The Top 10 Things You Should Immediately Do
If You're Interested in AI Workplace Productivity
Note: This has nothing to do with AI images, videos, or the flashy stuff dominating headlines.
After nine months of daily use in my consulting practice, I've noticed a pattern. The professionals seeing real productivity gains aren't the ones chasing the latest AI art generators or video tools. They're quietly mastering a set of practical fundamentals that compound over time.
What actually moves the needle? Start with these practices rather than waiting for the "perfect" moment.
1. Get a Pro-Level AI Account (Start Here, Don't Overthink It)
This isn't optional anymore. Free tiers of ChatGPT and Claude hit usage limits precisely when you need them most—during complex problem-solving sessions.
What to get: ChatGPT Pro ($20/month) or Claude Pro ($20/month)¹. If you're working in a company, get a Teams account instead—being able to share projects and maintain team context is just as important as the advanced models.
Why this matters: Claude handles Drive documents well and plays nicely with Workspace workflows, though it lacks native Gmail or Calendar integration like Gemini does. ChatGPT's Google Drive integration and project system excel at maintaining context across longer workflows. I find myself using Claude for research and document-heavy work, ChatGPT mostly for fact checking, image generation, and as a less expensive text and code analyzer.
Bottom line: This investment eliminates the friction that kills AI adoption. When you're not rationing usage, you actually use the tools.
2. Stay Critical of AI Responses (This Should Be Second Nature)
Before diving into advanced techniques, develop healthy skepticism. LLMs are sophisticated prediction machines, not truth engines.
Common failure modes I've observed:
Plausible-sounding but factually wrong technical details
Overly polished language that lacks authentic voice
Confident assertions about rapidly changing technologies
Code that compiles but misses edge cases
Making up supporting stories in an article (it just did it here)
Practical approach: Treat AI outputs as "smart first drafts" requiring your domain expertise. You bring the judgment, context, and taste. The AI provides the raw material.
After several months of daily use, I can spot AI hallucinations quickly. You develop an intuition for when responses feel too perfect or confidently wrong.
3. Master Prompting with AI Assistance
Here's the counterintuitive part: instead of manually crafting perfect prompts, describe what you're trying to accomplish and let the AI write the prompt.
Example workflow:
"I need a prompt that will help me analyze technical requirements documents for potential implementation risks"
Review the AI-generated prompt
Refine based on your specific needs
Save the final version for reuse
This approach teaches you effective prompting patterns while solving immediate problems. After a few weeks, you'll internalize the structure and won't need the training wheels.
Key insight: Good prompts provide context, specify format, and include examples. But you don't need to figure this out from scratch.
4. Create a Personalized Writing Prompt
This single investment saves hours weekly. Gather 6-8 samples of your writing—the longer the better—and ask the AI to analyze your voice and style. I went back and used Gmail and Google Drive to get 10 years worth of writing samples across emails, memos, and technical documents (yes that may be a bit extreme!).
The process:
Collect diverse writing samples from Gmail and Google Drive spanning several years
Upload writing samples to Claude or ChatGPT
Prompt: "Analyze these writing samples and create a style guide that captures my voice, tone, and structural patterns"
Test the resulting prompt on new content
Refine until the output requires minimal editing
What changed for me: My AI-generated first drafts now sound authentically like me, reducing editing time by 60-70%. The key is being honest about your actual writing style, not the style you think you should have.
5. Use AI as a Research Assistant
Stop manually synthesizing information from multiple sources. AI excels at first-pass research, concept disambiguation, and knowledge compilation.
Effective research patterns:
"Research X topic and provide a structured overview with key concepts, current trends, and implementation considerations"
Upload PDFs and ask for specific analysis
Request comparison frameworks between competing approaches
Where this shines: Exploring unfamiliar technical domains, competitive analysis, and building context before diving deep. Perplexity is particularly good at this type of research, though I use it less and less now that deep search is part of both Claude and GPT. I regularly use this for evaluating new development tools and understanding market landscapes.
Critical caveat: Always verify facts and cross-check sources, especially for recent developments or specific claims. For more on [effective AI research workflows], see my detailed breakdown.
