Pricing isn’t just a sales lever — it’s product strategy.
In this session from our Monetizing AI summit, Daniel Chalef, founder of Zep AI, joins Schematic to unpack what happens when you treat pricing as a living part of your product architecture (not a one-time decision).
Daniel shares hard-won lessons from building and monetizing Zep, an open-source context engineering platform powering AI agents. From free tiers that anchored the wrong expectations to metered billing that unlocked enterprise adoption, he walks through how Zep evolved its pricing and packaging model to balance PLG growth with enterprise scale.
You’ll learn:
Why limits are the connective tissue between PLG, usage-based, and enterprise pricing
How metered billing avoids anchoring (and when it doesn’t)
How “Time to Wow” became a north star for improving activation and conversion
Why hybrid pricing models might be the next frontier for AI infrastructure startups
How Schematic helped Zep iterate pricing architecture fast — running experiments, migrating plans, and building trust with enterprise buyers
If you care about pricing agility, packaging strategy, or the messy middle between product-led and enterprise sales… this one’s for you.
Pricing is not just a sales tool. It is product strategy. Pricing and packaging are never final — they are iterative. Packaging is as much about product structure and go-to-market as it is about designing adoption paths.
Daniel: I’m the founder of Zep, an open-source context engineering platform. We help AI agents get the right context at the right time so they’re accurate and personalized. My path: engineer → founder → product marketing and ML engineering → corp dev evaluating companies and pricing models. That mix shaped how we built and monetized Zep.
Goal from day one: be an enterprise software company solving mission-critical problems. Reality at day one: PLG motion with sub-$500 plans.
Free plan with generous limits to reduce friction and “land grab” devs. Outcome: lots of low-quality signups.
Prepaid monthly plans ($99 / $275 / $499) with included credits. Outcome: anchored enterprise buyers at the wrong price point.
“I see $499 on your site — are you really an enterprise product?”
Hard to bridge $0 → $499/mo → $50K+/yr.
Lesson: Published low prices create positioning and negotiation headwinds.
Reduced free limits to focus on experimentation over “free value.”
Added metered billing (familiar to AI buyers). Benefits:
Avoids anchoring. Buyers understand that 1M units will be expensive.
Self-limiting at scale. Enterprises expect volume pricing and will graduate to contracts.
Structure enterprise with a contracted usage amount (monthly on annual, or annual).
Use a soft limit: do not hard-stop usage; notify and renegotiate when they materially exceed the commit.
Overages become conversation triggers to trade: lower overage rate ↔ longer term, higher commit, etc.
Result: Limits create natural checkpoints to expand revenue and deliver more value.
Limits connect the whole journey:
Free: functional limits encourage activation without giving away the product.
Metered: price naturally scales; signals when to move up.
Enterprise: contracted limits + overage terms structure upsell and predictability.
Customer research found perceived pricing issues:
Metric confusion. We priced on megabytes. Devs don’t think in MB of text. It felt opaque and expensive.
Considering a shift to token-based pricing or a blended model (base fee + usage) for clarity and to raise the floor.
Metered drove the low end to $2–$3/month vs previous $99+.
Exploring hybrid: small base fee to lift floor + usage for scale.
Zep became a sophisticated enterprise product, which slowed time to value.
Launched “Time to Wow” initiative: simplify first-run, better defaults, visual feedback (knowledge graphs, walkthroughs), and reduce effort before code is written.
Faster activation improves conversion and makes pricing feel fair.
Unpaid metered bills (e.g., $200–$300) and card failures created loss.
Solution: threshold billing (charge at $10–$20), stronger dunning, earlier failsafes. Implemented with Schematic.
For metered: cost-of-service analysis + comparables in adjacent ICPs; limited survey value with developer users.
For enterprise: comparables and negotiated value; iterate frequently rather than “set and forget.”
Rapidly experiment, sunset, and migrate thousands of customers between plans.
Grandfathering windows, promotional grants, metered thresholds — without heavy engineering.
Pricing is product. Treat it as an evolving part of architecture and GTM.
Avoid anchoring. Be careful with public low prices that fight your enterprise story.
Use limits intentionally. They guide PLG → metered → enterprise and create expansion moments.
Prioritize clarity. Choose meters buyers actually reason about; expose usage and costs plainly.
Mind the floor. Pure metered can collapse low-end ARPU — consider a hybrid base + usage.
Optimize activation. Time to Wow drives conversion and supports your pricing narrative.
Daniel: Thanks all. Happy to take questions.