Everyone wants to build agents today. But does it make sense? How do you know that you really need GenAI to solve a problem? What will sustain? ..And what is just a fad?

✨ Three ways you can use Generative AI services today

1. Chat web interface: CoPilot, ChatGPT, Gemini, etc, which are instantly available and you get started quickly. Use AI in your day to day tasks.
2. Application programming Interfaces: Like OpenAI APIs, AWS Bedrock APIs; for developers and teams building products, gives you faster time to market; but keep check of cost and data sharing.
3. Direct or open-source: Run AI on your own infra, within your own network. This is expensive! Training requires compute power, but 100% secure as you own everything!

✨ Two ways you can launch your Generative AI app

1. For your organization internal use. Adopt existing products, use in every day tasks and implement specialised business solutions for internal products.
2. For external use, identify customer problems and solve them.

✨ What are the business drivers? But what are the business drivers?

• Automation: Automate manual and error prone tasks (Not robotic process automation)
• Augmentation: A helpful assistant, or co-worker, you get more done and level up your expertise
• Differentiation: Advisory, optimization, etc

Your GenAI product should either save time and money, or earn money, skills and expertise!

✨ What will (probably) not sustain in the long run

1. Generic wrappers on large language models
2. AI for just about anything without any unique value proposition
3. Anything that can be easily done on ChatGPT or Gemini by writing a prompt
4. Anything the big players may add as a feature

✨ What will (probably) sustain in the long run

1. Boring workflows that will indeed save money (enterprises make most of their money from boring stuff)
2. Domain specific workflows that are error-prone and best done with AI
3. Anything that needs specialized AI use, what ChatGPT or Gemini can’t do or won’t do

✨ Ask yourself…

Can AI help solve this problem Faster Cheaper Better ..and at scale

1. Map the problem
2. Identify pain points worth solving
3. Find where AI can help

✨ How AI Agents can help?

• Tasks human can do, but never find time for
• Tasks humans aren’t good at
• Tasks which should be best left to machines
• Tasks which humans simply can’t do