Third Year at SaaStr: Reflections

After reflecting on last year’s event, I wanted to do the same this year. Here are some key observations:

Insights:?

Attending sessions in person sparks my curiosity. I enjoy going to broad sessions to pick up new buzzwords, which I later dissect. Then, I focus more specific sessions or braindates to learn something right away. This year, AI was the big focus, and a few concepts caught my attention to dig deeper.?

• Onboarding with AI:?It’s obvious that having a strong AI-driven customer onboarding strategy is crucial, but how early AI should be involved? As a former SC, I’m thinking how AI can handle most of the disco, so folks interact with a rep/SC who already know their business. Credit to Seema Amble for the thought.
• AI as the "Main Branch":?Use AI to figure the core concept and then build a structured model around it. This ensures that everything developed is centered around the most important ideas. If customer retention is your software's core focus, AI would identify this as the main priority. It would then structure all features, enhancements, and strategies around improving customer retention. Honestly, I'm not sure how to implement this but will do some research - goal is to use AI to set up guardrails to ensure dev efforts align with your primary goal.?
• Model-Driven > SaaS & LLMs as a Step-Change: I like the buzzwords, but the key point is to focus on creating tailored models to solve specific problems, rather than just offering SaaS. The models are the core product, with the software being the tool to implement them. LLMs represent a major shift from traditional B2B software workflows, as they can analyze data to generate deeper insights and solutions beyond predefined processes. The takeaway for me is to focus on repeatable workflows and data that add value for the customer. Be specific and remember that customer 1000 SHOULD benefit even more than customer 10 because of the accumulated data.
• Multi-Modular AI:?Use different AI components, each specialized in a specific task, working together as one system. For example, one AI part might analyze data, another might handle customer queries, and another could predict trends. For example, you might have one module with a deep learning algorithm for data analysis, another module using natural language processing for customer support, and a third employing machine learning for predictive analytics. These talk with each other through APIs or a central orchestrator. This architecture allows each AI component to focus on what it does best, while the system as a whole delivers benefits.
• Models Over Prompt Engineering:?Strength of AI models is more important than the techniques used to create prompts. Well-developed model > the methods used to guide it. For example, if an AI model accurately predicts customer churn, it's more valuable than focusing on how the prompts were crafted to guide that prediction. The end result—the model's effectiveness—is what truly counts.

?Exhibitors:?

• Fewer exhibitors this year, and fewer companies doing the same thing as in prior years when VC capital rained on everyone. There were only two BI companies,?lowest I’ve seen at any recent conference. One standout company was Zencoder , which feels like a mix between Devin Devin and GitLab Copilot. Excited to try it.?
• Big emphasis on embedded integrations and analytics. "Embedded" means you can see reports and charts right in the app without leaving it. Embedded integrations mean vendors handle the integration work, so customers don't need extra middleware. As a B2B software vendor, I'd team up with a company that offers these solutions to make it easy for customers to integrate their apps instead of building connectors myself.
• Account Executives explaining products were much better this year.

Cool Concepts:

1. AI War Rooms:?Some companies have special teams dedicated to figuring out how AI will transform their entire business. This concept reminded me of what Reid Hoffman did at Netflix. Credit to Jason Boehmig for this idea.
2. RPA vs. Integrations: RPA (Robotic Process Automation) and integrations can either compete or complement each other in automating tasks. RPA mimics human actions across different systems, while integrations directly connect systems for smoother data flow. They compete when both can achieve the same task, but they complement each other when RPA is used to bridge gaps where direct integrations aren't possible.

Fun Use Cases:

1. AI Travel Recommendations: Sara Du from Alloy Automation uses AI for personalized travel tips by comparing places to familiar ones—like asking what the "Pac Heights" of Paris is to find a similar vibe in another city. This inspired me to embed AI in 3PA, allowing it to learn from user preferences and suggest similar metrics or goals based on their role/experience, adding more context during goal-setting.
2. AI Negotiating Bills: Joshua Browder discussed how his DoNotPay AI agents negotiate bills with Comcast on behalf of customers. Well funny enough, Comcast responded by deploying their own agents, so it became a scenario where an AI agent was negotiating with another AI agent. His challenge was ensuring the bot doesn’t lie, which got me thinking about making sure 3PA doesn’t make critical decisions without human oversight much deeper.
3. Tax Explanations with AI: from Kate Jensen Anthropic discussed how TurboTax’s AI explains why you’re paying a certain amount in taxes, which I found pretty cool. The added why with context should be embedded across every application.

Main Thing For Me:

Got a lot of insight into the real-life applications of OKRs and KPIs during my braindates. I also learned the most from discussing specific questions about my product with technical experts. I’ll write more about this later.

Overall, as I mentioned before, it was a great conference, and I enjoyed reviewing my notes.