May 30, 2025 · 28m · we-live-to-build
Why Founders Stopped Being the Human in the Loop
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Symfony founder Arpan Nanavati joins podcast host Sean Weisbrot to discuss how multi-agent AI architectures and an inverted human-in-the-loop model are disrupting traditional billable-hour legal services with affordable, outcome-based solutions for startups.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Sean holds 32.7% of the talking time here. How this is scored →
Sean as informed peer 3.8 Guest teaching 4.3 Guest disagreement 1.5 Sean pushing back 1.5
speaking balance: gold is Sean, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 22:03 Pricing taxonomy correction
Arpan immediately rejects Sean's categorization of their business as usage-based, firmly clarifying that charging occurs strictly upon delivering a complete customer outcome.
Hardest push from Sean ▶ 8:52 Confronting hallucinated legal precedentsSean directly challenges the premise of relying on AI in legal services by citing public incidents of generative models inventing case law.
Biggest teaching moment ▶ 11:15 Rethinking human-in-the-loop responsibilityArpan educates Sean on why standard SaaS self-service AI fails in law, showing why the vendor's licensed attorneys must act as the loop instead of the client.
Sean holds their own ▶ 24:04 Breakdown of UI regression and credit wastageSean demonstrates deep practical experience with LLM tooling by explaining how routing config deletions and UI rewrites make credit-based pricing models untenable.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Sean as informed peer | Guest teaching | Guest disagreement | Sean pushing back | Why |
|---|---|---|---|---|---|---|
| Disrupting Traditional Legal Billing Models for Startups | 2 | 4 | 1 | 0 | Sean asks broad opening questions regarding AI opportunities for SMBs and follows up casually about Arpan's parents. Arpan explains the economic structure of the twenty-trillion-dollar services market and how billable hours incentivize inefficiency. | |
| Balancing Deflationary Pricing with Venture Capital Expectations | 4 | 3 | 1 | 3 | Sean probes how Symfony balances deflationary pricing with venture capital growth demands. Arpan explains their transition from SaaS to outcome-based transaction models while maintaining 75 to 80 percent gross margins on a massive addressable market. | |
| Preventing Hallucinations and Inverting the Human-in-the-Loop | 3 | 5 | 2 | 4 | Sean pushes on the risk of AI hallucinating legal precedents in a high-stakes industry. Arpan details their 100-billion-token legal training dataset and reframes 'human-in-the-loop' by explaining that the provider, not the client, must serve as the expert loop. | |
| Open-Source Legal Data and Specialized Multi-Agent Systems | 5 | 4 | 1 | 1 | Sean draws from his own software development experience to ask about model data licensing and multi-agent architectural layers. Arpan educates him on how public legal data is open source while the proprietary execution architecture provides the true moat. | |
| Mid-Roll Audience Support and Channel Subscription Appeal | 5 | 5 | 2 | 1 | Sean discusses his previous startup and equates Symfony's approach to usage-based pricing, but Arpan directly corrects him to clarify the distinction of outcome-based pricing. Sean follows up with his own technical grievances regarding LLM hallucination issues in coding tools. | |
| Legal Accountability and the Long-Term Evolution of AI Services | 4 | 5 | 2 | 0 | Sean suggests acting as the human in the loop for his own software, prompting Arpan to point out regulatory constraints requiring certified attorneys in the loop. Arpan concludes with an overview of deflationary AI trends across professional services. |
Statements from this episode (14)
Nanavati: Professional services is a $20T market dominated by large corporations
“Services as an industry today is about a 20 trillion dollar market but most of that is consumed by larger corporations or HMX with small to medium sized businesses not really having an ability to reach higher quality services, in our case legal services.”
Nanavati: AI legal counsel delivers results equal to or better than human lawyers
“With AI, they have the ability to use a high quality AI legal counsel. Which gives them similar to sometimes even better, faster results than they would get from a traditional lawyer.”
Nanavati: Law firms' billable-hour model is ripe for disruption
“The business model of legal services and law firms is based on an hourly model. And it's all about how can you increase the billable hours? Which means you need a very, very fat wallet. You even pay for things like reading an email. Hey, can I get on a call ty…”
Nanavati: Cimphony moved from SaaS to outcome-based transaction pricing
“We've moved away from the SaaS model and it's a pure transaction based model. Like we pay for the outcome. You pay for alignment of the incentive with what the customer needs, rather than just paying for something when the customer does.”
Nanavati: Cimphony generates 75% to 80% margins
“And the fact that the cost of our infrastructure is low compared to the human power cost of the traditional model it allows us to generate very strong margins of 75 to 80% margins.”
Nanavati: Cimphony digested a 100-billion-token legal dataset dating back to 1975
“We've digested about a hundred billion token legal data set and case law from the last 50 years dating back to 1975 for example, across Supreme Court all the way down to the county courts.”
Nanavati: Legal AI requires the provider, not the customer, in the loop
“Generally when AI products sell human in the loop, what they're selling is the customer becomes the human in the loop, which is the customer still self-servicing the AI, which may make sense in a lot of, in legal industry, customer generally doesn't know what …”
Nanavati: Data is not a moat in legal AI, execution is
“Within the legal domain data by itself is open source and data in itself doesn't become the model. It's the application of that data set. It's the application of how you tune the models and the fact that you're building a tech layer between the customer and te…”
Nanavati: Cimphony has built 40 to 50 specialized AI legal agents
“What we've done is we've built about 40 to 50 agents that are highly specialized for their particular use case within their particular work call.”
Weisbrot: Stopped using Lovable due to severe hallucinations and broken routes
“This was a big reason why I was using lovable before and I got away from lovable because it was hallucinating. The deeper into the project, the more it hallucinated to the point that it was rewriting UI. It was removing pages from the route configuration.”
Nanavati: Cimphony considers itself a service company, not a software company
“We're not a software company. We're a service company is what I say.”
Weisbrot: Human-in-the-Loop Services Command Higher Prices Than Self-Serve SaaS
“I think you can make a lot more money a lot faster because you could charge more by having that human in the loop to provide the service.”
Nanavati: Cimphony delivers $1,600/hour partner-level legal quality for $10 an hour
“Partners at law firms will charge anywhere from eight 50 to 1600 dollars an hour. And with us, they still get the similar type of quality at 10 dollars an hour in type of cost, which is incredible savings for a similar type of quality, what they could expect f…”