Three former Gap Inc. executives who spent years deploying generative AI inside the retail giant have emerged with Marker, a startup that pairs software with embedded consulting teams to build custom AI agents for enterprises.
The Y Combinator-backed company announced its platform around August 20, 2026, positioning itself around a premise born from hard-won experience: that most companies don't need another dashboard to monitor AI experiments. They need people who understand how work actually gets done, combined with infrastructure that can turn those insights into production-ready automation.
David Pastewka, Will Drevno, and Richard Berwick landed at Gap through an acquisition rather than by choice. The retailer bought their previous venture, Drapr (also from Y Combinator), back in 2021. What followed was a three-year education in the messy reality of enterprise AI adoption. The trio eventually led Gap's generative AI program, navigating projects across product development, merchandising, and supply chain operations.
"We've sat on every side of this: founders selling into enterprises, operators transforming one from the inside, and executives responsible for enterprise AI," they wrote when introducing Marker to the Y Combinator community. "Marker is the company we wish we could have hired at Gap."
That statement carries more weight than the usual startup origin story. The founders aren't theorizing about what enterprises need; they've been the buyers, watched initiatives stall, and felt the friction between vendor promises and operational realities.
Consultants Who Code, Software That Adapts
Marker's approach splits the difference between traditional consulting engagements and pure software plays. Teams embed directly within client organizations to map existing workflows, then build custom agents on what the company calls its Command Center platform. The agents integrate with whatever tools employees already use, whether spreadsheets, reporting systems, or task queues.
"They work the way you already work. No new software to learn," the company's website promises.
It's a pitch calibrated for organizations wary of yet another tool that requires retraining staff or overhauling established processes. The model targets mid-size companies through Fortune 500 clients, though Marker hasn't disclosed pricing structures or minimum engagement terms.
As of August 28, 2026, the startup listed SOC 2 Type II and ISO 27001 certifications as in progress but hadn't provided completion timelines, a detail that larger enterprises typically scrutinize before contracts move forward.
Early Deployments Show Promise, With Caveats
Marker cites results from two unnamed clients on its website. A Fortune 500 retailer reportedly cut manual work in product operations by 80 percent and plans to expand the agents into assortment planning and buying functions. A second client in the property technology sector compressed onboarding from 120 days to 60 and expects to eliminate 95 percent of manual quality assurance tasks, though no specific date or customer name was disclosed.
The company hasn't disclosed when those deployments occurred or identified the customers, leaving questions about how mature the implementations are. Early wins often look different six months in, particularly when AI systems encounter edge cases or need to scale across departments.
Still, the numbers align with what enterprises increasingly expect from automation initiatives. Incremental improvements have given way to demands for substantial efficiency gains, particularly in operations that involve heavy manual processing.
A Crowded Field, But Different Angles
Marker enters an agent platform market that's become almost comically congested over the past year. OpenAI rolled out Workspace Agents for enterprise customers. Meta launched its Business Agent Platform in June, focused on customer interactions across WhatsApp, Messenger, and Instagram. Wix introduced Symphony for small businesses in mid-August, billing it as an "AI workforce."
Other players include Kortix with its command center approach, Outreach targeting revenue teams, and Gravitee building agent management tools. Meanwhile, the Agentic AI Foundation brought the A2A protocol for agent-to-agent communication under its oversight in August alongside the Model Context Protocol, a move that signals the ecosystem is maturing toward standardization.
What distinguishes Marker, at least in theory, is the consulting layer. Most platforms sell software and expect customers to figure out implementation. The founders argue their time inside Gap taught them that deployment expertise matters as much as the technology itself, maybe more.
Whether that thesis holds depends partly on whether Marker can maintain quality as it scales the human side of the business. Consulting models work beautifully until you run out of consultants who actually understand the domain.
All three founders studied at UC Berkeley, with Pastewka focused on business and computer science, Drevno in engineering, and Berwick in business. As of August 2026, the company operated with a team of three, according to its Y Combinator profile, and hadn't announced a priced funding round.
For now, Marker offers a 30-minute initial consultation through its website. The pitch is straightforward: if you're trying to deploy AI agents and don't want to waste six months figuring out why they keep breaking in production, here's a team that's already made those mistakes somewhere else.
