That’s where enterprise AI comes in.

What is consumer AI?

 Consumer AI is what most people think of when they talk about AI. ChatGPT, a Large Language Model (LLM) initially launched in late 2022, is a huge part of the current AI craze, and its large language model has been entertaining thousands. But it’s far from the only consumer AI on the market. Apps like Google Translate, Otter, and Grammarly have been running on AI for years. In short, consumer AI apps are writing, art-generating, or translation machines that scan huge amounts of data and use it to answer questions or respond to prompts, like a sophisticated predictive text machine. Thanks to the patterns consumer AI tools have learned from the data on which they were trained, these apps can serve as a chatbot, writing tool, portrait generator, translator, transcriber, grammar checker, and search engine (but definitely double check its answers; some days ChatGPT hallucinates like a raver). I’ve found consumer AI tools like ChatGPT or Perplexity useful in some cases and just entertaining in others. ChatGPT especially is creeping into the workplace. I’ve heard of people using it to help optimize schedules and summarize meeting notes. Some people, of course, take it even further, asking the tool to write reports, emails, or code for them. But this is where the main limitation for consumer AI tools shows up: they don’t have context for the tasks you’re asking them to do. Consumer AI tools can’t get much more specific than a general search engine because they don’t have any more context than, well, the entire internet. If you wanted to remember exactly how much vacation time you accrued at your company each month, ChatGPT would be able to give you a general trend in vacation time based on articles about vacation accrual. I asked ChatGPT how much vacation time Coda gets, and…
I tried using a more specific prompt, as you often hear that the limitations of AI are a user error issue, but it’s still not able to give me an answer because it simply doesn’t have access to this information. To get answers on vacation policy at your company from ChatGPT, you’d have to copy and paste the handbook into the chatbot. And if you were to ask it anything more sensitive or current than a static handbook policy? ChatGPT or any other consumer AI would have even less context from which to give you an answer.
So, you really get out of consumer AI what you put into it. If you wanted to transcribe a call and put the main points into a polished email with action points for a colleague, you’d need more than one tool. You could use Otter to transcribe and summarize a call, but then you’d need to copy and paste that summary into another tool, which would then generate a to-do list that you (finally!) copy and paste into your email. Having to provide all the context and connect disparate consumer AI tools is really not all that much more efficient than just sending an email with what you remember.

What is enterprise AI?

 Enterprise AI, on the other hand, is built with business in mind. There are AI app assistants like Notion, Atlassian, and Outlook, which can help employees organize their work data a little better than a chatbot can. But again, these tools are only able to help you with data that you input and organize, so we think of them as the bridge between consumer and enterprise AI. These apps are great for secretarial assistance, but they lack context and can’t drive action or give new insights. I’m sure you’ve noticed every company out there installing chatbots on their websites over the last few years. I consider these a sort of mid-level enterprise AI. Often, they’re LLMs that have been trained to use generative AI to answer customer questions, so they do have to have some specific context for your company’s policies and support options. Understanding user questions and pointing them to relevant help articles or sending their questions on to human staff is a great first use case for enterprise AI. But the most powerful versions of enterprise AI are models built to include proprietary data. That means the highest-level enterprise AI tools collect all the context necessary to give you actually useful answers to workplace questions.

What full-context enterprise AI looks like.

 Imagine an AI tool that could pull up your company’s sales numbers from years before you started there, turn calls directly into to-do lists, or tell you which of your co-workers would be most likely to know the answer to a technical question. That tool would be a bit more useful than one that writes a generic email based on millions of Wikipedia entries, right? For enterprise AI to actually help you in your work, it needs specific tailoring and access to proprietary information. Enterprise AI tools at this level often employ Retrieval Augmented Generation (RAG) techniques to add deep context to their LLMs. They may look and act like a chatbot, but they’re integrated into your company’s data and pull answers from accumulated institutional expertise in a way that would simply never be safe or possible with consumer AI. They can easily grab your latest vacation policies from the handbook. Turnkey tools like Glean can easily slot into your work databases and apps to act like a search engine for your institutional knowledge and your co-workers’ expertise, which is a fantastic way to use enterprise AI. ChatGPT is great for many content generation use cases, but with company-specific context built in, our own enterprise AI offering is much more powerful. Coda Brain is a turnkey AI platform that understands your company data and allows anyone in your org to act on it. It understands all your company knowledge and data by syncing all your tools into a comprehensive, always-connected company brain. More specifically, Coda Brain is a high-context connective tissue for our company: our comms team can get the impact or user numbers they need without bugging data teams, customer success teams can find perfect use cases and track customer feedback, and everyone can generate to-do lists and calendar pings from meeting transcripts. But it can also pull insights from all of our data and apps to spark creativity. I can ask it for ideas on which companies are the most likely potential clients (and then generate the perfect outreach email) or even for specific sales opportunities over $10k. That’s the future of enterprise AI—a team member who knows enough about what you’re doing to actually give you an assist.
AI is only as useful as the information it can access, and you shouldn’t have to spend your time copying and pasting sensitive, proprietary information into consumer AI to get that boost. With enterprise AI, you have much better control over what the model can access, leading to answers that are grounded in the specific context of your workplace. That means it’s a know-it-all you can trust, and one that’s actually useful to you. Which is why we’ve been building Coda Brain.
Learn more about Coda Brain

Related posts

Explore more stories about AI.
Introducing Coda Brain

Coda Brain turns data into action across the enterprise.

Coda + Snowflake partnership

Coda and Snowflake are partnering to turn data into action.

Coda’s enterprise security

Learn how to customize Coda to meet your security needs.