Home
Apr 13, 2026
View All

Automating Your Workday Without Code and the Agentic Commerce Shift

·1 underrepresented voice

The Short Version#

Two threads worth sitting with today: JJ Englert's Claude Cowork tutorial shows how non-engineers are now building real automation workflows — no code required — and Stripe's Veni Singh breaks down how agents, digital wallets, and trust are restructuring checkout in ways that go well beyond payment UX.

Lenny's Newsletter — Claude Cowork 101: Automating Your Workday Without Code#

Source: https://www.lennysnewsletter.com/p/claude-cowork-101-how-to-automate Credibility: High (video tutorial from a practitioner via Lenny's How I AI series)

What happened: JJ Englert from Tenex walked through how to go from zero to one with Claude Cowork — Claude's tool for connecting business apps and building AI "skills" that automate emails, Slack messages, and daily workflows. The tutorial is explicitly aimed at non-engineers. You connect your tools, define skills (essentially lightweight agents that can act on your behalf), and Claude executes against them.

Key capabilities:

  • Connect business tools directly to Claude (email, Slack, calendar, etc.)
  • Define "AI skills" — reusable automations triggered by natural language or schedule
  • Automate workflows like email triage, Slack drafts, and daily status updates
  • No code required; the tutorial assumes zero technical background

Why it matters for PMs: This is the clearest signal yet that agentic workflow automation is moving from developer playground to non-technical adoption. When your marketing lead or ops manager can build a Slack automation in Claude without filing an IT ticket, the implications for product teams are real: the tools your users are assembling around your product are changing. If your product touches any of the surfaces Claude Cowork can connect to (email, Slack, calendar), you need to understand what your users are automating — and whether that's happening with or without you. It also raises the bar on what "productivity feature" means in 2026. A PM who hasn't tried this has a blind spot.

Critical questions:

  • How reliable are these "skills" in practice? Automations that occasionally misfire will erode trust faster than they build value.
  • Is Claude Cowork building toward a platform (user-created skills shared across teams) or staying individual-only?
  • If non-technical users are assembling their own AI workflows, what's the product surface your team owns — and what's getting disintermediated?
  • What does user support look like when an AI-automated Slack message goes wrong?

Action you could take today: Watch the tutorial and map one workflow from your own workday against what Claude Cowork can handle. Even if you don't ship anything, knowing what's now table-stakes for "no-code automation" will sharpen your next prioritization conversation.

Stripe — How Agents, Digital Wallets, and Trust Are Rewriting Checkout#

Source: https://stripe.com/blog/product Credibility: High (first-party analysis from Stripe's PM on payments and checkout, Veni Singh)

What happened: Veni Singh, PM for OCS and Payments Dashboard at Stripe, published an analysis of how checkout is changing in 2026. The piece is based on Stripe's checkout activity data and covers three shifts: AI agents are beginning to act as purchasing intermediaries (agentic commerce), digital wallets have hit a tipping point in consumer preference, and trust signals (brand recognition, security cues) are increasingly decisive at the point of payment. This is not speculation — Stripe has visibility into actual checkout patterns at scale.

Key patterns:

  • Agentic commerce is real and growing: Agents are completing purchases on users' behalf. Stripe is actively building for this interaction model, not just anticipating it.
  • Digital wallets now dominant: Consumer preference for wallet-based checkout (Apple Pay, Google Pay) has crossed a threshold where friction in the wallet flow is a leading conversion killer.
  • Trust is the new conversion lever: Users abandoning checkout due to unfamiliar brand or security uncertainty — and familiar trust signals (known brand names, security badges) are measurably reducing drop-off.
  • The piece frames these three as compounding: agents need trusted payment rails, wallets need frictionless flows, and both need trust infrastructure to close.

Why it matters for PMs: If you're building any product with a payment or purchase surface, this is the context you need. Agentic commerce means designing for a buyer who isn't a human clicking buttons — your checkout flow needs to work for agents acting on behalf of users. That's a fundamentally different UX problem. The trust piece also lands in a specific place for fintech PMs: in a world where agents execute purchases, the human's trust in the agent's choices becomes a product design challenge, not just a UX one.

Critical questions:

  • How do you design a checkout flow that's optimized for both human and agent buyers simultaneously?
  • What authentication patterns work for agentic purchases without introducing friction for human users?
  • As agents become buyers, how does the relationship between consumer trust and brand recognition shift?
  • Stripe has the data advantage here — how long before they surface agent-specific checkout analytics as a product feature?

Action you could take today: Audit your current checkout or payment UX against these three vectors: wallet-first flow, trust signal placement, and whether your flow would work at all if an AI agent were executing it on a user's behalf.

