A casual commitment made in a Slack thread. A feature request buried three screens deep in last Tuesday's sales call transcript. The bug a new customer mentioned during onboarding that somehow never made it into the engineering backlog.
These are the hairline fractures through which B2B companies leak revenue, credibility, and trust. Requests that slip into the gaps between Slack channels, email threads, and recorded Zoom calls—noticed, perhaps even acknowledged, but never quite actioned.
Jeffrey Byun and Henry Lee have watched this story unfold before. The two founders sold their last venture, OrderAhead, to Square in 2017. They've seen what happens when customer operations buckle under scale. Now they're back with Akkari, building what they describe as an AI execution system for customer operations—an AI agent designed not just to surface customer commitments, but to close the loop on them through automation.
Execution, Not Just Detection
Strip away the pitch deck language and Akkari presents itself as an "AI execution system for customer operations." The premise is direct enough: the platform monitors conversations across Slack, calls, meetings, email, and Discord. It surfaces commitments and blockers. Then—and this is where it diverges from the pack—it acts on them.
That means drafting replies, creating pull requests, updating Linear and GitHub, scheduling meetings, provisioning access. When the fix ships, it pings the customer. The company frames this as an "always-on loop" that learns from outcomes, though what that learning mechanism looks like in practice remains somewhat opaque from the outside.
In theory, a feature request mentioned in passing during a customer Slack thread could spawn a Linear ticket, trigger a GitHub PR, and notify the customer when it's resolved—minimal human intervention required. The product site includes illustrative UI examples (such as mock interfaces showing "ext-replit"), though Akkari hasn't yet published formal case studies or customer testimonials to back the claims.
The platform surfaces primarily through Slack, offers REST API endpoints for managing accounts and commitments, and integrates with development tools including Claude Code and Cursor via MCP and CLI. For now, it's locked behind a private beta accessible only by request.
The Team Behind It
Byun, now CEO, and Lee are Y Combinator alumni—first from the Winter 2011 batch, and again from a more recent cohort. Between OrderAhead and Akkari, the pair co-founded BLADE, a crypto derivatives exchange. (The exact timeline of BLADE's operation and wind-down remains unclear from public records.)
The third co-founder, Michael Moore, was Chief Revenue Officer at OrderAhead, where he was involved in managing teams and operations. The current team is five people, all based in San Francisco, and they're working with YC partner Aaron Epstein.
The company's website lists backers including Tuesday Capital, Orange Collective (a fund focused on YC alumni), Liquid 2 Ventures (the seed firm backed by Joe Montana), and a handful of notable angels: DoorDash CEO Tony Xu, Opendoor founder Eric Wu, and Instacart CEO Fidji Simo among them. Those names appear as logos on the site, though no independent funding announcements have surfaced to confirm the full roster or terms. It's possible the round is still coming together, or that the company is keeping details close.
A Market Getting Crowded Fast
Akkari is entering a space that's heating up in real time. Zendesk unveiled what it calls an "Autonomous Service Workforce" earlier this year, expanding AI agents across messaging, email, voice—even ChatGPT and Gemini integrations. Intercom's Fin agent handles frontline support and now includes "Fin Operator," designed to orchestrate other agents. Decagon, which raised a significant late-stage round recently, offers proactive AI concierges that operate across channels. Then there's Sierra, the enterprise AI agent company, which reportedly raked in hundreds of millions in revenue last year.
The wedge here is narrower. Cursive focuses specifically on agentic customer success, triaging Slack, email, and support tickets. Centive AI positions itself as an "Autonomous Customer Success Platform." Anthropic itself waded into the fray with "Claude Tag," which lets teams summon Claude directly into Slack channels—a move that signals how seriously the foundational model providers are taking this layer of the stack.
What distinguishes Akkari, at least in its positioning, is the emphasis on execution and follow-through—not just detecting commitments or triaging inbound requests, but drafting the response, writing the code, staying on the issue until it's closed. Whether that constitutes a meaningful technical moat or just savvy go-to-market framing is the question early adopters will ultimately answer.
And the early adopters, it's worth noting, haven't spoken yet. No public testimonials, no named reference customers willing to go on record. That's not unusual for a company in private beta, but it does leave the pitch somewhat theoretical for now.
What Comes Next
Akkari isn't disclosing pricing details, and the product remains gated behind a waitlist. For founders drowning in customer Slack channels—watching requests pile up faster than they can triage them—the promise of an agent that doesn't just flag problems but actually ships fixes carries obvious appeal.
Byun and Lee have been through the scaling gauntlet before. They know what customer operations look like when they start to fracture under load, and what it takes to keep them running on something close to autopilot. Whether Akkari can deliver on that promise, and whether it can carve out defensible territory in an increasingly crowded field, will depend on execution.
Which, fittingly enough, is exactly what they're selling.
The company can be reached at [email protected] for those interested in joining the beta.
