The Senior Manager’s Guide to AI: From Hype to Real Impact
Lessons from a year of experimenting at work
AI is everywhere these days.
LinkedIn is full of prompts. Clickbait articles promise 2 hours saved a day. Newspapers call any restructuring “AI layoffs” (a few months ago it was “the death of the middle manager” - how times change!).
At work - AI is everywhere too. Each day there is another launch announcement, another hackathon, another AI training.
Yet for managers, little has changed. Meetings pile up. Decisions still need to be made, people hired, projects unblocked. IT spent six months rolling out “new apps” that look identical to the old apps except for an “AI assistant” button. You tried it: the email summary was junk, the meeting tool isn’t enabled in your region, Copilot features are grayed out. Not a minute saved.
AI’s rise is exciting, but also messy and frustrating. The guitar-around-the-campfire future isn’t here yet. Maybe in 2050. For now, grab a glass of water and dial into your next meeting.
If this sounds familiar, you are not alone. New technology always takes time to pay off. The FOMO is real, but most rabbit holes lead nowhere.
So I decided to put some structure to the chaos - after all, this is my speciality - and share how managers and organizations will go through the AI journey. I am observing this happening at Amazon now. If you are in a senior role in a big corp, leading a team, I hope this guide will be useful. Let me know your thoughts in the comments (or hit reply in the email)!
I have called my framework SPARK:
Start playing
Partner for decision-making
Activate your team
Replicate across org
Kindle the AI-first transformation
As a manager, you start playing with AI here and there, but you find it more useful outside of work context. At work you may use it every day, but see no real time saving. In fact, AI comes on top of the regular workload.
You level up when you find a “killer use case”. It enables you to deadlift 10x more or frees up a big chunk of your time. At this moment, the positive feedback loop starts building - you start seeing AI as practically useful, for you, now.
Next level of mastery is activating your team - people may be trying out various use cases on their own, but bringing them together and helping them accelerate the learning is something you can facilitate.
Fourth level is spreading best practices org-wide where cross-organisational impact and knowledge start to be activated. (Dare I say “synergised”? I don’t).
Final level is AI-first transformation, where roles, processes and operating models get redesigned around AI.
This post covers the first two levels. The rest will follow in Part 2. Subscribe so you don’t miss it. I also have an Easter egg at the end of Part 2.
Level 1: Start Playing
This is where most managers get stuck. You try the obvious use cases: summarising documents, drafting slides, rewriting emails, capturing meeting notes. The results are… meh.
There are three reasons:
The tools aren’t great yet. Many internal bots still run on older models. Using them feels like cutting tomatoes with a butter knife. The use of external tools is restricted, so you can only use them for generic prompts like “Give me a brief refresher on theory of constraints” or “Help me draft a job description for a Principal Product Manager role”.
You’re already skilled. Drafting an email is trivial for you but transformational for a junior colleague. For you, the AI rewrite usually takes longer than writing from scratch.
Your prompting is weak. Ask any tech team which AI skill to build and they’ll say: prompting. Better prompts mean faster progress to Level 2.
The hallmark of Level 1: you’re using AI often but seeing no real productivity gains. That’s normal. AI is not magical.
How to level up:
Take a free advanced prompting course (I liked this one).
Swap tricks with colleagues. Sit down with someone more advanced and ask them to show you what they do.
Browse 30 days of GPT by Hilary Gridley to get some ideas.
Keep playing. Block 30 minutes a day for AI. Repurpose part of a strategy block to do some training. Small daily reps compound — and eventually lead you to your first “killer use case.”
Level 2: Partner for Decision-Making
For individual contributors, the “killer use case” in AI is productivity. For managers, it isn’t. When was the last time you did a repetitive task? Filing expenses doesn’t count 🙂. As a manager, you have already eliminated, automated or delegated most repetitive tasks.
I bet that the “killer use case” for managers and executives is going to be AI as a decision-making partner.
