Convey, the enterprise AI platform that builds "digital teammates" to automate operational workflows, announced a $38 million Series A led by Andreessen Horowitz (a16z) with participation from Khosla Ventures and Pear VC. Joe Schmidt from a16z joins the board. The company disclosed that its AI teammates have already completed 1.1 million hours of real work (not demos or pilots) across customers including NBCUniversal, TelevisaUnivision, Unity, Samsara, ChargePoint, and Faire. One customer saved 450+ hours per week. Another—Savoya—achieved a 40% year-over-year EBITDA lift and is on track to save 10,000 hours this year.
Founded by Rohan Chopra (ex-DoorDash early engineer), Will Harvey, and Diego Canales, Convey positions itself as "teammates, not agents"—software that owns outcomes, not just tasks. Non-technical operators onboard Convey by screen-sharing and walking through a process (similar to onboarding a human employee). The platform then builds versioned, testable programs (not stateless prompt loops) that run reliably across hundreds of thousands of executions.
What Makes This Different from Agent Hype
Most AI agents run fresh prompt loops every time and "wing it" when something unexpected happens. You don't run business-critical processes on a vibe. Convey addresses this by compiling what it learns into real, versioned, testable programs that behave consistently. Each teammate has its own identity, permissions, and lives inside the customer's security perimeter—no workarounds, no shadow IT.
The ROI gap is closing. Global AI spend is projected to hit $2.5 trillion in 2026, but returns remain elusive in areas where most operational tasks occur (accounting, ops, campaign trafficking, invoice reconciliation). Convey's thesis: the best AI use cases are "rote and operational" with an objective right answer. Human workers keep strategy and customer relationships; AI takes the manual data entry that nobody wanted to do anyway.
a16z partner Joe Schmidt framed the opportunity: "This is going to be a way that every company across the economy can grow faster. The humans inside those companies can be doing the work they actually want to do." Schmidt cited two investment drivers: the founding team's pedigree and the massive addressable market of manual work waiting to be automated.
The "Steve" Problem and How Convey Solves It
The founding story traces back to DoorDash in its earliest days. According to a16z's announcement, the entire delivery network initially ran on "one guy named Steve" who manually texted drivers for every order. No algorithms, no dispatch system—just Steve matching orders to dashers on his iPhone. Rohan Chopra, an early DoorDash engineer, built the software that automated Steve's job, freeing him to solve bigger problems.
Chopra saw the same pattern everywhere: a smart person in accounting, ops, or account management "holding a critical process together by hand because the tools never caught up." These are your best people spending 2-3 hours daily clicking buttons, pulling reports, reconciling invoices, trafficking campaigns. It keeps the business running, but it's repetitive work and a poor use of talent.
Most companies don't have a Rohan sitting around to automate their Steves. That's why Chopra started Convey—to let non-technical teams build and run digital teammates without engineering help. You share your screen, walk through the process, answer a few questions. After that, Convey's AI owns the outcome, running inside existing systems and escalating to the right person when stuck.
What "100x Operators" Actually Means
Convey introduces the concept of "100x operators": someone who hands off grunt work and spends their time on things only they can do. This isn't about replacing workers (though the timing is sensitive—Snap, Block, and Wix recently cited AI as a factor in layoffs). Business Insider reports that Chopra found employees closest to the work are often most excited because they're handing off tasks they never wanted to do.
Example: A large restaurant chain employee who spent hours at the end of every day on manual data entry. "That is not how she wanted to spend her evenings," Chopra said. Convey freed her to focus on higher-value work.
The best use cases are tasks that are "rote and operational" with an objective right answer. Workers remain better at anything involving strategy, judgment, or customer relationships. "That's the piece that remains human well into the future," Chopra said.
Real Production Metrics vs. Pilot Theater
The distinction Convey emphasizes: 1.1 million hours of real work, not demos. Most enterprise AI announcements cite pilot programs or proof-of-concept wins. Convey's customers are running these teammates in production, at scale, on business-critical workflows.
Customer results disclosed:
- Large streaming service: Handed off reporting and ad ops workflows, recovered 450+ hours per week
- Savoya: 40% EBITDA lift year-over-year, on track to save 10,000 hours in 2026
- NBCUniversal, TelevisaUnivision, Unity, Samsara, ChargePoint, Faire: All named as production customers (no metrics disclosed yet)
What this signals to CFOs: The labor-savings ROI is measurable and material. At 450 hours/week (roughly 10 FTEs at 45 hours/week), even at conservative $50/hour loaded cost, that's $1.17M saved annually from one workflow handoff. Savoya's 40% EBITDA lift suggests operational AI isn't just cost reduction—it's margin expansion.
Why Convey Isn't Worried About OpenAI or Anthropic
Chopra dismissed platform competition by comparing it to DoorDash's early days. Skeptics asked what would happen if Uber competed directly. His answer: broad platforms don't always win every vertical. "The biggest advantage of Convey is focus," he said.
This mirrors enterprise software history: Salesforce built a CRM empire despite Microsoft Dynamics existing. Workday built HR/finance despite Oracle and SAP dominating enterprise. Snowflake built a data warehouse despite AWS, Google, and Azure all offering alternatives. Specialized tools win when the general-purpose platform lacks domain depth.
