Our AI Philosophy: Why We Haven't Built AI Into Basa...yet (But Use It Everywhere Else)

Three months into building Basa, I sat in a conference room watching a demo of AI-powered contract negotiation software. The sales pitch was polished: natural language processing that extracts key terms, algorithms that suggest optimal pricing, chatbots that handle back-and-forth with talent. Impressive capabilities, demonstrated with clean examples that made the technology look ready for prime time.

Then the sales rep pulled up a sample creator contract and asked the AI to identify potential issues. It flagged several concerns—some legitimate, some nonsensical. When pressed on accuracy rates, he admitted they were hitting about 97% precision on standard agreements.

"Ninety-seven percent is pretty good," he offered.

I thought about what 97% means when you're processing three hundred deals a month. It means nine contracts with errors—mistakes that could cost tens of thousands of dollars, damage relationships, or create legal liability. Those aren't acceptable odds when you're dealing with legally binding agreements where someone's career or brand reputation depends on getting the terms right.

That demo crystallized something I'd been grappling with since founding the company: the gulf between AI's capabilities and the precision threshold that contractual work demands.

The Seven-Month Deep Dive

After leaving Religion of Sports, I spent seven months researching AI's impact on content creation. Not browsing articles or attending conferences—actually diving deep into the technology, its limitations, its trajectory. At the end of those seven months, I came away with a hypothesis that shaped everything we've built: AI is going to slash production budgets and production timelines. Lower the barrier to entry for high-volume, high-velocity content creation, and you get an explosion in the volume of content.

That explosion in content volume won't be matched by an explosion in attention. Attention is finite. Over time, the value of individual pieces of content will gradually decline because there's just too much content competing for the same eyeballs. To replace lost revenue from less valuable content, producers are going to have to do a higher volume of deals. Orders of magnitude more deals.

Having coordinated thousands of deals on both the talent side and the producer side over fifteen years—from managing major label band Delta Rae to producing at Religion of Sports—I knew something immediately: there's no infrastructure for orders of magnitude more deals in the future. But what struck me harder was realizing there's also no infrastructure for the current volume of deals. That realization led directly to founding Basa.

But here's what people find counterintuitive: after spending seven months immersed in AI research, after building a company predicated on AI's transformative effects on content production, we deliberately don't build AI into our customer-facing platform. Meanwhile, internally, we use AI for practically everything.

When Complexity Resists Systematization

Creator deal flow—the process from initial outreach to signed contract—involves variables that resist encoding in ways that legal tech vendors consistently underestimate. You're not just negotiating commercial terms. You're managing relationship dynamics where someone's personality becomes their business model.

I watched this pattern repeatedly during my decade managing Delta Rae. When talent literally are the product, every business decision carries emotional weight that AI can't yet navigate. A brand wants to work with fifty creators on a trending campaign. Each creator evaluates the opportunity differently based on their audience demographics, their content calendar, their existing brand relationships, their personal comfort with the product category. One creator needs to check with their business partner. Another wants to modify the deliverables to fit their content style. A third is concerned about how the deal affects their positioning with other brands.

These aren't edge cases. This is standard deal flow at volume.

The AI negotiation demos I've seen handle the straightforward scenarios beautifully. But they break down when a creator's parent joins the call halfway through to ask about image rights, or when an agent representing multiple talents wants to restructure deliverables across an entire campaign. The problem isn't that these tools lack sophistication—it's that the negotiation patterns they're trained on don't capture the full complexity of how these transactions actually unfold in practice.

Where AI Actually Works

The distinction we've drawn is between automation and augmentation. We don't automate customer-facing negotiations because the processes aren't sufficiently repeatable yet and the accuracy requirements are absolute. But we use AI extensively internally, where our team can validate outputs and catch errors before they affect customers.

For our lean team, AI isn't optional—it's essential. We use it to scale our abilities, maintain consistency across outputs, and preserve focus by automating routine tasks so our attention stays on high-value activities that require human judgment. Everyone working with Basa is expected to try AI first before scheduling meetings or asking for guidance. This isn't efficiency theater—it's building genuine literacy in where AI excels and where it falls short.

Six months ago, a tool like V0—Vercel's AI design platform—didn't really exist. Today, it's completely changed our operational model. Traditionally, when someone who understands our customers identifies a product improvement, the process is broken and slow. The subject matter expert explains their idea to a product manager. The product manager translates that to a designer. The designer creates mockups. Back-and-forth refinement happens. Eventually it gets handed to developers. Weeks later, you might get the feature you originally envisioned.

But innovation happens bottom-up, not top-down. After months of frustration trying to capture what was already clear in her head through traditional design handoffs, our customer expert took it upon herself to test V0. Now she can turn her understanding of customer pain points directly into functional UI prototypes. Instead of playing telephone through multiple people, she generates working code that communicates her vision exactly to our development team. What used to take weeks of back-and-forth now happens in hours.

This creates what I think of as a testing ground. Before implementing any AI capability in our customer-facing product, we rigorously test similar tools in our own workflows to understand their strengths, limitations, and failure modes. We're accumulating trust calibration—learning through daily use exactly when to rely on AI outputs and when to question them.

The Precision Threshold

What became clear as we built Basa is that companies preparing for AI's second-order effects will capture more advantage than those chasing AI features. It's hard to predict exactly the pace of technological change with AI or how humans will respond to it. But the trajectory seems clear: when AI makes content creation nearly free and deal volume explodes, teams will need infrastructure designed for human judgment at AI-driven scale.

This framing shapes everything about our product development. We're not trying to automate negotiations that require human judgment. We're eliminating the administrative friction that prevents those negotiations from happening at the volume algorithmic distribution demands.

The average creator deal flow requires 60-80 emails at 5-10 minutes per email—that's 5-13 hours of coordination time per deal just managing communication. Scale that to fifty deals to catch a trending moment, and you're looking at 250-650 hours of administrative work. Think about what teams could do with that time instead. The brand manager who spent her week chasing signatures could be developing the next campaign. The talent representative coordinating forty deals simultaneously could be nurturing relationships with rising creators. The agency producer buried in contract revisions could be pitching new partnerships that actually drive revenue.

When you eliminate negative-value administrative work, you create space for the creative and strategic work that builds businesses. That's the infrastructure layer we're building—not AI-powered negotiation, but coordination tools that preserve space for human judgment while removing the busywork that currently buries it.

Building Toward Future Capabilities

Our internal AI literacy means we understand exactly which coordination tasks can eventually be augmented by AI and which will always require human judgment. We're accumulating advantages that can't be replicated quickly—not just technical capabilities, but calibrated understanding of where AI creates value and where it introduces risk. Our team understands AI capabilities and limitations through daily use, not theoretical knowledge.

When AI becomes ready for customer applications—when it can achieve the 100% accuracy that contracts demand—we'll implement it thoughtfully rather than reactively. Right now, AI can't deliver that precision, and creator negotiations involve too many variables that resist systematization. But the infrastructure those negotiations require—the coordination layer that eliminates administrative friction while preserving space for human judgment—that's something we can build today. And it's what teams will need when AI-driven content creation makes deal volume explode.

I sometimes wonder if we'll look back on this decision and see it as overly cautious. But I keep returning to what I learned during those seven months of AI research: the companies that win aren't necessarily the ones that adopt AI first, but the ones that understand its implications most deeply. We're not building AI into Basa today because we understand AI too well to implement it prematurely. That's the foundation everything else gets built on—infrastructure so well-designed for human decision-making at scale that it becomes the natural platform for AI enhancement when the technology catches up to the complexity of the problems we're solving.