How CEOs Build with AI - Sean Devine of XBE
Tue Apr 04 2023
AI in XBE
00:00 - 06:01
- Sean Devine, founder and CEO of XBE, is applying AI tools inside his company right now.
- AI is changing things in his hiring and team structure.
- Artificial intelligence will be a sea change on the order of mobile or maybe even computers themselves.
- Many people are overwhelmed with new possibilities and not sure where to start with AI.
- XBE provides mission-critical software to horizontal construction companies that build roads, bridges, tunnels, airports, etc., usually funded by the government.
- Sean talks about how he's applying AI inside his product in AI features and how they use AI to build the product.
XBE and Sean Devine
05:49 - 12:31
- Sean Devine, CEO of XBE, talks about his company that provides horizontal construction companies with operations management solutions.
- XB's software platform is wide and deep, covering scheduling and planning to dispatch and monitoring and analysis and continuous improvement and safety and administration.
- The team size is roughly 21 people, half in the US, half in India. The company is minimally funded with only $150,000 capital ever put into it.
- Devine believes in getting maximum leverage out of the team size they have now while continuing to grow. He learned programming as a way to create leverage for himself.
- He was self-taught by solving real problems he faced at work. He started learning how to program when he was 32 years old.
Sean Devine's Journey
12:12 - 18:14
- The speaker learned programming mid-career to solve a problem with money attached.
- He entered hackathons on weekends and became the host of the Ruby on Rails podcast.
- Eventually, he became a decent programmer and used his skills to differentiate his company Partidge through pricing sophistication.
- He built an automated pricing system but realized that the transportation management system they were using was not sufficient for customers to book shipments themselves.
- He started XPE without raising any seed money by leveraging his skills in product management, architecture, front-end development, strategy, and finance.
- He believes that pouring fuel on the fire can lead to closing deals that do not have product market fit and mistaking dynamic for product market fit.
GPT-3 and Content Production
17:53 - 24:38
- Early stage businesses are like being sick, it's helpful to know where you don't have it.
- Funding causes a break between experience and reality of the fit between you and the market.
- Expertise is now abundant, so why take money if you don't need it?
- Low capital requirement software businesses with high pace of iteration will change how people start companies.
- GPT-3 can be used as an aid to content production by collaborating with it.
- Communication is leverage, instead of having one-on-one conversations, use content to communicate with thousands of people at once.
- Use GPT-3 for chat as a collaborator starting from the most upstream place possible.
Creating a Newsletter with GPT-3
24:20 - 31:20
- The speaker discusses the process of creating a newsletter using GPT-3.
- They start with a big idea and create an outline, which they flesh out with examples and research.
- They highlight three software features to use in the newsletter.
- For each feature, they paste in the release note and write a how-to that can't be more than two sentences.
- The speaker rejects most ideas generated by GPT-3 and only uses the good ones.
- They find it fascinating that most people don't tell GPT-3 to generate lots of ideas.
- The current GPT-4 has 8,000 tokens or 6,100 words, but will soon increase to 32,000 tokens.
- Context is powerful for GPT-3 because it combines pre-trained information with specific content.
- The speaker finished the newsletter in about 45 minutes while watching a basketball game.
Using GPT-4 for Problem Solving
36:52 - 43:40
- Built a library that could solve the exact problem in one day
- Asked 180 questions during the day to technically implement everything
- Generated a ton of options and selected the ones liked instead of trying to get the answer right
- Asking many questions is a counterintuitive way to approach things but it works
- GPT-4 is more capable than GPT-3, but there's still a possibility of generating false information or being misled
- Be clear about your goal, provide context, and have it explain its thinking before doing any work just like a person would do when solving problems
- The secret to problem-solving is knowing what problem you're solving and making the problems smaller
AI in Content Creation and Collaboration
43:12 - 49:53
- Creating a project plan involves jotting down the required steps and going down the rabbit hole to find answers.
- The collaboration process is fluid and involves give-and-take.
- GPT-3 is good enough to fundamentally change everything, even if it never gets better.
- AI-infused features added in the last four months include rewriting incident headlines for clarity, suggesting safety risks based on past incidents and general knowledge, generating toolbox talks for foremen, making all generated content available in audio format, and using Google for audio generation while Open AI is used for other tools.
- Audio generation was added because many users don't like to read too much and it's more convenient when working at job sites.
AI Tools and Engineers
49:25 - 56:20
- Most of the open AI tools are now on GPT-4 beta.
- Engineers have mixed feelings about working with these tools.
- Some engineers are excited while others don't see it any differently than other features.
- Everyone has some fear of it all.
