Hear the latest insights on AI from leading investor and SaaStr fan-fave, Tomasz Tunguz of Theory Ventures. At SaaStr Annual, he was joined by Jordan Tigani, Founder and CEO of Mother Duck Maggie Hott, GTM at OpenAI, and Sharon Zhou, Co-Founder and CEO of Lamini to discuss the new architecture for building Software-as-a-Service applications with data and machine learning at their core. Together they explore the transformative role of data, which has become an integral part of the production stack, and delve into the implications of this shift and explore how leveraging data within the architecture empowers developers to create more robust, intelligent, and scalable SaaS solutions.

LLMs are transforming the way people use software, and these AI and data-driven companies share first-hand how they’re seeing this change take place. 

LLMs are Democratizing Tech

Someone was giving a course on data thinking at Princeton, teaching people how to ingest, query, and visualize data. He wasn’t teaching programming languages or higher-level tools that do this for you. He was just using ChatGPT to generate Python code. 

It was then that MotherDuck realized that AI would change who could do this stuff, the skills people would need, and the tools that would be built. 

There are a bunch of unknowns, but it’s clear how much LLMs are democratizing technology. It’s easier for someone who isn’t as sophisticated with SQL because they can ask questions of their data without the 3-6 month learning curve beforehand. This is being adopted broadly in the Enterprise. 

How Enterprises Are Adopting This Technology

Zooming out at a technical level, public data is running out for all of what LLMs can take advantage of. The next frontier is in Enterprise, and the best LLMs for the next wave will be Enterprise LLMs. 

The Enterprises at the forefront are starting to adopt techniques around fine-tuning their models to personalize them for every single user and customer, where data doesn’t leak between customers, so it’s highly personalized for any kind of complexity and use cases.  It has further democratized all possible use cases beyond what anyone can imagine. 

Who will be responsible for optimizing this software — the vendor or subject matter experts? 

There’s tension between the two. AI expertise might take a couple hundred top AI researchers to train them. However, it’s likely the domain experts who are driving the best models. The typical developer, infrastructure developer, or front-end developer who can access and manipulate data can help develop the trajectory of these models and build out the next generation of LLMs for the Enterprise. 

Revelations in the Market After the Launch of OpenAI for Enterprise

OpenAI shares three things about their journey to building ChatGPT Enterprise, learnings, and what is coming. 

1. When ChatGPT launched early last December, it saw millions of knowledge workers using this tool every day for their jobs. But there is the security IT side of the house saying, “We don’t want our corporate data trained into these models.” So, now it has data training turned off and has SSO, SOC2, and all the Enterprise features. When they set out to launch, they selected as wide a variety of design partners as they could in different industries and use cases to see how healthcare, banking, or consulting use these tools. 

2. A big learning during this process was the importance of developing alongside customers. They assumed everyone knew how to use it, but everyone uses it differently. So, they had to think about how to work with these design partners and the world to develop very specific use cases for their personas. Every single vertical can use it within every single company. Another learning was to invest in customer success work early on. They had Slack connect channels and weekly syncs and ensured everyone felt good about it. 

3. The biggest thing OpenAI is working on is customization. Imagine having an entire company’s data infused with ChatGPT Enterprise so they could query all the top sales deals and common trends and have all of that come back. 

Another thing they’re building is a business version, self-service for SMBs. There’s a huge demand for SMBs, startups, and founders, and the use cases are endless. OpenAI’s Advanced Data Analytics is the most powerful tool the team has ever seen. It’s essentially a super calculator that can be used to look at leads, opportunities, win and loss rates. Then, it synthesizes anything that would take someone a day to parse through and does it in minutes to seconds. You could feed in transcripts from Gong and have it analyze that and write a next steps follow-up email that’s beautiful and personalized. Usually, this takes someone 30 minutes to do, and it spits it out in 10 seconds. 

A Merger of Two Personas — Software Engineer and ML Engineers

Ten years ago, your data stack involved systems that produced data. Then, you put it in Cloud data lakes before it reached machine learning. Now, those ML engineers are becoming part of the core workflow. ML engineers are in the path of production, and it’s a mix between a classic software engineer who ships code and ML experts bringing in different tech and putting it into production. 

There’s a merger of two personas that historically have been separate. The landscape is rising up for how to make those people as successful as they can be. Merging these teams is really important to ship these models in production. Historically, it was an R&D team with a thin line to production and often failing that. Many teams failed before, and by merging the two and speeding up performance, many will succeed. 

