We're back.

I took a few weeks to step away from the noise and focus on what's actually working in AI workflows versus what's just filling up LinkedIn feeds. What I found changes how I'm approaching everything.

Most AI thought leadership right now is philosophical speculation dressed up as business strategy. I'm more interested in the practical reality of how AI actually impacts teams and workflows today.

Starting today, you'll see two newsletters per week as I share the breakthrough workflows I've been testing. Coming soon: Midjourney techniques that are transforming creative teams, advanced writing processes that compound your best ideas, and where agent platforms like n8n, Gumloop and Lindy actually deliver value.

Today's focus: deep research workflows that supercharge teams and preview the agentic future.

The leaders building real competitive advantage aren't just prompting AI for quick answers. They're developing systematic research processes that compound expertise and reshape how knowledge moves through their organizations.

This isn't just about better workflows. It's a thought exercise for what's coming. The deep research capabilities you build today mirror what agents will automate tomorrow.

If you're still thinking of AI as a productivity hack, what follows will shift your perspective toward building the foundation for autonomous knowledge work.

The Moment I Stopped Googling

I used to burn hours clicking through shallow search results, skimming recycled articles just to cobble together basic context for strategy documents.

Research wasn't intellectually challenging. It was just slow. And that killed momentum for everything I was trying to build.

The breakthrough came watching OpenAI's Deep Research tool at work. That loading bar ticking by while it went off to fetch structured, source-backed information from across the web.

There's something psychologically powerful about outsourcing that legwork—suddenly I had an invisible research assistant running in the background.

This wasn't prompt-and-return like basic ChatGPT. Deep research tools interrogate problems, gather actual sources, and come back with current, relevant context. Like having your own mini-Wikipedia delivered on demand, complete with citations instead of AI hallucinations.

The first real test: I wanted to know if those "AI agents run the world now" headlines were legit or just hype. Instead of spending my afternoon diving down rabbit holes, I fed the question to deep research and walked away.

Twenty minutes later, I had real company examples, recent implementations, specific use cases, and nuanced analysis I never would've found skimming Google's top hits. Suddenly I could back my writing with genuine, current evidence instead of running on gut instinct.

Here's what changed everything: I don't just run deep research once. I feed results back into further prompts, chaining them to create layered insight. What used to take a week now happens in a day.

Insights build on insights, compounding fast.

Why This Is Really Agent Training

Building deep research workflows isn't just about better information gathering. You're training for a fundamentally different way of working.

Every time you prompt deep research and watch it disappear to fetch, analyze, and package information, you're practicing agentic work. Learning the muscle of giving autonomy to machines—not just using a tool, but delegating an entire process.

Most people don't realize it yet, but prime model agentic workflows are rolled out piece by piece. Today's research tools are early glimpses of what's coming: AI that runs its own mini-processes, makes decisions, and returns structured output without constant hand-holding.

The manual chaining we do now will soon happen automatically. True agentic AI will run dozens of workflows simultaneously, linking them together without you touching a thing.

The real shift isn't technological. It's psychological.

Organizations starting today will adapt smoothly when agents handle entire knowledge cycles autonomously. The rest risk being blindsided.

Think about your future team composition. Digital teammates will likely outnumber human ones within a few years. The workflows you delegate today, the experiments you run, the comfort you build with AI autonomy. These habits will pay off exponentially as the technology accelerates.

You're not just learning new tools. You're learning how to scale up with AI instead of being replaced by it.

What Deep Research Actually Is

Let's be clear about what we're talking about.

Every major AI platform—ChatGPT, Claude, Perplexity, Google Gemini—now offers some version of "deep research" or "extensive research." The names vary, but the function is consistent.

This isn't just search with extra steps. You ask a complex question, and the AI scours the web, gathers real sources, distills key findings, and packages everything with citations. Like having a research assistant who actually knows how to synthesize information instead of just dumping links on your desk.

In practice, it's as simple as hitting a button in your chat window.

I use ChatGPT's deep research almost exclusively now—the O3 model is my go-to. But every major tool performs at a genuinely useful level. Some are faster, some more readable, some handle certain queries better.

Find the feature in your preferred platform and test it. They all work. The differences matter less than building the habit of actually using them.

My Workflow: Experience First, Research Second

My process doesn't start with research. That's backwards.

I usually begin by answering questions from personal experience, intuition, and years of pattern recognition. That's my 70% foundation—the insights I've earned through actually doing the work.

As I write or build a strategy, weak spots reveal themselves. Statements that feel thin. Claims I can't back up. Areas where my confidence dips. That's when I deploy deep research—not before.

I don't shotgun research across everything. I zero in on specific gaps where I need reinforcement.

Think of it like building a wall. My personal insights are the bricks and deep research is the mortar that binds them into something solid and defensible. Sometimes it flips: research becomes the brick, experience becomes the glue.

Pro tip: Pay for premium. Free models rarely deliver the depth or velocity you need for real work.

Here's where most people mess up: they treat this like a linear process. Fetch research, paste results, move on.

Wrong.

It's a feedback loop. I take what deep research surfaces and use it to prompt new questions, challenge assumptions, validate hunches. Each pass sharpens the insight. The real edge comes from chaining research and using each output as fuel for the next question.

You know you're done when your logic feels structurally sound. When every weak point has supporting evidence or clear, defendable rationale. You could always go deeper, but you've reached "enough" when the foundation is strong, not just patched.

Most people stop too early. They get one good answer our of the LLM agent and call it research. The competitive advantage lives in that second and third layer of inquiry.

