The 90-Day Lock-In Nobody Saw Coming
Wordware stumbled onto something fascinating with Sauna: users don't leave after 90 days. Not because they can't—there's no contract, no data hostage situation—but because the AI knows too much about how they actually work.
This isn't your typical feature-based moat. It's context-based retention. After three months, Sauna knows which meetings you'll actually attend, how you handle edge cases in support tickets, what "urgent" really means in your workflow. Users report they tried switching back to generic AI tools and just... couldn't. The loss of accumulated context felt like starting over with a new assistant who doesn't know your preferences yet.
That's a fundamentally different competitive position. As AI models commoditize, the richest user context becomes the differentiator. Wordware isn't winning because their underlying model is better—they're winning because their product remembers.
The pivot from Wordware (workflow builder) to Sauna (AI companion) tells you everything about what users actually want. People built sophisticated automations with the workflow tool, then drifted away. Not because the automations didn't work, but because building and maintaining them created friction. Users didn't want to learn triggers and actions—they wanted work done. The lesson: configuration is a retention risk. Every feature that requires users to set up, train, or maintain the system increases the chance they'll abandon before hitting that 90-day threshold where the magic happens.
The Volume Problem
Here's where it gets interesting. The agent learns your patterns beautifully for one-off tasks. But knowledge workers don't process things one at a time. They get 20 similar CVs to review, 30 support tickets with the same underlying issue, 15 emails that need the same style of response.
Right now, users teach the agent once, then manually apply that learning 20 times. That's the gap. The intelligence is there—the agent knows how you handle these—but you're still doing repetitive execution. This is where users hit a volume wall before they experience enough compounding value to stay.
The fix isn't complicated in concept: batch processing with pattern learning. Show me how you'd handle all 20 items based on what you've learned about my preferences, let me approve in bulk or flag exceptions. Human-in-the-loop at scale. Without this, you risk losing users who need the automation most—the ones drowning in high-volume, low-variability work.
The Trust Ceiling
Wordware raised the largest seed round in Y Combinator history ($30M), with backing from Paul Graham, Azeem Azhar, and Spark Capital. That's serious institutional validation. The pitch is compelling: give Sauna access to your full digital life, and it becomes indispensable.
But there's a trust ceiling forming. The product requires deep access—email, calendar, documents, communication patterns. Yet the security documentation remains vague: "variety of security measures" without specifying encryption standards, audit logs, or compliance certifications. For a product whose entire value proposition is becoming deeply embedded in your workflow, that's a gap.
This isn't about being insecure—it's about being visibly secure. Enterprise buyers and security-conscious users (exactly the people who'd benefit most from context-driven automation) need specifics: SOC 2 status, encryption standards, data residency, audit access, penetration testing cadence. The cost of not publishing this isn't security risk—it's capping your addressable market to early adopters while competitors with clearer postures take enterprise accounts.
The core insight is solid: context-driven retention is a real moat, and users genuinely can't return to generic AI after experiencing personalized intelligence. But scaling that requires solving for volume workflows and earning trust through transparency, not just assurances. We used Mimir to pull this analysis together, and the patterns are clear—Wordware has found something valuable. Now it's about removing the friction that keeps users from reaching that 90-day inflection point where they're locked in by choice, not circumstance.