932: Should You Build or Buy Your AI Solution? With Larissa Schneider

Unknown Source October 17, 2025 29 min

artificial-intelligence generative-ai startup investment anthropic

31 Companies

53 Key Quotes

4 Topics

🎯 Summary

Summary of Super Data Science Podcast Episode 932: Build vs. Buy in the Age of AI with Larissa Schneider (Unframe)

This episode of the Super Data Science podcast, hosted by John Cron, features Larissa Schneider, Co-founder and COO of Unframe, a startup that has raised $50 million to revolutionize enterprise AI deployment. The central theme is challenging the traditional “build vs. buy” dichotomy for AI solutions by offering a hybrid model that delivers the speed of buying with the customization of building.

1. Main Narrative Arc and Key Discussion Points

The conversation centers on Unframe’s novel Managed AI Delivery Platform, which leverages hundreds of pre-built, reusable “Lego brick” components. This approach allows them to rapidly tailor complex AI solutions to specific enterprise needs, promising ROI in weeks rather than years. A key differentiator is their no-commitment, no-cost Proof of Concept (POC) phase, contingent only on the customer realizing measurable business value. This model directly addresses the high failure rate of enterprise AI projects (citing the MIT/NAND group report showing 95% failure to show value).

2. Major Topics, Themes, and Subject Areas Covered

AI Solution Delivery Models: Comparing traditional build (slow, custom, high overhead) vs. buy (fast, generic, low customization).
Unframe’s Hybrid Model: Described as a blend of services and SaaS, offering tailored solutions built from standardized components.
Business Impact Focus: Emphasizing the necessity of starting with clear ROI and Key Performance Indicators (KPIs) rather than just exploring available technology.
Enterprise AI Use Cases: Categorized into Observability/Reporting (BI, search), Extraction/Abstraction (unstructured data from contracts, financials), and AI Workflows/Automation.
Founder-Led Sales: The importance of founders remaining involved in sales to identify repeatable patterns and guide the development of new platform “Lego blocks.”

3. Technical Concepts, Methodologies, or Frameworks Discussed

Lego Brick Architecture: A modular approach where complex AI capabilities are broken down into generic, reusable building blocks that can be quickly assembled and customized.
Business Impact Analysis: A structured upfront process to define measurable success metrics (KPIs) before development begins.
Handling Unstructured Data: Specific mention of extracting value from legacy documents like contracts, financial statements, and PDFs using abstraction techniques.
IT Operations Correlation: Using AI to correlate data across disparate enterprise tools (ServiceNow, JIRA, Datadog) to reduce manual effort and time-to-resolution.

4. Business Implications and Strategic Insights

Speed to Value: Unframe drastically cuts deployment time, moving from deep dive workshops to production-ready solutions in days, which is critical given how fast the AI landscape evolves.
De-risking AI Investment: The “pay only when you see value” model significantly lowers the barrier to entry and addresses executive fears about sunk costs in failed POCs.
Scalability Challenge: Large enterprises (like the Wall Street bank mentioned) face backlogs of thousands of potential AI use cases that internal teams cannot possibly handle, necessitating scalable external partners like Unframe.
Build vs. Buy Strategy: The general advice is to buy for quick wins and non-core functions (e.g., marketing content, ticketing search) and build only for core IP or strategic competitive advantage (e.g., core banking systems). Unframe positions itself as the “perfect middle ground.”

5. Key Personalities, Experts, or Thought Leaders Mentioned

Larissa Schneider: Co-founder and COO of Unframe.
John Cron: Host of the Super Data Science podcast.
Shy: Co-founder and CEO of Unframe (previously co-founded NoName Security).
NAND Group (MIT): Referenced for the report detailing the high failure rate of AI projects.
Claude (Anthropic): Mentioned in a sponsor segment as a powerful AI collaborator for research and strategy.

6. Predictions, Trends, or Future-Looking Statements

AI necessitates a fundamental reset in how enterprise software is built and delivered, moving beyond “one size fits none.”
The pace of AI innovation means in-house solutions risk becoming outdated almost immediately after launch.
The future of enterprise AI adoption relies on scalable, partnership-based execution rather than isolated internal projects.

7. Practical Applications and Real-World Examples

IT Operations: A Fortune 10 insurer uses Unframe for a single pane of glass to correlate logs and tickets across their entire tech stack, drastically improving resolution times.
Contract Management: Abstracting data from large volumes of contracts (leases, MSAs) stored in SharePoint to provide scalable insights on trends and expiration dates.

8. Controversies, Challenges, or Problems Highlighted

AI Project Failure Rate: The industry standard of 95% of AI projects failing to deliver value is the primary problem Unframe seeks to solve.
Executive Pressure vs. Strategy: Many leaders rush to execute on AI without defining clear KPIs, leading to wasted resources on non-impactful POCs.
Agent Proliferation Risk: Concerns about business leaders building thousands of unmaintained, autonomous agents without proper governance, especially in regulated industries.

9. Solutions, Recommendations, or Actionable Advice Provided

Define KPIs Upfront: Whether building or buying, success hinges on clearly structured metrics (accuracy, completeness, cost reduction,

🏢 Companies Mentioned

TraceBlip ✅ tech

West Coast ✅ unknown

Five Minute Friday ✅ unknown

Claude Pro ✅ unknown

When I ✅ unknown

Wall Street ✅ unknown

In HR ✅ unknown

Like ChatGPT ✅ unknown

NoName Security ✅ unknown

If I ✅ unknown

And I ✅ unknown

New York ✅ unknown

Tel Aviv ✅ unknown

Unframe AI ✅ unknown

And Berlin ✅ unknown

💬 Key Insights

"They said, 'Here are a thousand data samples. When we go through the testing for this POC, we want 96% accuracy at a completeness rate of responses of over 90%.' I'm like, 'Okay, wow.' This is like the first time... that we saw someone was so prepared, and they really knew what they were working towards."

Impact Score: 10

"If people want to have successful AI projects in their organizations, they need to be clear on what the KPIs are upfront. So, whether that's cost or time or accuracy, they need to have that clearly structured so that whether they're working with Unframe or whether they're doing this on their own, how do they know the project is a success unless those KPIs are defined up front?"

Impact Score: 10

"AI moves so fast. By the time that you launch something in-house, it's best-in-class on that day, but the following week, I can promise you there's a new model, there's new AI capability, new research that has come out. It's probably already outdated again, right?"

Impact Score: 10

"I think most cases like this are probably the 80/20 rule: buy the things that you need to get quick impact on your business—stuff that is simple use cases, maybe quick wins—those things. But what you could consider building, and I totally understand when you want to build that in-house, is anything that has IP ownership concerns, you want to increase your strategic positioning, things like that."

Impact Score: 10

"NAND group... said that 95% of AI projects fail to show value in production or never make it to production. It's only 5% of AI projects that get into production and then provide business value."

Impact Score: 10

"My backlog currently is 1,670 use cases that the business has brought to me that I have to execute on by the end of 2026. So, I could have as many internal engineers as I possibly wanted; I will never get them done in-house. So, I need a scalable approach."

Impact Score: 10

📊 Topics

#artificialintelligence 71 #generativeai 13 #startup 8 #investment 3