6. Utilize Multiple AI Models
Each model has distinct strengths. Even if you start with just one platform, understanding what each excels at helps you make better decisions about when to expand or switch tools.
My current allocation:
Claude: Research, document analysis, writing assistance, complex reasoning
ChatGPT: Code generation, project management, technical explanations
Perplexity: Real-time information, fact-checking, current events²
Gemini: Google Workspace integration, alternative perspectives
Practical tip: If you have access to multiple models, test the same prompt across them for new projects. You'll quickly identify which tool fits best for different work types. For a detailed comparison of capabilities, see [my analysis of Claude vs ChatGPT for technical work].
7. Learn to Use Claude's Entity Editor and GPT's Canvas Editor
These AI-native editing environments are becoming central to knowledge work, but they require different skills than traditional word processors.
Claude's Entity Editor: Prompt-only editing with no direct text manipulation. Tedious for small changes but powerful for structural revisions and iterative development.
GPT's Canvas Editor: Hybrid approach allowing both direct edits and prompt-based changes. Better for shorter documents but has length limitations.
Essential skill: Get fluent with Markdown. Both platforms rely on it heavily, and knowing basic formatting dramatically improves your editing efficiency.
Pro tip: Learn the import/export patterns. You'll often start in one tool and finish in another.
8. Understand Context Windows
Every AI conversation has a memory limit. Long chats get truncated, causing the AI to "forget" earlier context and produce inconsistent responses.
Practical implications:
Complex projects benefit from shorter, focused sessions
Summarize key decisions before context limits hit
Reset threads when you notice quality degrading
Structure prompts to be self-contained when possible
What I've learned: Better to have three focused 20-minute sessions than one meandering 90-minute chat. Plan your AI interactions like you'd plan meetings.
9. Leverage Folders and Threads for Context Management
Modern AI platforms offer organizational tools that maintain continuity across sessions. Use them strategically.
ChatGPT Projects: Maintain persistent memory and custom instructions. Excellent for ongoing technical work where the AI needs to remember your preferences, codebase structure, or project constraints.
Claude Projects: Folder-based organization with shared context documents. Perfect for research projects or complex analysis where you want consistent context across multiple conversations.
Implementation tip: Create separate projects/folders for different types of work. Don't mix code debugging with strategic planning in the same context.
10. Use AI for Ad-Hoc Data Analysis
Instead of wrestling with Excel pivot tables or learning new dashboard tools, export your data as CSV and upload it directly to AI.
Common scenarios:
Sales data analysis and trend identification
User behavior pattern recognition
Performance metrics interpretation
Budget variance analysis
Why this works: AI can write Python/pandas code on the fly, clean messy data, generate visualizations, and explain findings in plain English. No need to set up complex analytics infrastructure for one-off questions.
It's important for non-coders to become comfortable with AI coding for them on the fly for quick projects. We're not talking about building an app—we're talking about another language that will help hugely with more involved data work.
Example: I regularly upload project metrics (deployment frequency, bug rates, cycle time) and ask for trend analysis and anomaly detection. Gets insights in minutes that would take hours in traditional tools.
The Real Productivity Multiplier
Here's what I've observed after implementing these practices across my consulting work: the compound effect matters more than individual techniques. Each practice builds on the others, creating workflows that genuinely transform how you approach knowledge work.
The professionals seeing 2-3x productivity gains aren't using AI for everything. They're strategically applying these fundamentals to eliminate friction in their existing workflows.
Start with items 1-3 this week. Add one new practice every two weeks. By month three, you'll have developed intuitions about when and how to leverage AI that will serve you for years.
Bottom line: These aren't cutting-edge techniques. They're practical fundamentals that work consistently across different roles and industries. The question isn't whether AI will transform knowledge work—it's whether you'll lead that transformation or react to it.
¹ Teams accounts start at $25/month for ChatGPT and $30/month for Claude but include collaboration features essential for companies.
² Perplexity Pro runs $20/month with access to GPT-4, Claude, and real-time search. Worth it if you do heavy research work.
For more practical insights on AI tools and development productivity, subscribe to HyperDev. I test emerging technologies so you can focus on what actually delivers value.