The AI incumbent race

Hubspot, Slack, Salesforce and Discord announced AI features. There’s a lot to unpack in these announcements and I’m going to dig into them in this week’s essay.

Before diving into this week’s piece, I wanted to acknowledge the rollercoaster ride in tech over the last few days. Silicon Valley Bank, a bank for startups and technology, collapsed on Friday. Many people were able to get their money out, but many others were not. I don’t have much to add to the discussion. It just felt awkward to write this week’s piece without acknowledging it. If you are impacted as an investor or founder, I’m sorry and I hope the situation is resolved soon. If there is anything I can do to help, please reach out.

In the meantime, here are my thoughts on the AI announcements from this week.

Chatspot AI <> Hubspot

Hubspot’s CTO Dharmesh Patel announced Chatspot AI via a tweet and on ProductHunt. It’s inspiring to see the CTO of a $18 billion SaaS company hacking away at an AI tool. The announcement is here and I recommend watching it. It’s one of the best AI demos I’ve seen.

One of my biggest takeaways from the demo is Dharmesh’s articulation of why this is impactful. To date, our experience with software is to point and click. AI transforms this into a declarative approach: ask and you will receive.


The Hubspot demo focusses on 3 buckets: sales / crm, reporting and marketing. Imagine you’ve just had a conversation with a prospective customer. This information needs to be added to your CRM. According to Dharmesh, it can take 20+ clicks using the imperative approach. With Chatspot AI, it’s a prompt and single click.

The second callout from the demo is in the “reporting” bucket. With a single prompt (”How much revenue did we make last month?”), a user can get critical data out of the CRM. Although this sounds trivial, many users within an organisation might not know how to get the information out of Hubspot. They might need the help of other people in the team to get the information out. AI can help save time and make this more efficient.

So my two main takeaways from this are that AI can:

  1. Streamline a specific task or process
  2. Enable people to do a task they could not otherwise do

Slack & Salesforce

Slack and Salesforce (which acquired Slack) announced a bunch AI features this week. They also announced a $250 million fund that will invest in AI companies.

OpenAI apparently build the feature with Slack that allow users to do 3 things:

  • Summarise conversation from a thread
  • Draft messages
  • Instantly find answers to a question using conversational history from Slack

Separately, Salesforce also announced Einstein GPT. The primary goal of Einstein is to help users with AI-created content. This could be a sales email or a response to a customer support request. It sounds like Einstein is basically a conduit between a customer’s proprietary data on Salesforce and an AI model. Users are able to choose OpenAI’s model or an alternative.

Two main takeaways from this one:

  1. OpenAI’s GTM strategy is interesting. They built the product for Slack. They are big users of Slack and probably figured that it was a great way to test ChatGPT’s capabilities in a business context. It’s also a great way to test, learn and iterate quickly.
  2. The approach with Einstein to choose OpenAI or a different AI model is interesting. Salesforce is essentially saying, we’ll do the work to combine your external and internal data, and you choose the AI model you want to use. A big reason for this might be privacy. Whilst OpenAI’s policies prevent it from using your data for training, some companies might want full control via a custom, open-source model running on their own servers.


Discord announced 3 AI features this week:

  • Clyde AI: Clyde is an AI bot that lives in Discord and people can talk to. It leverages OpenAI’s technology.
  • AutoMod AI: to help moderate discussions on Discord servers.
  • Conversation summaries: to generate quick summaries based on the conversation in a server.

Separately, the following paragraph also stood out to me from the Discord announcement:

more than 30 million people already use AI apps on Discord every month. Midjourney’s server is the biggest on Discord, with more than 13 million members bringing their imaginations to pixels. Overall, our users have created more than 1 billion unique images through AI apps on Discord

One big takeaway from the Discord announcement is the focus on privacy. Despite leveraging OpenAI’s technology, data will be private and opt in. Your Discord conversational history will not be used to train OpenAI’s models.

The second takeaway for me is the focus on tedious tasks. As a Discord user, I have felt the pains first hand of not being able to keep with conversations in a server. From talking to friends, I know how difficult moderation is. Discord has gone after tasks that it knows are the most painful for users, and hopes to use AI to solve them.

To close

A few common themes from these announcements that I think are worth calling out:

First, every product has looked at it’s user experience and the things that are frustrating or taking time. They are experimenting with AI to see if they can resolve them.

Second, they are taking risks. They are moving fast even if the solution isn’t a 100% perfect. This is clear indication that the technology is top of mind for most of these companies.

Third, there is a clear focus on privacy. Business users are going to have questions on how their data is processed and stored. They will want to know if it’s used for training of models. Two of the announcements (Slack + Salesforce and Discord) address these concerns directly in their announcements.

Finally, OpenAI’s GTM strategy here is interesting. They are actively working with other tech companies to develop business applications using their technology. It makes sense because this is category creation — we’ve never really had the ability to generate or summarise things with LLMs like we can today.