Barry O'Reilly's new book Artificial Organizations lands on a question most leaders are quietly avoiding. If you're using it well, AI takes the busywork off your plate, but what do you actually do with the extra time?
His answer is that you use it to lead with better judgement. You sense what's happening in your organisation, think clearly about it, decide quickly, and act with confidence. He calls AI a judgement support system: the machine handles the synthesis, the human handles what matters. Or as he puts it: "AI isn't replacing leaders. It's exposing them."
I love that framing. But there's a follow-up question hiding inside it that most of us aren't answering. If we're going to spend reclaimed time on better judgement, what's the most valuable thing we can apply that judgement to?
The data suggests we're failing this question badly.
BCG's most recent AI at Work study found that around half of employees using generative AI save more than an hour a day. When asked what they did with the time, only 39% said they spent it on strategic work. The rest went into producing more output or finishing the day earlier. Microsoft's Work Trend Index calls the result the "infinite workday." The average knowledge worker is interrupted every two minutes, gets 117 emails and 153 Teams messages a day, and over half of their meetings happen ad hoc. Microsoft's own warning to leaders is blunt: bolt AI onto the current setup and you just accelerate a broken system.
It gets worse when you look at the agent side. MIT's NANDA initiative found that 95% of enterprise GenAI pilots deliver no measurable P&L impact. Gartner predicts over 40% of agentic AI projects will be cancelled by the end of 2027, mostly due to unclear business value and inadequate risk controls.
These aren't technology failures. They are clarity failures. We're handing AI tools to organisations that can't articulate what they're trying to achieve, and asking those tools to make them better at it.
The leader's job isn't to "do strategy" once a year. It's to maintain it.
So when Barry says AI gives you time back to lead with judgement, the obvious follow-up is: judgement about what?
The most valuable thing you can do with that reclaimed time is be more present with your team and put the work into your Decision Stack. Update it. Maintain it. And above all, communicate it.
That distinction matters. Most leaders treat strategy as a thing you do at an offsite once a year. The leadership team disappears for two days, comes back with a deck, presents it once on Monday morning, then everyone goes back to whatever they were doing. The Decision Stack gets the same treatment in most organisations. You ladder vision to strategy to objectives once, slide-ify it, assume the work is done.
It isn't. A Harvard Business Review study found that 95% of employees can't articulate their own company's strategy. That isn't because employees aren't paying attention. It's because their leadership team isn't repeating it. A strategy that lives in a doc, an offsite recap, or one all-hands deck from January doesn't survive contact with the daily decisions everyone has to make.
The real leadership work is the maintenance and the repetition. Updating the Stack as the business changes. Walking through it in your one-on-ones. Pressure-testing it with your peers. Making sure every team can trace what they're working on this week back to the vision and the strategy. And then doing it all again next week, because the new joiner missed it and last week's debate exposed a principle you hadn't articulated yet.
This is exactly the work AI is freeing you up to do.
Your clarity is the condition for your team's speed
Here's where the math starts to compound. When you spend your reclaimed time being present with your team and communicating the Stack, you're not just making your own decisions better. You're creating the conditions in which everyone below you can use their reclaimed AI time well.
A team that understands the strategy doesn't have to relitigate it every week. A team that knows the principles can make calls at the edges without escalating. A team that can trace its work back to the vision doesn't waste its reclaimed AI hours on the wrong problem.
The flywheel reads like this:
- You get an hour back from AI.
- You spend it being present with your team, sharpening and communicating the Stack.
- Your team gets clarity on what matters, what doesn't, and how to choose.
- Your team gets hours back from their own AI use.
- They spend those hours making better decisions, faster, all towards the organisation's goals.
Skip the first step and the flywheel runs in reverse. Your team's reclaimed hours go into producing more of the wrong thing at speed. AI becomes an output multiplier rather than a decision multiplier.
And the agent fleet you're so excited about needs exactly the same clarity. Anthropic's research on multi-agent systems shows that orchestrator agents need explicit strategic context to coordinate the rest of the fleet: what matters, what doesn't, how to allocate effort, when to go deep versus wide. The same context your team needs, in slightly more structured form, is the spec your agents need. A human team can muddle through misalignment with hallway conversations and good intentions. A fleet of AI agents will scale your misalignment at machine speed and hand you back a confident, well-formatted mess.
What this actually looks like
Barry's case studies in Artificial Organizations are instructive. American Airlines didn't go from two software releases a month to forty in three weeks because they bought better AI. They reorganised around clearer decisions and shorter feedback loops. Amazon's "decision velocity" works because Jeff Bezos was clear from day one that most decisions are reversible and don't need consensus. Skyscanner's experimentation engine works because their teams know what they're trying to learn before they run the experiment.
In every case, AI added speed. But strategic direction created velocity. And alignment, bringing the whole organisation along the journey, created momentum.
So here's the challenge. The next time AI hands you back an hour, resist the urge to fill it with more meetings or another sweep through your inbox. Spend it on the Stack.
Spend it in your one-on-ones, walking through how your team's current work connects up to the strategy. If you can't trace it cleanly, the gap is yours to close.
Spend it pressure-testing whether your strategy is actually making choices. If it's "grow 25%," you don't have a strategy. You have a target.
Spend it writing down the principle behind a decision your team keeps relitigating. "Speed even over perfection." "Self-service even over white-glove support." A principle you can write on a postcard saves you a hundred hours of debate later, and it's exactly the kind of artifact your teams and your AI agents need to make consistent calls without escalating.
Or do the most uncomfortable exercise I know. Get five of your executives in a room and have each of them, on their own, write down what your company's strategy is. Then read the answers out loud. The look on their faces when they hear five different stories will do more for your organisation than another quarter of OKRs.
This is the work AI is freeing you up to do. The fires don't go away. AI just makes them smaller and easier to ignore. So when the time appears, use it for the thing only a human leader can do. Be present with your team. Maintain the Stack. Communicate it again, and again, and again.
That's the higher-value work Barry is pointing at. Everything else is just acceleration.
Sources and further reading
- Barry O'Reilly, Artificial Organizations
- BCG, AI at Work 2025: Momentum Builds, but Gaps Remain
- Microsoft, Breaking Down the Infinite Workday
- MIT NANDA, State of AI in Business 2025
- Gartner, Over 40% of Agentic AI Projects Will Be Canceled by End of 2027
- Anthropic, Multi-Agent Research System
