I recently curated the perfect Pandora station: Bossa Nova radio. It started out okay. João Gilberto, Stan Getz, a little Antônio Carlos Jobim. Then the Eagles showed up. Then Kenny G. doing “The Girl from Ipanema”.
I hit skip. Then thumbs down. A few songs later, more classic rock. Skip again. Then finally, a rhythm clicked. It wasn’t perfect, but I could leave it on all day without interference.
That’s what training AI feels like. You don’t program it. You shape it. You notice the outliers. You tune the feedback loop. You say yes to nuance and no to nonsense. And if you don’t? You get entropy: clutter, drift, and results that feel almost right but totally wrong.
Which brings us to your inbox.
You’ve probably seen the updates: Microsoft 365 Copilot is embedding AI into Outlook, Teams, Excel, and Word. Gmail’s Duet AI and Slack’s summaries are also gaining traction in preview environments. But here’s the reality: unless you’ve configured, licensed, and enabled these features, they’re not doing anything for you.
For most organizations, the AI inbox isn’t replacing your assistant. It’s waiting to be trained.
The New Inbox Isn’t Just Email
These tools are capable of more than autocomplete. Microsoft’s semantic index makes Copilot aware of content across files, chats, and history. Gmail’s “Help Me Write” pulls context from threads and previous interactions.
But none of this works by default. IT needs to enable it. Admins need to configure it. Teams need to guide it. Until then, it’s a feature with potential, not a function with results.
But Here’s What AI Still Can’t Do
Microsoft’s new Dragon Copilot is already being tested in healthcare. It drafts clinical summaries, generates referral letters, and creates after-visit notes for physicians (especially residents and private practice docs) who are drowning in administrative tasks. But early users say it’s not plug-and-play. Success depends on training, feedback, and constant iteration.
Meanwhile, in logistics, a startup called Augment is deploying its AI assistant “Augie” to respond to routine shipping inquiries and automate email flows. Companies like Arrive Logistics and Armstrong Transport Group are integrating it with their customer ops teams. But again, they aren’t letting it run solo. Every response still gets a once-over.
The lesson? AI helps when you shape it. Not when you just switch it on.
Who’s Actually Using It and What They’ve Learned
Some firms will try to ignore this shift. They’ll double down on what worked before and hope clients don’t notice the difference. But others are doing something smarter.
They’re not just learning how to use AI. They’re embedding it into their delivery. They’re turning it loose in the research phase, using it to challenge assumptions in real time, and relying on it to pressure-test their own thinking before it gets anywhere near a client.
I’ve watched teams with offshore analysts outpace more expensive domestic groups by combining skilled judgment with smart automation. And I’ve seen business units shift from reactive to proactive, not because they hired someone new, but because they deployed the right AI with the right thinking behind it.
It’s like handing a Formula One driver a car that suddenly goes twice as fast. If they know how to handle it, they’re not just faster : they’re more strategic, more precise, and more dangerous to the competition. That’s the advantage.
Because when you’ve spent decades helping shape platforms, building IP, and delivering major enterprise solutions around the world, you already know how to win. AI doesn’t replace that. It accelerates it. It lets you skip the fluff and focus on what really moves the needle.
That’s the real differentiator. It’s not about automating the whole job. It’s about shortening the distance from data to decision, without sacrificing insight.
AI and the Danger of Drift
Your inbox isn’t just busy. It’s personal. When AI drafts messages or responds on your behalf, it learns from what you let slide.
We’ve seen real cases of entropy: a Copilot that keeps promoting old templates, a Slack assistant that pushes irrelevant updates, a Gmail reply generator that forgets the client’s tone.
The common factor? No feedback. No editor. No ownership.
AI gets better with attention. And worse without it.
Practical Steps to Make AI Work for You
If you want the AI inbox to actually help:
- Choose one workflow: like customer updates or internal recaps.
- Assign a human to train the AI: someone who knows what “right” looks like.
- Give feedback constantly: thumbs up, thumbs down, rewrite.
- Set expectations: this is not fire-and-forget. It’s curate-and-refine.
Same as my Bossa Nova station. Expect to dedicate the time to get the results you want. You didn’t get magic on day one. You taught it what mattered. One skip at a time.
Final Thoughts
The AI inbox doesn’t replace your judgment, but it, at best, responds to it.
If you’re not tuning it, you’re just hoping it stays useful.
So take a moment. Shape it now. And by the time it matters most, it’ll sound like you want it to.
Just minus the Eagles.
