Zapier’s Cofounder Shares How He’s Implementing Light-Weight Project Management Systems and Agentic Workflows

Faye Wai

Last updated on Nov 4, 2025

QUICK SUMMARY

Wade Foster is a cofounder of the tool that many of us already use to automate workflows and connect software: Zapier. He is currently heading the charge to integrate AI deeply into both the organization and the tool.

TABLE OF CONTENTS Founding Zapier Agentic Workflows AI-assisted Delivery Rituals Zapier's "AI Actions" Feature AI-Adoption Culture AI Automation Engineer as a Role Future of Project Management

We caught up with Wade to get a peek at how project management is changing within the organization that is, itself, changing project management.

Here's what he had to say.

Founding Zapier, The Business that Transformed Automation

I'm the cofounder and CEO of Zapier.

While I’m not a project manager by title, I’m helping with delivery at scale and championing AI orchestration across the org. In fact, we published our internal playbook on driving 97% AI adoption across the employee base.

My role is to set direction, remove roadblocks, and make sure we’re shipping the highest impact work for our customers — faster. At Zapier, project delivery is less about rigid processes and more about building systems and automation that empower people to do their best work.

Shifting From Traditional Project Management to Lightweight AI-driven Systems

Traditional Gantt charts and heavyweight status decks don’t cut it in a remote, async-first world. We’ve shifted to lightweight, automation-first systems.

For example, instead of weekly status meetings, updates are auto-generated in Slack channels from project tools. Leaders only jump in if there’s a flag. That saves meeting hours and makes progress transparent by default.

Here's a breakdown of what we are doing — and what we aren't.

What we’re moving away from:

Ad‑hoc team-by-team conventions for Jira projects, roadmaps, and reviews that created inconsistency, hidden dependencies, and reporting churn across EPD.
Status pages and long update docs that drifted from reality because upstream data wasn’t standardized or automated.
Unenforced delivery hygiene (e.g., epics in progress without a target date), which made planning and downstream coordination unreliable.

What we’re moving toward:

Standardized “golden paths” for core delivery tools—like shared Jira project structures, Productboard roadmap intake pilots, and opinionated Figma structures—so status and dependencies are consistent and computable across teams.
A build‑wide operating system. This is a single flow that consolidates weekly demos and a public‑facing changelog (including WIP), with release notes automated end‑to‑end to cut reporting waste and increase visibility.
Lightweight delivery guardrails in the flow of work—e.g., a Slack‑announced validator requiring a due date when a Jira Epic moves to “In Progress”—so plans stay real without adding meetings.
A clear remit for Build Ops to run and maintain the shared configs and defaults (the “platform for how we work”), evolving them with teams while tracking core delivery, quality, and capacity metrics in Zone MORs for transparency.

Steps Zapier Took to Implement Lightweight Project Management

Here's what we've done to make that transition:

Established Build Ops’ roles and scope to “drive scalability, automation, and efficiency across EPD,” then staffed for engineering‑leaning ops, with a tool stack spanning Zapier, ChatGPT, Productboard, Figma, Jira, Jellyfish, Coda, OpsLevel, and Looker/Looker Studio to wire together delivery and reporting flows.
Shipped golden‑path templates and shared configurations: standardized Jira structures, Productboard pilots for roadmap intake, and Figma org patterns. Build Ops owns upkeep so squads don’t reinvent processes locally.
Ran the “BuildOS” pilot in the Editor subzone, meaning that we stood up a unified changelog workflow that consolidates internal WIP with weekly demos and public‑facing updates, with an automated release‑notes system to reduce manual comms overhead.
Added delivery validators and norms in‑tool — not in a doc. That means we announced the “Epic must have a due date when moving to In Progress” rule and created guidance for communicating date changes, including references to forecasting support in Jellyfish.

And this is what we're seeing as a result:

Less administrative and reporting waste, with roadmaps and release notes auto‑generated from the actual system of work (rather than hand‑assembled decks).
Clearer cross‑team coordination via shared structures and metrics, improving planning and dependency management across teams as golden paths expand.
Better delivery hygiene and stakeholder alignment as due‑date expectations are enforced at the point of change and surfaced consistently downstream.