Language Matters — Will It Be English or Python? 

As those teams merge, an interesting debate about what language they will write in has emerged. Will it be Python or English? 

As people start to use natural language to do tasks, that can help accelerate the merger. As we move towards typing out a paragraph in English and turning that into a production system, that might further peel off workloads and help merge the teams. 

If someone who isn’t proficient in Python needs access to more sophisticated data, they can speak English, which could be 10x or 100x productivity. 

LLMs Change How Companies Go to Market

Massive changes are happening on the tech side of an organization. What about the GTM side? 

The short of it is that AI will fundamentally change how sales roles operate and look. First, let’s touch on how sales has changed over the years. Every year, there is a new step. In 2017 or 2018, it was easy to prospect in companies. The bread and butter of any sales org was to generate opportunities and customers. Then, tools like Salesloft and Outreach emerged and took prospecting mainstream. 

What do sales reps need to do next to get in front of buyers? Personalization. It’s all about researching a company or customer, determining if you went to the same school, and finding ways to connect. Research and personalization can happen in less than a minute with AI. You can write an email that sounds very personalized yet was written by an LLM. 

The next era of sales has to think deeply about how to interact with the customer. You can no longer ask a million discovery questions. You need a strong point of view on how your product will save x company money or solve their pains. The beauty of LLMs is they can help you do that research in advance. Gone will be the days of spending 2-3 hours researching and practicing discovery questions before a meeting. 

Now, it’s 20 minutes max, and the LLM can roleplay with you on questions and value statements. On the other side, LLMs will save so much time for the sales professional in terms of administration tasks like follow-up emails and logging notes into Salesforce. You can feed the info back into the LLM and prompt engineering to write summaries and build slide decks. The sales rep and sales team will have to step up with value selling. 

Ways We Use AI Daily Now

People interacting with data are changing their lives. Some people do a lot of slide decks, and Midjourney and generative AI make it easier to have a good image. As the tech gets better and the operator gets better at prompt engineering, some of the quirks, like too many fingers or ducks without wings, will go away. Other panelists use ChatGPT Enterprise, Slack, and Gong daily. And others use AI to generate blog titles with better clickthrough rates. 

With AI, industries like education will be completely upended, democratizing the future of education for everyone and making access equal and equitable for people all around the world. No one really knows what the next five years will look like with AI. But the only clear thing is that people will be doing things differently. Roles, how businesses operate, and who they need to hire will dramatically change. 

Key Takeaways

  1. Democratization of Technology through LLMs: LLMs are making technology more accessible. For example, ChatGPT simplifies tasks like writing Python code, reducing the need for extensive programming knowledge.
  2. Adoption in Enterprises: Enterprises are at the forefront of adopting AI, with a focus on fine-tuning models for personalized user experiences. This includes ensuring data privacy and customizing applications for complex use cases.
  3. Role of Domain Experts and Developers: There’s a growing importance of domain experts in optimizing AI models, alongside traditional developers. This collaboration is driving the development of next-generation LLMs tailored for enterprise needs.
  4. OpenAI’s Journey with ChatGPT Enterprise: OpenAI’s experience with ChatGPT Enterprise highlights:
    • Rapid adoption by knowledge workers.
    • The necessity of balancing security and functionality.
    • The importance of customer feedback in developing specific use cases.
    • Ongoing efforts to customize and enhance the tool for various business sectors.
  5. Integration of ML Engineers and Software Engineers: The line between ML engineers and software engineers is blurring. The integration of these roles is crucial for efficiently deploying AI models in production.
  6. Language of Future Programming: There’s a debate whether future programming will be dominated by Python or natural language like English, with the latter potentially offering greater productivity gains.
  7. Impact on Sales and Marketing: AI is revolutionizing sales processes by enabling rapid personalization and research, transforming the way sales professionals prepare for and engage with customers.
  8. Daily Uses of AI: AI tools are being used in various daily tasks, from creating slide decks to generating blog titles. This is changing roles in businesses and potentially disrupting industries like education.
  9. Uncertain Yet Transformative Future: The exact future of AI is uncertain, but it’s clear that it will bring significant changes in how businesses operate, the nature of various roles, and the skills required in the workforce.

Related Posts

Pin It on Pinterest

Share This