Management is lonely. Imagine having a trusted companion available 24/7 - one that can play board member, executive coach, McKinsey partner. You can talk - not type, not prompt - anytime. That’s not science fiction. You can build one today.
Yes, confidentiality matters. A small startup CEO can experiment freely on public tools. Just follow this guide from
and start building. A Fortune 500 executive must stay within policy and guardrails. Your company probably hasn’t built an internal tool like this yet. But you can. All you need is a lawyer and a developer. It can be ready faster than you think.And you can still do a lot on public LLMs. As managers, we face common situations as we grow our people.
Try this conversational prompt:
“What is the essence of Matt Mochary’s approach on developing team culture and employee performance?”
Follow up: “Double click on clarity of roles and expectations. How should it be tailored relative to employee seniority?”
Then: “Create 1-pager guideline I can give to a new manager on my team on managing a junior employee. Reference 3 authoritative sources with links that they can explore on their own in addition to 1-pager.”
Stage a debate between Bill Campbell (“people first”) and Jack Welch (“up-or-out”).
Ask Kim Scott and Marcus Buckingham for contrasting feedback styles.
People are creating AI personas -
, ex-Amazon VP, has a good one and his advice is applicable for large companies.The hallmark of Level 2: you have at least one AI persona you trust to improve your decisions. Your default mode of thinking now evolves using it to widen your perspective, stress-test the ideas and accelerate the synthesis. You spend less energy on grunt work and more on judgement.
How to level up:
Share decision-making prompts not just outcomes. When a prompt saves you an hour, pass it to your team. Team them to get advice from Andy Grove and other management gurus.
Build a personal prompt library. Stop reinventing each time.
Run a team AI session. Collect best use cases in a shared doc. Collective learning compounds.
Add AI to existing rituals. Add “AI wins & learnings” as agenda item to the weekly meeting, setup a regular TGIF demo, or a “Friday AI hour.”
Clarify safe use. People often don’t know what’s allowed. Set guardrails and lower the barrier to experimentation.
Interlude: a note on productivity
I said AI for managers isn’t about productivity, but it can help with IC-style work.
Some time ago I wanted to write a proposal on a program sponsor role. A very simple prompt gave me a great first draft and removed the “blank page” syndrome. It turned a six-hour chore into a two-hour job. I would struggle to find these six hours in the manager schedule, stretching this task for weeks. Two hours are much easier to find.
It also helps with people work: summarising interview notes for debrief, or sharpening a performance summary. Precise words matter in such cases and AI helps improve the quality of my feedback. I can also point it at all of the past examples of my interview feedback and tell it: “This is what good looks like. Here are my raw notes, here is my decision, now turn them into the feedback summary”. All your past artifacts become your secret weapon.
One more reflection on the personal productivity front. I haven’t figured out a way to apply this to work, but it is so good I wanted to share.
We all have certain “Eat the frog” tasks that need doing, but we keep procrastinating on them because they are complicated and bureaucratic.
Recently I needed to submit an amendment to a tax declaration. Figuring this out in French across multiple websites can be a nightmare. So normally I would just delay this until the last responsible moment.
This time, AI gave me a summary of steps. In fact, it hallucinated a feature on a government website that didn’t exist. Their IT department was quite puzzled by my detailed request to activate it 😀. At the same time, it gave me an alternative path that worked. I got it done in one sitting. I don’t dread such tasks anymore.
What’s next
That’s it for today. In Part 2 I will cover the three remaining levels of SPARK: Activate your team, Replicate across org, and Kindle the AI-first transformation.
Subscribe so you don’t miss it.
Till next time and happy AI exploring! 🙂
Additional resources, I found useful:
Ethan Mollick’s Quick Start Guide (June 2025, free)
Hilary Gridley’s How to Become a SuperManager with AI (Nov 2024, partly paywalled, even free part is useful)
Torsten Walbaum’s Using AI as a Thought Partner (July 2025, free) In depth guide how to 10x your thinking with AI. This is gold.
If you know other great guides for managers and executives, please share.