Case Study: Cutting Through Agent Hype

As I mentioned, I recently kept seeing headlines about "AI agents running businesses end-to-end." The kind of breathless coverage that makes you wonder if you're missing something or if it's just vendor hype dressed up as breakthrough news.

I wanted the real story.

Instead of falling down the Google rabbit hole, I fired up deep research in ChatGPT and asked:

Find recent, real-world examples from the past 6 months of companies actually using AI agents to automate substantial business operations—not just pilots or isolated use-cases, but actual deployed workflows. What did they automate, what tools did they use, and what results did they see?

Twenty minutes later, I had concrete case studies with sources:

A SaaS company using AI agents to handle 80% of customer onboarding without human intervention—complete with implementation details and success metrics from the CEO's own interview.

A logistics startup chaining agents together to automate freight booking—with actual cost and time savings reported in their Series A announcement.

A mid-sized marketing agency using agents for campaign setup, asset distribution, and reporting—sourced from a detailed case study they published, not just a press release.

Each example came with links to original documentation, interviews, and data. Not third-party summaries or recycled takes from the same three industry blogs.

This wasn't just faster than my old approach—it was better. Instead of managing browser tabs and cross-referencing sources, I got structured context and recent data I could actually trust.

The result? My next article and the advice I gave to people was grounded in real implementations, not industry noise. That's the difference between research that compounds your expertise and research that just fills time.

Start Small, Build the Habit

Like I said, every major LLM now has a "research" or "deep research" button. If you're not using it, you're leaving value on the table.

There's no secret handshake. It's sitting right in front of you.

The real move isn't finding the perfect use case—it's using deep research constantly. If you're not maxing out your monthly usage limits, you're not exploring the edge of what's possible.

Reps are everything.

Top operators use deep research five, six, ten times a day. They're not saving it for "important" projects. They're building muscle memory on everything from competitive analysis to market research to fact-checking their own assumptions.

Push the tool until you hit its limits. That's how you learn both strengths and weaknesses before it matters. Encourage your team to do the same. This isn't about individual productivity, it's about leveling up the whole organization.

Want to know if your deep research result is actually better than Google? Gut-check it the way you would if a team member handed you their work. Does it surface something you couldn't get from a lazy search, or is it just rehashed noise from the same top-ranked articles?

You'll know.

There's no special workflow or secret trick here. It's about building confidence through daily use. The tools are still cheap—now's the time to practice before pricing or access changes.

These repetitions aren't just about immediate productivity. They're training you and your team for when these tools become true agents, chaining complex workflows together automatically.

You don't want to be the leader scrambling to catch up when that future hits.

Build the habit now. Make it survival later.

The Copy-Paste Trap

The biggest mistake I see with deep research tools? People treat them like a copy-paste shortcut.

It's the modern version of submitting a Wikipedia article as your own work. Just as lazy. Just as obvious.

Dumping raw AI output into a report or sharing it directly with your team signals complete lack of respect for their time and intelligence. If you're not distilling and re-contextualizing insights, you're not leading—you're mailing it in.

Clients and teammates can smell bot-generated copy from across the room. Generic phrasing, surface-level analysis, insights that sound impressive but lack real depth. You risk your reputation fast when you rely on unfiltered output as your final product.

People want your perspective, not regurgitated facts.

Here's what too many leaders miss: AI research is a springboard, not the finish line. The real competitive advantage comes from using these tools to jumpstart your process, then layering your own expertise, narrative, and context on top.

You still have to do the work.

Deep research should cut the time you spend fetching information, but you can't skip synthesis. You can't skip connecting the dots. You can't skip making it yours. The moment you try to shortcut those steps, you lose everything that makes your output valuable.

We’ll start seeing many teams lose credibility because they got lazy with the "glue"—the human layer that turns research into insight, insight into strategy, strategy into action.

You can move faster with these tools. But you can't skip the parts where your judgment, voice, and accountability actually matter.

That's not just a workflow mistake. It's a leadership failure.

Scout Report

Here are two small model updates that have me excited this week:

🔗 Claude Projects Handle 10x More Content → Anthropic just made Claude way more usable for heavy workflows. Projects now support 10 times more content, and the new voice mode puts it closer to the top LLMs. → Read more

🔗 ChatGPT’s Advanced Voice Mode Just Got an Upgrade → The latest release notes (June 7) detail upgraded intonation, realistic cadence, better emotional expressiveness (sarcasm, empathy), and even built‑in translation. → Read more

Forward Signal

If you're not already using deep research workflows, I hope this opened your eyes to the competitive advantage sitting right in front of you. Start incorporating these processes now.

But here's the bigger picture.

This is just one step toward true agentic AI. What you're learning today—sending AI off to research and return with structured insights—is training you for a fundamentally different way of working.

Eventually, autonomous AI will chain these processes together automatically. Research connects to analysis, analysis feeds strategy, strategy informs execution. All without your prompting.

The framework matters: get comfortable with AI agents doing meaningful work and coming back with genuine value. When we start chaining all of this together, that's where you'll see the real magic in scaling organizations and reshaping what teams look like.

Start with deep research. Think like an agent architect.

Next up later this week: the innovation coming out of Midjourney that's transforming creative output for design teams. Some of these capabilities will surprise you.

If this edition of Frontier Notes for AI Leaders gave you great insight, please share it with someone who leads through change.

Share Frontier Notes for AI Leaders

And if you haven’t subscribed or upgraded to premium yet, we’re delivering frameworks and systems that make AI practical every week—sign up below.

Subscribe now

Frontier Notes

Deep Research Is the New AI Operator's Edge