Superpowers for your team - Vladimir Druts
AI's Transformative Impact on Work
Introduction to AI Developments
- The speaker introduces themselves and their experience in AI since 2022, highlighting the early days with ChatGPT 3.5.
- They express excitement about recent advancements in AI, likening the current changes to a "tidal wave" that will significantly impact the economy and productivity.
Personal Experience with AI
- The speaker shares how Agent AI has increased their work volume by 100 times without compromising quality, allowing them to focus on impactful tasks.
- A client reports a dramatic increase in productivity, stating they accomplished more work in one week than in the previous six months.
Presentation Overview
- The speaker outlines the agenda for a brisk 30-minute presentation, including a Q&A session and live demos of CoWork and other tools.
- They mention using Claude CoWork, an anthropic system that operates best on Mac but is being demonstrated on Windows.
Current State of AI Usage
- The speaker engages the audience by asking how many have ChatGPT or Claude open, noting that most people use AI at some level but haven't fully transitioned from chatting to delegating tasks.
- A chart illustrates the timeline of software task completion rates, marking significant inflection points in AI development.
Inflection Points in AI Development
- The first major inflection point was chat capabilities around 2023 with GPT4; reasoning abilities emerged around late 2024 to 2025.
- By 2026, there is expected to be exponential growth (hockey stick trajectory), particularly concerning agent-based systems capable of autonomous actions.
Adoption Curve and Workforce Impact
- Jensen Huang from Nvidia emphasizes that every company needs an open claw strategy as part of its operational framework.
- Statistics reveal that while many workers use free chatbots (1.3 billion), only a small fraction engage deeply with coding scaffolds (2 to 5 million).
Time Efficiency Insights
- A study indicates significant time waste within teams due to tasks manageable by AI; recovering just 10% could save nearly $94k annually for teams of at least 30 people.
Layers of Claude Operating System
- The operating system consists of three layers: basic chat functionality (Claude), collaborative work management (Claude CoWork), and advanced experimentation tools (Claude Code).
- Claude CoWork can manage files and produce deliverables autonomously, while Claude Code offers sophisticated features beyond coding.
Understanding Claude Co-work as a Team Member
Reframing AI Interaction
- Instead of viewing Claude Co-work merely as a tool, it should be considered a team member that synthesizes information and cross-references documents to produce work products.
The Mechanics of LLMs
- Large Language Models (LLMs) operate using tokens, which represent the brain power of the model. A token is approximately three-quarters of a word.
Energy Consumption in Conversations
- Every interaction with an LLM consumes energy; thus, prompts should be clear and detailed to minimize energy expenditure.
Context Management
- Maintaining focus on one topic per conversation is crucial for effective communication with LLMs, as context switching can lead to confusion and loss of information.
Model Variants: Sonnet vs. Opus
- Two primary models are discussed: Sonnet for everyday tasks and Opus for deep analysis, with Opus recently upgraded to handle 1 million tokens.
Exploring the Claude Co-work Ecosystem
Application Overview
- Claude Co-work is available as a desktop application for Windows and OSX, allowing users to chat, co-work, or code while accessing files securely.
Memory Management in Projects
- Unlike traditional chat interfaces that remember all interactions over time, Claude Co-work retains memory only within specific projects through MD files.
Skills and Connectors Explained
- Skills function as training modules (SOPs), connectors allow integration with platforms like Google Drive and Slack, while extensions enhance capabilities.
Utilizing Skills and Extensions Effectively
Creating Effective Skills
- Users can create skills by instructing Claude on specific tasks or methods they want it to remember across sessions within projects.
Expanding Functionality Through Connectors
- Connectors provide access to various applications enhancing productivity by linking tools like calendars or project management software seamlessly.
Community Contributions
- There exists a vibrant community creating skills that can be shared among users, significantly boosting the functionality of Claude Co-work.
Claude Co-Work: Integrating Tools for Enhanced Productivity
Overview of Claude's Capabilities
- Claude Co-work can connect with various tools like Google Drive, GitHub, Excalidraw, and Notion, allowing it to log in under your account and utilize these resources effectively.
- It can create plans in Notion, update projects, access real-time calendar information, and provide prompts about upcoming meetings while integrating into company Slack directories.
File Management and Automation
- Claude has the ability to read various file types including PDFs and spreadsheets. It can also schedule activities based on this data.
- The system operates locally by reading and writing files without requiring constant user input; it can browse the web and send texts via phone through a dispatch function.
Data Control and Project Structuring
- Users maintain control over their data; Anthropic does not train on team plan data unless opted out in personal pro plans.
- For project organization, starting with three MD files is recommended: one for business context (claude.mmd), detailing priorities, guardrails, roles, responsibilities, and preferred communication styles.
Dynamic Learning and Adaptation
- These MD files are designed to evolve as Claude learns from ongoing interactions within projects. They are plain text files that require no coding knowledge.