Microsoft Copilot Studio — Automate Business Processes by Mixing AI Agents and Workflows#

Source: https://www.microsoft.com/en-us/microsoft-copilot/blog/copilot-studio/automate-business-processes-with-agents-plus-workflows-in-microsoft-copilot-studio/ Credibility: High (first-party product announcement from Microsoft)

What happened: Microsoft shipped new capabilities in Copilot Studio that let organizations mix AI agents with traditional workflow automation. The core change: you can now combine structured, rule-based process flows with dynamic AI agent behavior in a single automation. This means a process can hand off from a deterministic step to an agent decision and back — rather than forcing a choice between "pure automation" and "pure AI."

Key capabilities:

  • Mix rules-based workflow steps with AI agent actions in a single business process
  • Agents can handle judgment calls mid-process, then return control to structured steps
  • Targeted at enterprise business process automation (HR, IT, operations workflows)
  • Builds on the Copilot Studio platform, so IT-governed rollout with existing Microsoft security controls

Why it matters for PMs: The "AI agent vs. workflow" framing is a false choice — and Microsoft just shipped a product that makes that concrete. The practical PM takeaway is architectural: if you're designing business process automation, the right model is usually hybrid. Agents handle ambiguity; workflows handle compliance, auditability, and repeatability. Building them as separate systems is technical debt. This also signals where enterprise AI tooling is headed: not "replace your workflows with agents" but "give your workflows judgment where they need it."

Critical questions:

  • How does Copilot Studio handle failure modes when an agent makes a wrong call mid-process? What does rollback look like?
  • Is the "mix" genuinely flexible, or are there constraints on where in a workflow an agent can intervene?
  • How does this compare to LangGraph's approach to human-in-the-loop agent workflows — is this the enterprise-packaged version of the same idea?
  • What's the governance and audit story for agent decisions embedded in regulated business processes?

Action you could take today: If your team has any existing workflow automations (Zapier, Power Automate, internal scripts), sketch what one of them would look like with an AI decision node inserted at the step that currently requires human judgment. That's the mental model Microsoft is now productizing.

Quick Hits#

  • Karri Saarinen (Linear): Posted on LinkedIn (April 12) that "many AI workflows still lose context at handoff... Linear is for the continuity of the thread and compounding of the context." Clear signal that Linear is positioning itself as the memory layer for AI-assisted development teams, not just an issue tracker: https://www.linkedin.com/in/karrisaarinen/

  • Vercel: Shipped "Copy-to-Prompt instructions for Flags" — the feature flags page now generates prompt instructions so you or your agent can install the Flags SDK and configure feature flags directly from AI context. Small but meaningful: feature flag tooling is now explicitly designed for agent-driven workflows (April 13): https://vercel.com/changelog/copy-to-prompt-instructions-now-available-for-flags

  • Lenny's Newsletter / Keith Rabois: New episode on "hard truths about building in the AI era" covers his view that AI is collapsing the PM role, the barrels vs. ammunition hiring framework, and why talking to customers hurts consumer products. Worth a listen if you're thinking about how PM scope is shifting (April 12): https://www.lennysnewsletter.com/p/hard-truths-about-building-in-the-ai-era (Note: URL covered in prior update — skip if deduplication strict; content is new episode)

  • Notion: Announced data residency expansion to Japan and South Korea (May 2026, Enterprise plan). Not AI-specific, but a consistent enterprise signal — data sovereignty is now table stakes for global SaaS expansion (April 13): https://www.notion.com/blog/notion-expands-data-residency-to-japan-south-korea

  • Harrison Chase (LangChain): Tweeted on the importance of model-agnostic harnesses — arguing that the harness layer is where teams should invest for durability as models keep changing. Connects directly to the continual learning framework he published earlier this week (April 9): https://x.com/hwchase17/status/2042793872456130660

The Thread#

Agentic workflows are moving from framework to product. This week: Copilot Studio ships hybrid agent+workflow automation, Vercel makes feature flags agent-readable, Claude Cowork lets non-engineers build agent-driven automations, and Stripe's PM describes agents as active participants in checkout. These aren't separate trends — they're the same transition arriving at different layers of the stack simultaneously. The infrastructure for agents-in-production is now shipping as packaged product, not just as developer frameworks.

Sit With This#

Stripe's Veni Singh describes AI agents completing purchases on behalf of users — and the trust and authentication challenges that creates. Your checkout flow was designed for a human making a conscious decision to buy.

For your product: If an AI agent were executing purchases on your users' behalf, what would break first — the UX, the auth flow, or your fraud detection? And which of those is a 2026 problem, not a 2028 one?