From Context to Execution: How AI Agents Understand, Remember, and Act
An AI agent can reason well and still make the wrong call. Often, the problem isn’t the model. It’s what the model knows when it acts.
A relationship agent needs to remember what happened the last time you met someone. A GTM agent needs to see what changed in the market this week, not last quarter. A workplace agent needs to know how the team actually gets things done. A shopping agent needs product facts and judgment that only the merchant can provide.
Each case depends on context, but the context comes from different sources, changes at different speeds, and carries different risks.
This panel brings together four founders working on those problems from different angles:
Private memory for professional relationships
Live market and audience intelligence
Shared memory for teams and agents
Merchant-authored knowledge for AI commerce
We’ll compare how these systems capture, update, retrieve, and protect context, and ask a practical question:
What does an agent need to know before we trust it to act?
Panelists
🎤 Felix Meng
Founder & CEO, XTrace
Felix is the founder and CEO of XTrace and a Stanford alumnus. XTrace is the shared memory infrastructure for AI-native teams. XTrace captures knowledge from work that people and agents have already completed, making successful procedures, decisions, and artifacts available for reuse.
His lens: How teams and agents can remember what worked, transfer that knowledge, and stop solving the same problems from scratch.
🎤 Jing Kuang
Co-Founder, Cresca | Founding Partner, Y+ Ventures
Jing works at the intersection of consumer AI, professional relationships, and trust. She is currently building Cresca, a private relationship memory layer that helps professionals remember who they meet and follow up with the right context. She is also the Founding Partner of Y+ Ventures and a Stanford GSB alumna. Y+ Ventures is an early-stage venture firm investing in consumer AI startups at the angel, pre-seed, and seed stages.
Her lens: How personal relationship context can be captured and surfaced without giving up privacy or control.
🎤 Joe Tam
Co-Founder & CTO, Upriver
Co-Founder & CTO, Upriver Joe is the co-founder and CTO of Upriver. He previously worked on large-scale multimodal data and consumer products at Instagram, Google, and Spotify.
Upriver gives agents real-time context on emerging trends, brands, and creators. They partner with developers automating GTM and marketing, with clients including Clay and Unify.
His lens: Agents need real-time market and cultural signals to act in a relevant way for target audiences that doesn't feel generic or like slop.
Co-Founder & CEO, Sartorial
Sumit is the co-founder and CEO of Sartorial and a Northwestern alumnus. Sartorial is the brand-truth layer for agentic commerce: it turns a merchant's product catalog into structured reasoning that AI shopping agents can understand, remember, and act on across every AI surface.
His lens: How an AI agent goes from finding a product to choosing it, and what context a brand must supply for that last step to happen.
Moderator
Co-Founder, Sartorial
Stella is the co-founder and GTM lead at Sartorial and a Northwestern Alumna. Previously she built the demand-generation function from zero and ran the Shopify storefront end to end for hardware tech.
Agenda
2:00–2:30 PM — Check-in and networking
2:30–3:40 PM — Moderated panel discussion, Q&A
3:40–3:50 PM — Open mic
3:50–5:00 PM — Networking
Who Should Attend
This conversation is for:
Founders, product leaders, and engineers building agents, copilots, memory systems, retrieval infrastructure, or agent platforms
Enterprise AI, data, and knowledge teams connecting agents to information spread across tools, workflows, and departments
GTM, marketing, and commerce teams exploring how agents can use current market signals and authoritative business knowledge
Investors looking at agent infrastructure, vertical AI, memory, and context engineering
Product, privacy, and governance leaders deciding what agents may remember, share, and act on
It will be especially useful if you’re working through questions like:
What belongs in the prompt, the memory layer, the retrieval system, or the source application? Who gets to change it? When should it be forgotten? And when is it reliable enough to support a real action?
🔥 Limited spots.
About the Host
Sartorial is the agent-native product intelligence helping brands market to AI buyers.
Sartorial helps retailers and D2C brands restructure their old catalogs to be machine-readable and marketing them to AI agents to drive more sales and increase revenue as shoppers increasingly favor agentic commerce over Google search.
Reach out if you're interested in collaborating, sponsoring future events, or just chatting about agentic commerce, startups, and community building!
Message Stella Xiao or email apps@sartori.al 💌
About the Sponsors
XTrace is the shared brain for AI-native teams.
It connects the tools your team already uses — ChatGPT, Claude, Notion, Slack, Drive, Granola — and turns them into one brain with shared intelligence. Every doc, every decision, every conversation, every AI interaction: captured, shareable, and portable across teammates and across agents.
Cresca is a private AI memory layer for professional relationships.
People can capture notes through voice, text, photos, and event interactions, organize them around their contacts, and bring the right context back before a meeting or follow-up. Relationship data remains private by default.
A small, intentional community of founders, investors, and builders, many of us YC and SPC alumni.