- Establishing a single source of truth within an organization is crucial to ensure consistency across teams while allowing individual contributions.
Practical Demonstration of Features
- A demonstration involves creating a morning brief by prompting Claude to review all relevant files in a designated folder.
- The AI generates a structured morning brief based on user preferences outlined in the claude.md file using either the sonnet or opus model.
Summary of Generated Brief
- The generated brief includes details such as today's calendar events, Slack highlights, open tasks prioritized according to urgency.
- This functionality allows users to receive updates tailored to their needs across connected tools while maintaining flexibility in how information is presented.
Quarterly Priorities and Team Context
Overview of Team Structure and SOP
- Discussion on quarterly priorities, emphasizing the importance of understanding team context and roles within the organization.
- Introduction to "clawmd," a tool designed to facilitate communication, ownership, and prioritization of speed over perfection in workflows.
Demonstration of SOP Creation
- Example presented where fragmented files are used to create a cohesive Standard Operating Procedure (SOP), highlighting the challenges with existing materials.
- Explanation of how Claude Co-work can aggregate information from various tools across an organization to produce a comprehensive work product.
Automated SOP Generation
- The process involves reviewing multiple files and sources such as Slack or Notion to compile step-by-step instructions for an SOP.
- Claude identifies gaps in source materials during its analysis, showcasing its ability to enhance content quality by flagging inconsistencies.
Transforming SOP into Branded Outputs
Creating Professional Documents
- Capability of turning generated SOP documents into branded PDFs, enhancing presentation quality for stakeholders.
Daily Content Generation
- Description of using Claude Code for curating relevant news content tailored to personal preferences, demonstrating its adaptability in generating ideas based on trending topics.
Implementation Strategy for Organizations
Structuring Internal Rollout
- Proposal for organizations to start with one team as early adopters who can evangelize the use of Claude within their workflows.
Building Organizational Assets
- Emphasis on creating a master "claude.md" file that standardizes processes across teams while preventing context drift and ensuring consistent outputs.
Advanced Features and Automation
Scheduled Tasks and Updates
- Highlighting the ability of Claude to automate routine tasks like drafting executive briefs or board updates without prompting, thus saving time for users.
AI Tools and Their Impact on Productivity
The Current Landscape of AI Utilization
- The speaker discusses the ability of AI to categorize unread emails and generate P&L summaries, emphasizing its role in automating tasks based on predefined rules.
- Highlights that current AI companies are subsidizing token costs, making advanced tools accessible at a fraction of traditional costs, akin to having a full workforce for minimal expenses.
- Compares the current state of AI with the early days of Facebook ads, where low costs allowed early adopters to build significant personal brands and capitalize on cheap advertising opportunities.
Urgency in Adopting AI Technologies
- Encourages immediate engagement with available AI tools before prices rise, suggesting that users should start leveraging these technologies to enhance productivity.
- Introduces the concept of using multiple parallel agents for different projects, which can lead to improved outcomes through collaborative efforts.
Enhancing Output Quality Through Collaboration
- Describes how creating competitive personalities among agents can improve work product quality by simulating diverse perspectives during evaluations.
- Suggests forming specialized sub-agents (e.g., CMO, CFO) for targeted insights and better content generation while acknowledging ongoing challenges with hallucination in outputs.
Practical Steps for Implementation
- Offers practical advice on starting with Claude's tools by creating folders and drafting initial company descriptions as a way to familiarize oneself with the technology.
- Notes that many individuals feel overwhelmed when starting but often experience breakthroughs once they begin experimenting with AI tools.
Real-world Applications and Benefits
- Shares personal experiences using Claude for presentation creation, demonstrating its capability in generating branded materials quickly and efficiently.
- Discusses how Claude assisted in planning demos and managing presentations effectively through automated features like timers across slides.
Future Prospects of Digital Assistance
- Envisions a future where digital assistants function as chief-of-staff roles within organizations, aiding decision-making processes and streamlining operations.
- Mentions conducting competitive audits across multiple companies simultaneously using advanced frameworks provided by Claude Code.
Project Management Agents: Transforming Business Operations
The Role of AI in Project Management
- The speaker discusses the utility of a project management agent that assists with follow-ups, deliverables, risks, and status updates. This agent is always accessible via phone and possesses extensive knowledge about the user.
- The speaker mentions using "OpenClaw" on Telegram, where they have virtual coaches with unique personalities who provide 24/7 business coaching. These agents are designed to understand the user's context and assist in their work.
- The transformative impact of these AI agents is highlighted; they help identify gaps in the user's workflow and support ongoing projects effectively.
Practical Steps for Implementation
- The speaker recommends visiting horizon.ai to access a calculator that estimates operational friction within teams based on size and business type. Users can download a comprehensive report for further insights.
- A call to action encourages users to start utilizing these tools immediately by downloading "Claude," emphasizing its potential benefits for enhancing productivity.