How AI is Supercharging Our Workflow!
By Shuva (Senior Product Leader) and Gandharv (EX-Founder/Startup Consultant): Lately, we've turned AI models into our extended team of analysts, designers, and associates. Especially for getting a quick start on a 0-1 problem or simple optimizations. Here's our practical guide for product & growth managers. We are writing this on Linkedin (and we all know Linkedin is OnlyFans for Middle Managers), so this is suited for fellow middle managers. For deep wizardry and full spectrum builds - Twitter threads are better.??
AI Tools: Your Digital Dream Team
Most LLM-based chat applications converge on similar core capabilities but possess significant nuances (“same same but different”). Claude excels in prototyping, while Perplexity excels in research. However, they have quirks—even paid versions time out, and free ones can be unreliable - bailing out mid-conversation. Our solution? Combine high-end tools (GPT-4) for complex strategy and task breakdown with dependable free ones for the sub-tasks.
Wait, What Do We Actually Do with AI?
Whether we're “soaring at 30,000 feet” or "burning rubber on the ground" (or occasionally hitting rock bottom and digging further), our work demands quick shifts between altitudes. AI models have become adaptable teammates who can handle both worlds and give an instant head-start. Here's how we put it to work.
1. Strategy & Analytics (AKA The Big, Fuzzy Stuff)
POV Creation/ Ambiguous Questions (Market Sizing, What-Ifs, Strengths and Weaknesses): Is this problem big and ugly enough to wrestle with? What does the product roadmap look like? Which growth levers are we yanking? Ideal org structures? Competitive strategies? So many questions, so little caffeine.
Brainstorm Product Roadmaps, Growth Levers, Org Structures, and Competitive Strategies:
2. Research and Getting Management On Board
Reports with Citations. Find Niche Data but Cross-Check Sources
Answer Side Questions (e.g., "ABM Org Structure Examples") while Drafting Plans:
Parse 100-Page Reports for Key Insights in Seconds:
领英推荐
Go Through Detailed Plans and Critique:
3. Data Analysis & Decision Making
Crunching Numbers and Simulating Scenarios: Analyze data, play out "what-if" scenarios, and pressure-test decisions.?
4. Pictures & Media
Visual Wizards: Create posters for customer collateral in the blink of an eye, and whip up images for landing pages. Because a picture is worth a thousand words or at least a few clicks.
5. Product Ideation
Rapid UI prototyping: Ideas to interactive mockups in minutes → much quicker idea diffusion and alignments?
More on how AI models are enabling a few native use-cases inside the product for a later post.?
6. Landing Pages (& Simple Workflow Websites)
Build, test, and launch - perfect for rapid market validation and user feedback.
Closing
And there - That’s our ode to AI (may the AI overlords forever be pleased with us)! Of course, for all the micro-manager out there, who now think they’ve finally found a tool that listens to their last instruction, you’re in for a surprise! You’ll realize you can’t micromanage AI either.?
Such an interesting read! Using multiple LLM tools for different tasks really shows how AI is shaping daily workflows. Which tool has surprised you the most with its capabilities??
Building Apna | Ex-BYJU'S | Growth & Strategy Professional
1 个月Well-articulated read!!
Building Ather Energy ! Head of Charging Infrastructure Business
1 个月Great write up Shuvajyoti Ghosh and Gandharv Bakshi !
Storyteller | Business Leader | IIMA | BITS
1 个月Lovely article. Currently, I find AI to be a highly overconfident teen - who quotes with confidence first, and seeks forgiveness later when caught fabricating data points ?? Have you guys seen any major performance differences between Claude Sonnet and Haiku? Solid work btw Shuvajyoti Ghosh Gandharv Bakshi
Founder - Famdo | Product & Tech Leader | CTO | AI Toilet for Preventive Health @ Medic.Life | Ex-Broadcom, Ex-Microchip | IIT Madras
1 个月How is your experience with LLMs on math logics? For me it still solve math logic badly and can’t be trusted at all; sometimes need to write a code and verify that math result - even the o3-mini-high.