For Convey, the domain is operational workflows. Non-technical operators don't want to prompt-engineer or build agent workflows in LangChain. They want to onboard a teammate, hand off a process, and never think about it again. That's a product problem, not just a model problem.
What Enterprise Leaders Should Do
For COOs: Identify Your "Steves"
Run an audit of where smart people waste time on repetitive work. Look for:
- Manual reporting (pulling data from 5 systems into Excel every Monday)
- Invoice reconciliation (matching POs to receipts across multiple vendors)
- Campaign trafficking (setting up the same 50 ad placements for every launch)
- Data entry (transcribing forms, updating CRMs, syncing systems that should integrate but don't)
Prioritize by hours spent and talent wasted. If a $150K/year ops manager spends 10 hours/week on manual reconciliation, that's $37,500/year in opportunity cost. If you can automate it for $10K/year, the ROI is immediate.
For CTOs: Evaluate Security and Identity Models
Convey's positioning is smart: each teammate gets its own identity and permissions inside your existing security perimeter. This avoids the "admin account with full access" anti-pattern that killed early RPA deployments.
Questions to ask:
- How does the AI authenticate to our systems? (SAML/SSO, service accounts, delegated permissions?)
- What happens when an AI teammate encounters a new scenario? (Does it escalate to a human or make a guess?)
- Can we version-control and test AI workflows like we do software? (Convey says yes; verify this in a POC)
- What audit trail exists for every AI action? (Compliance needs this)
The versioned programs claim is significant. If true, it means you can treat AI workflows like production code—tested, reviewed, deployed with rollback capability. That's fundamentally different from prompt-based agents that improvise every time.
For CFOs: Model Labor Savings vs. Platform Costs
The $38M round at undisclosed valuation suggests pricing won't be trivial. Enterprise AI platforms typically charge per-seat, per-task, or consumption-based. Convey hasn't disclosed pricing publicly.
ROI modeling framework:
- Baseline labor cost: Hours spent on manual work × loaded hourly rate
- Automation cost: Platform fees + integration/setup + ongoing maintenance
- Net savings: (1) minus (2)
- Payback period: Automation cost ÷ monthly net savings
Example (Savoya's 10,000 hours/year saved):
- Baseline: 10,000 hrs × $50/hr = $500K/year labor cost
- Automation cost: Assume $100K platform + $50K setup = $150K first year, $100K recurring
- Net savings: $500K - $100K = $400K/year (after year 1)
- Payback: 3.75 months
At 40% EBITDA lift, the calculus is even better. If Savoya's baseline EBITDA was $1M, a 40% lift is $400K. Even if half of that came from other initiatives, the AI contribution is $200K—well above platform costs.
For CIOs: Watch the Competitive Landscape
Convey is not alone in operational AI. Competitors include:
- Zapier Central / AI Actions: Workflow automation adding AI
- UiPath / Automation Anywhere: RPA vendors pivoting to AI
- Salesforce Agentforce, Microsoft Copilot: Incumbents embedding agents
- Sierra, Regal, Decagon: Vertical AI agents (customer service, sales)
What differentiates Convey:
- No-code onboarding for non-technical users (vs. developer-first platforms)
- Versioned, testable programs (vs. stateless prompt loops)
- Outcome ownership (vs. task completion)
- Focus on operational workflows (vs. general-purpose agents)
Strategic question: Do we bet on a platform like Convey, or do we wait for Salesforce/Microsoft to build this into their existing stacks? The risk of waiting: competitors adopt faster and capture 18-24 months of margin expansion before you catch up.
The $2.5 Trillion AI Spend Gap
The a16z announcement cites a critical stat: Global AI spend will reach $2.5 trillion in 2026, yet returns remain lacking "in the areas where most operational tasks occur." This echoes the 2025-2026 enterprise AI narrative: massive investment, limited production ROI.
Convey's bet: The ROI gap exists because most AI tools are contextless point solutions that companies must adapt to use, rather than AI adapting to the company. Convey onboards like a human (screen share, walkthrough, Q&A) and then runs inside existing systems with existing permissions.
If this model works at scale, it validates the "AI coworker" category. Not copilots (suggestions you act on), not agents (tasks you delegate), but teammates (outcomes you hand off). The semantic distinction matters: a coworker owns a job function end-to-end, escalates when stuck, and improves over time.
What Happens Next
Convey will use the $38M to expand into more enterprise markets and accelerate product development. With a16z's enterprise network and Khosla's enterprise SaaS expertise, expect aggressive go-to-market in 2026-2027.
The AI employment debate intensifies. Snap, Block, and Wix citing AI as layoff factors creates political and PR risk. Convey's positioning—"freeing employees to do work they actually want to do"—is the optimistic framing. The pessimistic framing: 1.1 million hours automated = ~500 FTE jobs eliminated (at 2,080 hours/year per FTE).
The truth is somewhere in between. Some roles will shrink. Others will expand (managing AI teammates, handling escalations, solving the problems AI can't). The companies that figure out how to redeploy freed-up talent toward growth (not just cost reduction) will capture the 40% EBITDA upside Savoya demonstrated.
Expect competition to heat up. Salesforce acquired Fin for $3.6 billion this week. If operational AI becomes table stakes, incumbents will buy or build aggressively. Convey's window to dominate before Microsoft embeds this into Dynamics 365 or Salesforce ships Agentforce 2.0 is 12-24 months.
