News
Research Report
The AI Transformation 100: Concrete Ideas for Fixing How We Work
In Brief
- Claude orchestrates multiple subagents for complex coding tasks.
- Faster maintenance and modernization reduce manual effort and turnaround time.
- Parallels to how modern knowledge work is coordinated suggest dynamic workflows may move beyond developer tool.
Anthropic launched dynamic workflows in Claude Code on May 28, which allows the system to construct and execute task-specific structures at runtime rather than following a fixed sequence.
Instead of handling problems linearly, Claude decomposes tasks, assigns subtasks to multiple parallel subagents and synthesizes results after internal comparison and verification.
The feature — currently targeted towards developers — is available in research preview across the Claude Code CLI, Desktop and VS Code extension for Max, Team and Enterprise plans. It is also accessible via the Claude API, Amazon Bedrock, Vertex AI and Microsoft Foundry.
The company explicitly noted that dynamic workflows can consume significantly more tokens than a typical Claude Code session, advising users test the tool on a task with limited scope to understand the costs.
Claude Code Dynamic Workflows Overview
| Capability | Description |
|---|---|
| Dynamic workflows | Orchestrates parallel subagents for complex, multi-step coding tasks |
| Parallel subagent orchestration | Runs tens to hundreds of subagents simultaneously in one session |
| Adversarial verification | Independent agents attempt to refute findings before surfacing results |
| Progress persistence | Interrupted jobs resume where they left off rather than restarting |
Dynamic Workflows Reshape Large-Scale Code Work
Dynamic workflows means Claude can take a large task and split it into smaller parts, then run those parts at the same time using multiple internal agents instead of solving everything sequentially.
The tool is meant to be used for complex tasks which typically take hours or days. You need multiple agents because complex problems aren’t just one step, they involve different types of work happening in parallel. One agent might read a codebase, another might test assumptions, another might look for bugs or inconsistencies. Doing that sequence in parallel not only provides speed, it lets the system explore more possibilities at once.
The system then compares and reconciles those outputs, checking them against each other before producing a final answer. That verification step increases the chances the system will catch inconsistencies before producing the final result.
The feature, part of a broader Claude Code release, supports migrations across hundreds of thousands of lines of code, according to Anthropic. Dynamic workflows uses an existing test suite as the completion standard, which moves validation from a developer-managed checkpoint to an automated arbiter of success.
From Coding to Knowledge Work
If dynamic workflows follows the same trajectory as Claude Cowork, we may soon see it crossover into knowledge work.
Dynamic workflows operate in a similar way to how enterprise work happenss: fragmented information across systems, parallel inputs and verification across teams, and constant reconciliation of conflicting inputs before decisions get made. Incident response, compliance review, financial analysis, sales operations, internal research are all coordination-heavy problems. The bottleneck typically happens with structuring information, validating it and turning the information into a decision.
Dynamic workflows would map directly onto that coordination layer.
Webinar
Jun
3
From AI Investment to CX Results: What Enterprise Leaders Need to Know
Move beyond experiments. See how top enterprises scale AI for CX results.
Register
Webinar
Jun
4
Content Strategy Leaders Live: Managing Risk, Compliance & AI in Financial Services
Learn how financial services leaders are modernizing content systems without disrupting trust, compliance or experience.
Register
Webinar
Jun
16
From Legacy to Launch-Ready: How Gainbridge Made Its Website a Marketing-Led Growth Engine
Join in to learn how a D2C annuity brand gave marketing full website ownership — without slowing down or risking compliance.
Register
Webinar
On demand
Content Strategy Leaders Live: Managing Scale, Safety & AI in Manufacturing
How manufacturing leaders update content systems to handle product complexity and scale while integrating AI safely.
Watch Now
Webinar
On demand
AI With Intent: How CX Leaders Decide Where Automation Belongs (And Where It Doesn't)
Senior CX leaders from REI & Hyatt reveal their framework for using AI to boost retention.
Watch Now
Webinar
On demand
From Content Sprawl to Competitive Advantage: A Knowledge Operations Roadmap
How automation, AI and a centralized content platform turn knowledge fragmentation into a scalable competitive advantage.
Watch Now
But it will only work if Claude is connected to the workplace stack. Without integrations into tools like Slack or Teams (where plans and actions are discussed), Glean (where knowledge lives) or ServiceNow and Salesforce (where actions are executed), Claude would be operating on limited context. It can analyze work, but it cannot participate in it.
Anthropic's Potential 2026 IPO
Anthropic's actions of late have all the signs of a business moving towards an IPO. The company has quickly amassed funding — its most recent round valuing the company near $1 trillion dollars — and launched features to improve its value for the enterprise.
Key launches include the Claude Code Slack integration (December 2025), Claude Cowork (January 2026), Sonnet 4.6 (February), Managed Agents and Claude Design (April 2026). Enterprise integrations span Google Workspace, Asana, Apple's Xcode 26.3 and Goldman Sachs, while the acquisition of Vercept strengthened Claude's ability to interact with live software environments.