An AI-Powered Workflow for Team Transparency

Let's dive deeper into the release notes system I mentioned. At Zapier, we have a value: Default to transparency.

That means we share a lot of our works in progress. We have a strong culture of sharing demos, changelogs, and results across the company.

As we've grown, we’ve done this in many different ways and those artifacts ended up logged in a bunch of places, making it harder for anyone to get a pulse on what’s happening inside the company. So our Build Ops team fixed this by creating a simple flow where a team can submit updates once, and the system automatically routes those artifacts to the right tools and Slack channels.

We built it with Zapier Interfaces, Tables, and Zaps. Submissions go into Tables as the source of truth, sync to Coda for tracking, and post updates into #feed-delivery-updates in Slack, so everyone sees the same information in real time.

It wasn’t perfect at first. But once we ironed out the alignment and edge cases, it gave us 100% coverage of shipped work, reduced duplicate reporting, and made it easy for downstream teams like Support or GTM to get what they need without chasing PMs.

We’re now layering in AI to auto-summarize updates and even draft help docs, so communication scales with less manual effort.

How Zapier Uses Agentic Workflows to Boost Productivity

We've been experimenting with agentic workflows, and the biggest wins so far have been in two areas:

Front-door orchestration: We’ve deployed agents that classify, enrich, and route requests coming from Slack, forms, or email. They push them into the right queues with SLA-based nudges and even trigger escalations when deadlines slip. This cuts down on triage overhead and ensures requests don’t get lost in the shuffle.
Research and content ops: Our marketing team uses agents to gather context, prep campaign assets, QA drafts, and push them into the right tools. Humans approve in Slack, and then distribution runs automatically. It’s a good example of compounding value when the whole flow is wired end-to-end.

What’s worked best is combining deterministic Zaps, AI steps, and constrained agents. With approvals and observability built in, these systems are reliable enough for production use.

The adoption pattern has also been clear: Start small with one workflow that saves real time, prove it works, and then expand scope. That incremental approach has made adoption stick across teams.

Wade Foster

Co-founder & CEO at Zapier

AI-assisted Delivery Rituals For Optimal Project Execution

Our delivery rituals haven’t disappeared in an AI-first world, but they’ve evolved to be lighter weight, more transparent, and AI-assisted.

Defining scope: We use our ZIP Execution Model, which requires epic one-pagers, DACI roles, and explicit pre-commit sign-off before anything moves to “Planned.” Every Jira epic links back to a Productboard item, so scope is traceable end-to-end. AI shows up here by helping teams explore options and draft those one-pagers faster, but approval gates and traceability stay the same.
Aligning teams: We apply DACI explicitly and keep a steady roadmap cadence — quarterly reviews in Productboard, biweekly reviews for high-visibility work. Because AI fluency is high across the org—engineers, PMs, designers—alignment benefits from AI-generated drafts and prototypes that speed up decision-making.
Validating work: As I mentioned, we consolidated demos, changelogs, and results logs into a single Build Ops–run submission flow that posts to Slack, giving 100% coverage of shipped work. AI helps generate summaries, diffs, and call-outs from artifacts, but final quality and safety approvals are always human-owned.
Managing execution: Intake is normalized across channels and funneled into Productboard and Jira. Execution telemetry stays live and lightweight, with Build Ops “golden paths” keeping hygiene consistent. AI is embedded “in the flow” — engineers rely on AI dev tools daily, product teams use copilots to cut down admin toil — but ownership, approvals, and audit trails remain deterministic.

The bottom line:

AI accelerates the rituals but doesn’t replace them. Humans still own the judgment calls, while AI reduces friction so teams can focus on the highest-value work.

Rapid Prototyping in Product Planning with AI Tools

But even with all of these big changes, I think the biggest shift for me has been adding prototyping in the product planning process.

With vibe-coding tools like v0, you can skip a lot of lengthy reading, writing, and documentation. Instead, you build a clickable prototype.

There's a saying, “A picture is worth a thousand words.” Well, a prototype is worth a thousand pictures. So it's really powerful for condensing timelines. It cuts out weeks from most features.

