loading...
Relay extends beyond Zapier-style data routing to embed AI generation and human approval steps inside the same automation sequence – bridging the gap between moving data and running real processes.
ENTRY ANGLES
Vertical-specific workflow platforms (e.g., e-commerce operations, legal intake, professional services client onboarding) · Dual-mode workflow systems distinguishing routine processes from exception handling with different escalation and notification logic · AI operations with human approval gates as first-class platform elements
VERTICALS
CAPABILITIES
AI operations and workflow automation, Deep vertical domain expertise, Exception handling and escalation logic design
Zapier is widely used, but its architecture has a ceiling – it's good at moving data between apps, less good at building processes that involve human decisions or AI generation along the way. Relay is building in the space above that ceiling.
The platform is a workflow automation tool with two features that its founder argues distinguish it from the crowded Zapier-adjacent market. The first is native AI operations that can be dropped into any automation sequence. The second is the ability to insert human decision points into those same sequences.
The first feature works like this: Relay can be instructed to summarize incoming customer emails in a single sentence and push that summary into a CRM activity log – automatically, every time. Or to draft a response to the email and send it without human intervention. These are AI steps embedded in a procedural flow, not one-off prompts.
The second feature is what makes this genuinely different. Rather than sending that AI-drafted response directly, the next step can route it to a human reviewer who approves or edits before the email goes out. This isn't just about AI oversight – most real business processes already include human approval gates. A new client added to the CRM might trigger an AI-generated presentation draft, which goes to the sales rep for editing, then to management for sign-off, and only then is sent to the client. Relay can model all of that as a single executable workflow.
The app review response scenario illustrates the depth: a new review appears, gets published to an internal Slack channel and logged in a database, triggers an AI-generated reply draft routed to the product team, passes through customer support for tone review, and after final approval is automatically posted to the app store. Each step is connected; nothing falls through the cracks.
Pricing starts at a free tier, with the standard configuration at $9 per user per month. More advanced tiers with custom modules and branded interfaces are available on request. Relay has raised $3.1M from Andreessen Horowitz in this round, on top of $5M raised a year ago from Khosla Ventures.
AI tools are entering business workflows, but most companies are using them opportunistically – someone manually copies text into a chatbot, gets output, pastes it somewhere else. That's useful but not scalable. The productivity gains from AI come from embedding it in structured processes that run consistently, with the right people intervening at the right moments and AI handling everything else.
Relay's bet is that small and mid-sized companies – which represent 99.9% of businesses by count – need a tool simple enough to configure without an engineering team but powerful enough to handle multi-step, multi-participant processes. Current enterprise workflow tools tend to be expensive, complex, and oriented toward technical buyers. Simpler tools lack the human-in-the-loop and AI-step functionality that make workflows genuinely useful in 2024.
By 2026, market research projects that 75% of organizations will have undergone meaningful digital transformation – which increasingly includes AI adoption, not just cloud adoption. The companies moving through that transition need infrastructure to do it. The investment signal is consistent with this reading: Andreessen Horowitz and Khosla Ventures don't invest in workflow automation tools unless they see a category expansion underway.
A related startup, Flawless – [covered here](/review/vovremja-bystro-i-djoshevo) – is addressing exception handling in automated workflows for e-commerce operations. Relay focuses on standard process automation; Flawless on edge cases. The distinction is useful but the technical infrastructure is similar, and Flawless's natural next step is adding AI to detect what constitutes an exception in the first place.
The opportunity is building simple, flexible workflow platforms that small and mid-sized companies can actually configure themselves – with AI operations and human approval gates as first-class elements, not add-ons.
The most defensible version of this product serves a specific vertical deeply rather than starting as a horizontal tool. The general case is competitive; a platform purpose-built for e-commerce operations, legal intake, or client onboarding in professional services has a tighter value proposition and a clearer reason for someone to trust it with their processes.
There's also an interesting product direction in the distinction between routine processes and exception handling. Standard workflows and exception workflows need different triggering logic, escalation paths, and notification behavior. A platform that models both – and knows which mode it's in – would be meaningfully more complete than either Relay or Flawless in their current forms. The infrastructure for this already exists; it's primarily a product framing problem. Even in its current simplified form, Relay has demonstrated that there's demand and investor conviction in this direction. The category is early enough that well-designed new entrants still have room.