Zeta Global
From AI adoption to operating leverage: How Zeta Global scaled engineering with Span
~30%
reduction in pull request cycle time over the year
20%
improvement in Engineering productivity
2X
higher throughput of consistent AI users vs. non-AI users
"Span gives us the ground truth behind our AI transformation. We can finally measure where AI is helping us move faster, think smarter, and deliver bolder."
Patrick D’Souza
Senior Director of AI Enablement
Background
Zeta Global is an AI-native marketing cloud helping enterprises deliver intelligent, personalized customer experiences at scale. As AI became central to both its product strategy and internal engineering workflows, Zeta faced a familiar challenge. Early AI success was real, but uneven.
Some engineers were moving faster with AI. Others hesitated, unsure where AI fit, how much to trust it, or whether it would introduce risk. Leadership lacked a reliable way to separate signal from noise.
Zeta partnered with Span to turn AI from an individual advantage into an organizational capability that could be governed, measured, and scaled with confidence. The result was a shift from anecdotal acceleration to a disciplined, compounding engineering flywheel where velocity, quality, and reliability improved together.
What Zeta encountered was a classic AI adoption chasm. Early adopters were moving quickly and accepting risk, while the broader engineering organization required AI to be reliable, governed, and embedded into existing workflows. Bridging this gap between individual experimentation and organizational trust became the central challenge of Zeta’s AI transformation.
Challenge
AI Adoption Without Visibility
By the start of its AI enablement journey, Zeta engineers were already experimenting with modern AI coding tools. However, leadership lacked answers to fundamental questions:
Where was AI actually helping teams move faster?
Were velocity gains coming at the expense of quality or stability?
Which workflows benefited most from AI, and which needed better structure?
How could AI investment decisions be made with confidence rather than intuition?
Surveys and anecdotes provided directional insight, but they could not support executive-level decisions. As Zeta set an ambitious mandate to materially improve engineering productivity, the absence of a system of record became a strategic constraint.
AI was accelerating individuals, but not yet the organization.
The Insight: Crossing the AI Chasm Requires Measurement
Zeta recognized that scaling AI was not primarily a tooling problem. It was an operating model problem.
Early adopters were willing to experiment, tolerate friction, and self-correct. Most teams were not. They needed AI to work inside established workflows, with guardrails, clarity, and shared standards.
Crossing this AI chasm required two things. Enablement, so engineers could build confidence using AI in real production work. Measurement, so leadership could see where AI created durable leverage and where it introduced risk.
Rather than accelerating AI usage indiscriminately, Zeta chose discipline over speed. Leadership recognized that without trusted baselines for flow, quality, and reliability, scaling AI would amplify noise rather than impact. Measurement had to come before acceleration.
Solution
Span as the Engineering Intelligence Layer
Zeta’s goal was not to optimize individual AI tools, but to evolve its engineering operating model. That required a system that could connect AI adoption directly to execution outcomes across the organization.
Zeta selected Span to serve as the system of record for engineering execution and AI impact.
Rather than starting with AI metrics alone, Zeta anchored measurement in core engineering health signals such as flow, quality, and operational stability, then layered AI adoption and usage on top. This sequencing ensured trust in the data before scaling AI-native workflows further.
With Span, Zeta gained the ability to correlate AI usage with real delivery outcomes across teams, identify where rework, review latency, or workflow friction constrained throughput, distinguish healthy acceleration from risk-inducing speed, create a shared executive-ready view of engineering health and progress, and move from opinion-led debates to evidence-based decisions.
As Patrick D’Souza, Senior Director of AI Enablement at Zeta Global, explains:
“Span gives us the ground truth behind our AI transformation. We can finally understand where AI is helping us move faster, think smarter, and deliver bolder, without guessing.”
Results
From Experimentation to a Compounding Flywheel
With visibility in place, Zeta shifted from ad-hoc experimentation to a coordinated, system-level approach.
Enablement drove adoption, measurement drove focus, and together they created a compounding flywheel where velocity, quality, and reliability reinforced one another rather than trading off.
Enablement Meets Evidence
AI enablement focused on real delivery work, including live coding sessions, peer-led coaching, and workflow-embedded guidance, rather than abstract training. Span’s insights allowed these efforts to be targeted where friction actually existed, increasing impact without broad mandates.
Quality and Velocity Improved Together
Span revealed that many delays stemmed from rework and review cycles rather than implementation speed. This insight fundamentally changed where Zeta focused its efforts. Rather than optimizing for raw coding speed, teams shifted quality and standards enforcement earlier in the lifecycle, where defects are cheaper to fix and flow improves systemically. The result was reduced friction without lowering standards.
Leadership Confidence Increased
For the first time, executives could see how AI adoption translated into execution outcomes. This clarity unlocked trust. For the first time, leadership could confidently connect AI investment decisions to real execution outcomes, enabling informed prioritization rather than intuition-driven debate. Teams were empowered to go further with AI, knowing that performance and risk were visible and manageable.
Bharat Goyal, EVP and Head of Engineering at Zeta Global, notes:
“To scale an AI-native engineering organization, you need clarity on how teams actually work and where leverage exists. Span makes that clarity possible.”
Organizational Alignment at Scale
Span is now embedded in Zeta’s internal engineering operating rhythm. It supports cross-functional alignment between engineering, product, and leadership by providing a shared view of engineering flow and delivery health, AI adoption maturity across teams, areas where further automation will produce the highest return, and how execution improvements tie back to business outcomes.
Rather than being a reporting tool, Span functions as a decision engine that guides where to invest, where to standardize, and where to slow down to protect long-term reliability.
Looking Ahead
From Adoption to Operating Leverage
By the end of this phase, success was no longer measured by AI usage alone. The focus shifted to operating leverage. Faster delivery of customer value. Higher quality built in by default. More predictable operations. Reduced cognitive load on senior engineers. Span provided the connective tissue that made this shift measurable and sustainable.
Zeta’s AI journey is ongoing. With measurement and trust established, the organization is now scaling agentic workflows across code review, testing, diagnostics, and operations, while keeping humans firmly in control.
As AI becomes infrastructure rather than novelty, Span ensures that progress remains visible, governable, and compounding.
Patrick summarizes the shift:
“AI stopped being something a few people were good at. It became something the organization could rely on. Span made that transition possible.”