- CoPilot is good for 25% productivity increase, while chat GPT and other zoomed-out tools are good for another 25%.
- Chat GPT is good at refactoring code and answering questions about algorithms.
- Chat GPT was used to find a production bug that took down buttons in the app. It wrote a list of the most likely culprits, one of which was the issue.
- Chat GPT was also used to help debug a drag and drop bug in a feature being written by Pankaj.
- The speaker is obsessed with motivating everyone to fully embrace these tools and communicate what's possible with them.
- The speaker's mission is to make clear
Growth, Leverage, and AI
55:53 - 1:02:18
- Growth solves the scaling problem caused by a doubling of capacity.
- If not growing, productivity must be absorbed into growth, lower costs or better prices.
- Expectations of output for each individual person have gone up due to new tools and increased efficiency.
- Quality jump is not talked about as much as it should be.
- Prepping before meetings can make you five times better at attending them.
- Creative work can also benefit from the method of generating multiple options and fleshing out the best one.
- 'Product ties operation' means taking all the work that operations staff members do to help train new customers, provide customer support, explain things, problem solve and automating it using AI.
- The company used a big corpus of knowledge they had in order to build an AI version for most of what their operations staff does.
Recruiting and Problem-Solving Skills
1:08:25 - 1:14:51
- Recruiting is changing a lot right now. Problem-solving skills matter more than technical expertise.
- Technical roles are not that valuable in the very near future.
Success in AI and Characteristics
1:14:23 - 1:21:02
- Recruiting is focused on problem-solving skills, not technical expertise.
- Identifying people who are good at asking questions is a skill to interview for.
- The style of work required in this field is physically difficult and requires stamina and work ethic.
- Finding people with low ego that won't be upset by the identity deconstruction caused by AI advancements is important.
AI and Human Input
1:20:39 - 1:27:14
- AI can't work without human input and presence.
- Subscribing to AI-centric newsletters is a good way to learn about AI.
- Ben's bites is a recommended newsletter for AI learning.
- Reading details on AI and implementing them creates its own energy and feedback loop.
- XV is in a good spot because the product is complicated and expansive.
- AI helps make non-AI features, but there's insufficient talk about it in the community.
- XV deploys AI in sales and marketing game to research prospects and create bespoke plans for target markets.
- The threat of being caught by competitors is real due to the speed of innovation, but XV can press their advantage by going faster.
- Margins in software industry will increase due to AI innovation.
Impact of AI on Industries
1:26:57 - 1:34:03
- The price of things should collapse in a lot of areas.
- Revenue and margin percent will decline given leverage.
- Leverage works both ways, everything that was good is going to turn bad.
- Those who are many years into a very complicated product within the software industry, especially if they have existing customer relationships and integrations and practices around it are relatively safe.
- The simpler the product, the weaker the moat for sure.
- There's a lot of software with what is truly like a technical edge and a foundation that is like an engineering problem that hasn't been solved before, some anyway, maybe not the majority. But enough that that's also a different case.
- I would be very concerned if I was a point solution. Like incredibly concerned. More the more unit task the thing is, the more specialized, the more even if it's technically advanced. I don't think there's any point solution that's just very good at
Facing the Changes in AI
1:33:40 - 1:40:03
- The degree of change is scary and will take a lot of courage to face.
- Running from it, denying it, or lashing out at it are futile and cowardly.
- It's important to acknowledge that many people will be impacted by the changes.
- Problem solving, stamina, and identity are key characteristics for success in this space.
- Technical expertise is not necessary for success in AI.
- AI has applications across all industries and fields.
- The speaker is interested in connecting with others on the topic of AI.
1:39:33 - 1:45:28
- The guest's Twitter handle is 'barely known'
- Hey Kayla, the AI feature of the guest's company, is named after his daughter who works for XBE
- The guest created a website called BayKayla.com to help his daughter find a suitable match
- The website has a gallery of photos and a text box where interested people can ask her out
- Listeners of this podcast are the target demographic for finding a match for his daughter
- Episode 38 with Chris Ho and Robert Hayes of Athena and episode 28 with Janine Sidel are recommended for listeners interested in people leverage and HR function respectively
Sponsorships and Recommendations
1:45:02 - 1:47:16
- Accredited investors can invest in early stage companies alongside the host and his partners in rolling fun.
- Sean, the guest, is an investor in the fun.
- Bread is a sponsor of this episode and can help you hire a talented team to build excellent software.
- Leverage is possible in modern society and there's a revolution going on. Get exposed to it and make money along the way.
- The podcast Super Friends is a monthly meeting of five podcast producers who share best practices, insights, and discussions to help make you a better podcaster.