Enterprise AI, Real Talk: Unlocking Next-Gen Workflow Productivity
Recap from EntreConnect – April 18th, 2025, Mountain View, CA
What happens when four powerhouse product minds—from Series E giants to YC rocketships—sit down to talk AI agents? You get fire. And you get the kind of hard-earned insights no one’s tweeting.
At EntreConnect’s latest event, "AI for the Enterprise: Building Next-Gen Tools for Workflow Automation & Productivity", founders and operators broke down how agentic AI is moving from hype to production—across hiring, customer service, infrastructure, and documentation. These weren’t theoretical musings. This was a download from the trenches.
Let’s dig in.
🎙️ The Panel
Jaclyn Zhuang – VP of Product, Eightfold.ai (Series E, $2.1B valuation), formerly Head of Product at Atlassian
Sinduja Ramanujam – Senior Staff Product Manager, ServiceNow
Sarthak Srinivas – Co-founder & CEO, Quantstruct (YC W25), ex-PM at Moveworks (acquired by ServiceNow)
Deepti Srivastava – Founder & CEO, Snow Leopard; Founding PM of Google Spanner
🤖 Let’s Be Real: “Agent” Is a Fuzzy Word
The term agent got politely dismantled at EntreConnect—and rightly so. Everyone on stage had a different reaction to it, but one thread united them: the word has become marketing mush.
Sarthak cut straight through it:
“I don’t actually like the word agent too much. It can mean whatever you want it to mean.”
Instead, he outlined what matters: LLMs with tool-calling, memory, and context. That’s the atomic unit of functionality. Call it an agent, automaton, or workflow—but if it can’t reason, act, and close the loop, it’s just a chatbot in a costume.
Jaclyn agreed:
“Even if you ask ChatGPT or Claude what an agent is, you won’t get a consistent answer. Some people use it for automation. Others call simple chatbots agents. It’s a huge spectrum.”
Sinduja highlighted the confusion in practice:
“In CSM, we’ve always had human agents. Now, when I say 'agent' in a PRD, I get asked: ‘Do you mean a human or an AI?’ That’s how unclear it is.”
And as Deepti put it:
“We’ve built multi-agent systems before the term was cool. They’re just programs that do stuff. The new part is using LLMs as components in those workflows.”
Bottom line: The label doesn’t matter. What matters is this:
✅ Understands input (text, image, sentiment)
✅ Retrieves relevant knowledge
✅ Plans actions
✅ Executes real tools
✅ Closes the loop autonomously
If it doesn’t do that, it’s not an agent. It’s an experiment.
🧱 AI Demos Are Easy. Production Is Brutal.
Everyone agreed: shipping a slick demo is simple. Shipping a reliable agent into production? That’s where the real battle begins.
Deepti didn’t sugarcoat it:
“People build POCs with dummy data. But in production? You need reliability, determinism, and zero hallucination. That’s not solved yet.”
Sarthak put it even more plainly:
“Half my day is spent labeling agent outputs in Google Sheets. That’s the unsexy work that gets you to production.”
He described Quantstruct’s agentic systems as directed graphs where each decision must be checked against business logic. The key is QA, scaffolding, and trust-building—not clever prompts.
Deepti added the enterprise lens:
“No AI system today has five nines of availability. Enterprise apps can’t tolerate models that give different answers every time.”
If your AI tool only works in a sandbox, it’s not a product. It’s just a project.
📉 Retention Is the Real KPI
Here’s the quiet truth in GenAI adoption: getting people to try your AI tool is easy. Getting them to keep using it is hard.
Deepti nailed it:
“Last year, everyone adopted something. But what’s the churn curve?”
“The first thing to go when value isn’t real is the tool that only adds productivity.”
Sinduja shared how ServiceNow tracks whether the AI:
Picked the right tool
Executed correctly
Delivered what a human would
It’s not just usage—it’s observability and outcome accuracy. That’s retention.
Jaclyn tied retention to business survival:
“Companies are keeping employees they were going to lay off—because our agent helps reallocate them internally. That’s not just retention. That’s survival.”
And Sarthak brought it full circle:
“We don’t measure success by usage spikes. We measure it by trust. And that’s built manually, day by day.”
🚀 Startups vs Enterprises — Who Wins AI?
Here’s the real story: in AI, it’s not size that wins. It’s speed, focus, and being willing to bet big when everyone else hesitates.
Startups are built for this moment.
Enterprises... not so much.
Sinduja nailed it:
“Big companies are like cruise ships. Startups are jet skis.”
Enterprises have the obvious advantages:
Deep customer trust 🤝
Huge distribution 🌎
Decades of credibility 🏛️
But they’re weighed down by:
Endless approvals 🧱
Slow innovation cycles 🐢
Cultures allergic to real risk 😬
Meanwhile, startups move fast, experiment aggressively, and jump on the opportunities big companies can’t even see yet.
Deepti called out the real advantage:
“After every platform shift, it’s the new ideas—not the old winners—that dominate.”
The playbook for startups:
Find the gaps giants are ignoring
Move faster than corporate red tape
Ride the AI platform shift to invent new categories
Earn trust by solving real, painful problems
Bottom line: Startups don’t need the biggest brand or the deepest pockets. They need speed, clarity, and hunger.
💥 Tactical Takeaways
This wasn’t a hypefest—it was a reality check. Here’s what real builders had to say:
✅ Start with pain, not novelty
✅ Build native workflows—not bolt-ons
✅ Invest in infra, not just models
✅ Track trust, not just usage
✅ Production is the real milestone—not the launch video
“The AI that survives will create new value—not just speed up the old way.” — Deepti
“No one has 100% product-market fit. Find the 20% they ignore—and own it.” — Jaclyn
🚀 Final Thought: This Is the Platform Shift
Deepti closed it out with a challenge:
“This isn’t about what big companies left behind. It’s about what’s newly possible now.”
That’s the real opportunity. It’s not about retrofitting GenAI into old workflows — it’s about rethinking the systems entirely, and building for the world that AI makes possible, not the one that came before. The founders and builders who will win this next era aren’t the ones patching the past — they’re the ones inventing the future. 🚀
A huge thank you to our incredible speakers — Arthur, Jaclyn, Sinduja, and Deepti — for sharing real insights, hard-won lessons, and bold ideas. And big thanks to everyone who joined us, asked sharp questions, and stayed late to keep the energy high. 🙌
🗓️ Upcoming Events
Follow us on Linkedin and Luma to stay updated on future events!
May 13, Sunnyvale, CA | SaaStr Side Event: The AI Frontier – Scaling Agents & Driving Innovation in SMB and Enterprise with Unicorn Founder @Plug & Play. (🤝 Open for partnership & sponsorship! For more details, contact: team@entreconnect.us to secure your spot!)
Jun 13, San Francisco, CA | How Agents Will Revolutionize the Future of Data @AWS GenAI Loft, Dive into the future of AI Agents in Data — with C-suite leaders from MongoDB, Gen, and beyond! 🚀
July 15, New York, NY | 2025 AI for Greater Good. EntreConnect is thrilled to co-host the AI Hackathon with Gemi.AI, harnessing the transformative power of AI to drive ethical innovation, social impact, and sustainability.
🧳 Career Opportunities
EntreConnect is looking for volunteers! Join us to help shape the future of our founder community, drive meaningful impact, and grow alongside some of the most inspiring entrepreneurs out there:
👉 Check out this LinkedIn post, this and this one for more opportunities from our amazing panelists' companies!