There’s a saying, “A picture is worth a thousand words.” Well, a prototype is worth a thousand pictures. So it’s really powerful for condensing timelines. It cuts out weeks from most features.

Wade Foster

Co-founder & CEO at Zapier

And that means there are opportunities to spend time with customers, trying to understand the problems they face and where the product helps or hurts them in their workflows.

Zapier's "AI Actions" Feature Allows AI to Take Action For You

No surprise here… a lot of our tech stack is Zapier: Zaps, AI by Zapier, Agents, Tables, Interfaces, plus MCP to connect external AI tools to thousands of actions. This has evolved in the last 6-12 months by us doubling down on a unified builder surface (Zaps + Tables + Interfaces) and adding Agents to bring judgment into structured workflows.

We also leaned into MCP so builders using tools like ChatGPT, Claude, and Cursor can safely take action across apps without net‑new integrations. We introduced plan changes that package the core building blocks together for simpler rollout and governance across teams. Impact is highest where teams use the full stack:

Deterministic automation for the backbone
AI steps for inference
And agents for the “last mile” of variable work—with approvals and observability built in.

Collaboration/ops surface varies by team. For example, our marketing org connects Slack, Coda, Iterable, HubSpot, and Databricks via Zapier to compress handoffs and ship faster with auto‑generated email assets, approvals, and distribution.

Of all the tools we've used, Zapier's own “AI Actions” feature has honestly been the biggest game-changer. It lets us build workflows where AI doesn’t just give you direction on how to do it yourself but takes action for you, like updating a project tracker or flagging a missed dependency. And non-technical teams can set this up themselves without waiting on engineers.

Building a Strong AI-Adoption Culture at Zapier

Tooling is important, but even more important is the culture you set.

Tooling is important, but even more important is the culture you set.

It's so easy in a large organization to keep doing what you've always done in the past, but when a game‑changing technology like AI comes along, you have to find mechanisms to get the organization to operate differently.

We've done that by investing a lot in hackathons, show-and-tells at our all‑hands meetings, and Slack channels where people can share workflows or learnings.

All these cultural moments start to compound, where it becomes more and more normal for folks to turn to AI and say, "Hm, how might I do my work differently to solve this problem?"

Why AI-first Delivery Orgs Need an AI Automation Engineer

If you want an AI-first delivery org, consider hiring a new role: AI Automation Engineer (or “orchestration engineer”)

Wade's Tip

If you want an AI-first delivery org, consider hiring a new role: AI Automation Engineer (or “orchestration engineer”)w

It’s the connective tissue between strategy and shipped systems, and it accelerates both delivery and culture change.

Use AI to Handle Tedious and High-Cost Tasks

As you can see, a lot of our AI adoption is meant to help humans with the work that we already do today, and there's a ton of value to be gained from that.

There are also opportunities in the things humans can't do. What are we unable to tackle because it costs too much or is too tedious? Invest in those.

A great example is support documentation. For most companies, keeping a knowledge base up to date is a tedious, time‑consuming task that humans are not particularly good at. However, with AI, you can take all your support tickets and automatically turn them into a rich knowledge base.

These are the types of projects we are now looking at. We identify something we’re simply not doing but should be, and allow AI to make it much easier.

The future of Project Management UX with AI and Orchestration

Here's the future, as I see it.

Backlogs will have owners — some human, some agentic. PMs and leads become outcome designers and orchestration architects.

And most “project management” UX becomes a thin layer on top of orchestration platforms: Live systems run the work, status is computed, humans provide goals, guardrails, and judgment.

Lead with orchestration; not a chatbot

If I can give any advice, it's this: Lead with orchestration, not a chatbot. Get AI into the flow of work where it can take action and be governed.

Pair top‑down urgency with bottom‑up building. Run a hackathon, publish guardrails, appoint champions, and share playbooks. That’s how we drove 97% adoption internally.

And start with two “wedge” workflows that matter. Make them production‑reliable with approvals and logs; measure cycle time, error rates, and stakeholder NPS; then scale patterns org‑wide.

Follow Along

You can follow along as Wade continues to change the game of automation on X and LinkedIn. And, of course, check out Zapier!

More expert interviews to come on The Digital Project Manager.

Unlock for Free