I use just about every major AI assistant that’s out there: ChatGPT, Gemini (including Nano Banana), Siri, Claude, Microsoft Copilot, Perplexity, Grok, plus a few iOS-only tools. For the ones I use a lot, I pay for the full version; for the occasional ones – like Claude – I stick with the free tier.

Within CGNET, people gravitate toward different tools, but most start with Copilot because it’s built into Word and Outlook and feels easy and familiar.

I’ve developed my own workflow. I usually start with Perplexity for research. I’ll write a page or so of what I’m trying to say, then ask Perplexity to turn it into a more formal report or paper. After that, I run it through ChatGPT, where I’ll ask it to make the writing edgier, funnier, or more conversational – while also cleaning up grammar, spelling, and all the little tense and phrasing problems that creep in.

When I describe this to people, they often say, “Oh, I just use Copilot. I don’t need all those others.” And that’s probably fine for many use cases. But honestly, I don’t think you get the full value of what these tools can do if you stick with only one.

Management, of course, would love a single AI that can do everything – write, research, code, summarize meetings, and run the business. It’s an understandable instinct.

It’s also about as realistic as waiting for pigs to fly.

You wouldn’t ask one employee to run finance, design your brand, and manage your network. But that’s what leaders expect from a single AI model. The payoff? Average work, wasted time, and credibility leaking out one bad answer at a time.

The truth is simpler and more powerful: Different AI systems are built for different jobs.

Some are trained for live web search and citation accuracy. Some are designed to work inside your private data – email, documents, calendars, and chat. Others are optimized for software development, debugging, and long technical contexts.

Treat them as interchangeable and you get weaker results. Treat them as specialists and you get leverage – the kind your competitors don’t.

Why One-Size-Fits-All Is a Trap

There are three big reasons the “one AI for everything” idea breaks down.

1. Different training, different superpowers:  An AI optimized for fast conversational answers is rarely the best at grounded research. A system trained to synthesize citations may not understand your internal Slack threads. A coding model thinks very differently from a writing model. These differences are not bugs – they’re the point.

2. What an AI can see matters as much as how smart it is:  An AI’s real power comes from its context. Microsoft Copilot plugged into your Microsoft 365 tenant can see your emails, files, and meetings. Perplexity can see the live web. Claude inside your IDE (Integrated Development Environment) can see your codebase. Each one operates in a different universe.

3. Task-specific optimization wins:  A model that shines at conversational writing may stumble on structured research. A reasoning-heavy model may be awkward at creative formatting. Acknowledging this isn’t weakness – it’s how you get better results.

The Stack: Which AI for What

Here’s a practical way to think about routing your work.

Research with citations → Perplexity (or ChatGPT)

When you need to know what’s actually happening in the world – and you need to prove it – Perplexity is hard to beat. It’s built as an answer engine, not just a chatbot. It runs live web searches, pulls from multiple sources, and shows inline citations you can click and verify.

Use it for:

“What are the current options for X?”

“Compare Y and Z.”

“Summarize the state of the art in [topic].”

“Draft a memo with references.”

I personally lean toward Perplexity for this kind of work. It’s a bit deeper and more rigorous about sourcing. That said, ChatGPT competes in the same tier – it just brings different stylistic strengths and ecosystem advantages. The gap isn’t huge, but when accuracy and citations matter, Perplexity usually gets the nod.

Organization-internal questions → Microsoft 365 Copilot

If the question is “What’s going on inside my organization?” Copilot is in a different category entirely. It can search your Outlook, Teams, SharePoint, OneDrive, and calendar – based on your permissions.

Use it for:

• “Show me all documents about feature X”
• “Summarize meetings where Y was discussed”
• “List emails about rollout risks in the last month”
• “What has the team said about Z this quarter?”

No external AI can do this, no matter how hyped it is. But that doesn’t make Copilot great at web research or academic writing – it’s a specialist in internal content.

Code generation & debugging → Code-tuned models

For real development work, models built for coding (Claude in dev tools, GitHub Copilot, and similar) consistently outperform general chatbots. They’re optimized for:

• Long-context code understanding
• Refactoring
• Test generation
• Step-by-step debugging
• API and library comparisons

The winning pattern: use a coding assistant inside your IDE, and keep a research AI (Perplexity, ChatGPT, Gemini) open in a browser tab when you need to look things up.

Writing, editing, and polish → Any top-tier model

For brainstorming, outlining, rewriting, and grammar, almost any modern LLM works well. What matters most is where it lives. If you write in Word, Copilot is convenient. If you live in Google Docs, Gemini fits naturally. Pair it with a research AI in a second tab when you need facts or inspiration.  My go-to choice, however, is ChatGPT.

Long-form research & reports → Deep research models

When you need a serious, structured, citation-rich report, use tools built for it: Perplexity Deep Research or Google NotebookLM paired with Gemini. These systems are designed to:

• Build comprehensive outlines
• Research multiple angles
• Gather and cite sources
• Organize everything into a coherent document

The Power Move: Use Them in Parallel

Most days you actually need all three: your email, your documents, and the web.

The pragmatic workflow:

1. Start with the AI embedded where you’re working (Copilot in Outlook, Gemini in Docs).
2. Open Perplexity or ChatGPT in a second tab for research.
3. Use your coding assistant when you’re in code.
4. Recompose in your primary editor.

This beats forcing one system to do everything it’s bad at.

So… Which AI Is “Best”?

That question misses the point – and wastes time.

• Best for research? Perplexity, because of its answer-engine design and citations.
• Best for org data? Microsoft Copilot, because only it can see inside your tenant.
• Best for code? Code-tuned specialists.
• Best all-arounder? Gemini – solid across the board, deeply integrated into Google Workspace.
• Best for you? The one closest to where you’re already working.

Don’t wait for one AI that does everything

Don’t wait for one AI that does everything. That’s a flying pig.

We’ve arrived in the Age of Aquarius (everyone sing along!): a team of specialist AIs, each doing what it does best.

The era of one general-purpose AI handling all your cognitive work either ended, or never really existed in the first place. The era of AI as a bench of specialists is here.

So the next time someone asks, “Which AI should I use?” the real answer is: “What are you trying to do?”

Pick the right tools.

Let each do what it does best.

Stop waiting for pigs to fly.

This is the age of specialists.

Want to learn more? AI has been a subject of my writing for several years, and CGNET has offered AI user training and implementation for both large and small scale organizations.   I would love to answer your questions! Please check out our website or drop me a line at g.*******@***et.com.