Most SaaS companies still run like static machines. Markets do not. Buyers, data, and decision cycles shift weekly. AI compounds that gap: it’s easy to spin up a model, but hard to turn it into a repeatable business impact.
Dael Williamson argues that to thrive in this new era, SaaS leaders must treat their company like a living system. That means thinking in agents, networks, evaluation loops, and UX—not just models or algorithms.
His message is simple: models are commodities, systems are moats.
Dael Williamson is EMEA CTO at Databricks, where he helps large organizations scale data and AI responsibly. With over 24 years of experience in architecture, analytics, and digital strategy, he has guided countless enterprises through data-driven transformations.
Databricks now operates at around $4 billion ARR, with 140% net revenue retention and sustained 50%+ annual growth—proof that AI-first systems can scale profitably.
In his SaaSiest 2025 keynote, Dael shared what happened when Databricks turned the microscope inward—using AI on Databricks itself to explore how AI, networks, and people truly interact.
The core shift: from models to systems
AI’s next competitive edge won’t come from who builds the smartest model, but from who designs the smartest system around it.
Dael outlined four principles that define that shift:
1. Agents that act like teammates
Soon, every employee will have their own fleet of AI agents. These agents won’t work through hierarchies—they’ll work through networks, communicating dynamically to complete tasks and make decisions faster. The challenge for leaders is building the connective tissue that lets them collaborate safely and productively.
2. Data that reflects reality
Machine data is clean. Human data is messy. Sales data is chaos. Without structure, agents amplify the noise. With governance and lineage built into workflows, they accelerate real outcomes.
3. Eval-driven development
Agents behave differently each time you ask a question. You need automated evals—accuracy, latency, stability, and safety checks—to ensure reliable outputs. Think of it as test-driven development for AI.
4. UX that replaces meetings, not people
AI adoption fails when interfaces add friction. It succeeds when they save time. Databricks replaced status meetings with AI interfaces that summarize context, show what’s changed, and highlight what decisions are needed.
As Dael put it: “AI isn’t replacing people—it’s replacing meetings.”
What Databricks learned by experimenting on itself
Over the past two years, Databricks has been its own lab for AI transformation. Here’s what Dael’s team discovered:
- AI interfaces saved time, not jobs. Teams using AI copilots cut meeting load and improve focus.
- The edge outperformed the center. A UK-built internal tool—nicknamed “Alice”—went viral across 5,000 employees without any promotion. Centralized dashboards lagged behind.
- Finance adopted AI fast. The field did not. Real-time revenue analytics worked beautifully, but only 25% of the company used AI weekly. Even AI-first companies face adoption barriers without the right UX and incentives.
- Networks scale better than hierarchies. Inspired by companies like ASML, Databricks found that distributed ownership and peer-to-peer design fueled faster innovation than top-down coordination.
Three experiments every SaaS company can replicate
1. Supply chain stress testing
Three people recreated a bank-grade stress-testing system in 20 hours, using MIT’s resilience framework. Then they built an AI agent to explain the results in plain language—28 hours total to create a working prototype that once took banks years.
2. Sales system diagnostics
Dael exported time-series Salesforce data, built an agent to run experiments, and used PySpark to analyze bottlenecks. The result: clear insights into pipeline delays, queueing issues, and revenue acceleration opportunities—all built in a weekend by someone who hadn’t coded in 10 years.
3. Agent labs as the new R&D
Instead of big one-off AI projects, Databricks now runs “Agent Labs”—small teams that ship weekly, measure adoption, and improve based on usage data. It’s the AI version of continuous deployment.
Your moat isn’t your model. It’s the network of workflows, data, and behaviors that agents learn from.
The 5-step framework for SaaS leaders
1. Map your work as networks, not org charts
Identify critical flows—discovery, decision, delivery, learning—and locate bottlenecks. These are agent opportunities.
2. Launch a small Agent Lab
Pick 2–3 use cases with real impact (pipeline triage, customer ops, or forecasting). Staff lean: one product owner, one data engineer, one designer, one SME. Ship weekly.
3. Build an eval harness
Define success for every task: accuracy, latency, safety. Track eval pass rates just like uptime or conversion.
4. Embed AI in daily tools
Integrate where your team already works—no new logins, no friction. Replace meeting decks with AI interfaces that deliver the brief, the diff, and the decision.
5. Reward usage, not experiments
Add “agent-assisted outcomes” to your dashboards. Measure how AI improves cycle time, accuracy, and team capacity—not how many pilots exist.
Metrics that actually prove adoption
- Agent Adoption Rate: active users / eligible users
- Cycle Time Delta: before vs after automation
- Agent Assist %: tasks completed with AI help
- Eval Pass Rate: stability across versions and prompts
- Decision Quality: win rate or SLA improvement
What to avoid
- Building everything centrally—scale what the field already loves.
- Assuming your model is your moat—your system is.
- Skipping evals—leaders won’t trust untested outputs.
- Letting hierarchy slow agent networks—agents need distributed ownership to thrive.
The takeaway
AI will not reinvent your company on its own. It will amplify whatever system it enters.
If that system is rigid and hierarchical, you’ll get expensive demos.
If it’s adaptive, evaluated, and networked, you’ll get compounding advantage.
Start small. Measure relentlessly. Treat your company like a living system—because in the age of agents, only living systems grow.
Watch the full session from SaaSiest Amsterdam 2025 here: https://saasiest.com/the-role-of-ai-in-changing-company-structures-and-dynamics/
