Partner at Kleiner Perkins
Check size: No verified personal check range or standard firm-wide range is published. Her official portfolio spans early and growth investments; disclosed company rounds are not personal allocation limits.
Leigh Marie backs AI infrastructure and applications, with a particular eye for the systems that close the gap between frontier-model capability and reliable, mission-critical use. Her portfolio includes tailored enterprise AI, agent workforces, learned memory, runbook automation, AI sales, developer infrastructure, databases, observability, identity, biopharma knowledge work, and financial products. Her Scale AI background gives her direct fluency in data and ML workflows for autonomy and robotics.
Show why the model itself is not enough and what deployment, data, tooling, or product layer you own. Bring production reliability, performance, evaluation, proprietary context, integration, security, and evidence that users trust the product with a mission-critical workload. Highlight talent density and why this team can keep pace with a weekly-changing research frontier.
Small talent-dense teams, world-class technical recruiters, infrastructure that reliably ships frontier capabilities, mission-critical workloads, deep product craft, and founders who pair technical depth with exceptional speed and hustle. Her writing highlights real production use over pilots and talent density that is difficult to manufacture.
Her published work implies caution toward generic one-size-fits-all AI, research demos that never reach production, pilots without mission-critical adoption, products that ignore proprietary company context, fragmented point solutions without workflow ownership, and teams unable to recruit exceptional builders.
Evaluate the deployment infrastructure, proprietary-data integration, evaluation, reliability, and in-house expertise required to turn a frontier model into a mission-critical system.
Treat repeated recruitment of independently world-class builders as evidence of founder quality and ambition, then verify whether that density translates into shipping velocity and production quality.
Test whether the product continuously learns from a customer's processes, data, expertise, and goals so performance and switching cost improve with legitimate use.
Find a high-demand physical service where missed calls, scheduling, or manual coordination directly loses revenue, and let AI own a narrow measurable outcome before expanding.
“For most companies, the models aren’t the bottleneck.”
— https://www.kleinerperkins.com/perspectives/applied-compute-closing-the-gap-between-frontier-ai-and-real-world-impact/
“Talent density is nearly impossible to manufacture.”
— https://www.kleinerperkins.com/perspectives/applied-compute-closing-the-gap-between-frontier-ai-and-real-world-impact/
Enterprise AI is constrained less by model capability than by reliable deployment, proprietary-data integration, and scarce expertise; the durable layer closes that production gap.
AI agents can capture missed demand and automate overloaded office workflows for physical service businesses that remain far less digitized than knowledge work.
AI-native knowledge systems can remove manual regulatory and clinical information bottlenecks that consume large portions of drug-development timelines.