Imagine being three coffees deep into your morning and then realising your calendar had rearranged itself.
Not asked you to approve changes. Not sent you polite notifications. Just… done it. The AI agent had spotted a conflict, found a better slot, updated three people, and left you a note: “Moved your 2pm to Thursday. You needed prep time before the board meeting.”
It was right. And this was me last Tuesday, though I hadn’t asked it to do any of that. For about ten seconds, I felt a primal flicker of unease. Then I felt something else entirely: relief. Because I’d just got back an hour I would have spent playing email tennis, and I could use it to actually think about the board meeting instead of rushing to be prepared for it.
That’s when it hit me. Most conversation around agentic AI (software that doesn’t just answer questions but takes action) has been about speed. Do more tasks. Faster outputs. Increased efficiency. I think we could be having the wrong conversation.
The Exhaustion Economy
Here’s what nobody’s saying out loud: people aren’t necessarily inefficient. They’re exhausted. And they’re exhausted because roughly 60% of their workday is what could be called “admin theatre”, activities that look like work, feel like work, create the performance of productivity, but generate almost zero strategic value.
Scheduling meetings. Chasing approvals. Reformatting reports. Updating spreadsheets. Copying information from one system to another. Sending “just checking in” emails. Reading “just checking in” emails. It’s coordination debt. And it’s killing productivity.
A 2023 Microsoft Work Trend Index study found found that information workers spend 57% of their time communicating (meetings, email, chat) and only 43% creating. But here’s the bit that hurts: when researchers dug deeper, they found that most of that “communication” wasn’t strategic collaboration. It was administrative coordination. Moving pieces around the board.
Meanwhile, the actual strategic work ie the thinking, the innovation, the problems only humans can solve gets shoved into the margins. Early mornings. Late nights. Weekends. We’ve built an economy where our most valuable asset, human attention, is increasingly being spent on our least valuable work.
When Software Stops Suggesting and Starts Acting
For years, AI has been a brilliant assistant. Ask it a question, get an answer. Feed it data, receive insights. It’s been a tool that waits for you.
But something fundamental shifted in the last 18 months. Google’s Gemini 2.5 Computer Use can drive the browser clicking buttons, filling forms, navigating interfaces on your behalf. Windows 11 Copilot Actions can run multi-step tasks across apps with user-granted permissions. In materials science, MIT’s SpectroGen system acts as a virtual spectrometer, inferring hard-to-get spectra (e.g., X-ray) from easier scans.
These are not chatbots. They’re agents. Software that acts. And the difference is everything. A chatbot tells you how to schedule a meeting. An agent schedules the meeting. A chatbot explains how to analyse your sales pipeline. An agent analyses your sales pipeline, identifies the bottlenecks, and drafts the strategic action.
The shift from “assist” to “act” sounds incremental. It’s not. It’s a bit like the difference between a calculator and an accountant.
The Wrong Question (and the Right One)
When many organisations evaluate agentic AI, they ask: “How much faster can we do X?”
That’s an efficiency question. And it’s not wrong, exactly. It’s just small. A transformational question is: “What are we NOT doing because we’re drowning in X?” Let me give you a real example, a mid-sized professional services firm. When we audited how their senior consultants spent their time, we found that partners were spending 14 hours per week on what they called “project admin”—updating client dashboards, scheduling follow-ups, collating team inputs, formatting deliverables.
Fourteen hours. Per partner. Per week.
When I asked what they’d do with an extra 14 hours, the managing partner didn’t say “take on more clients.” She said: “I’d finally have time to think about where this firm should be in five years instead of just keeping the trains running.”
That’s the drudgery dividend. Not doing the same work faster. Reclaiming cognitive capacity for the work that actually matters.
The Attention Audit: A Framework
If you’re serious about capturing the drudgery dividend, you need to start with brutal honesty about where attention currently goes.
Here’s the framework I use with clients:
Step 1: The Time Audit
For one week, track time in three categories:
- Strategic Work: Thinking, creating, problem-solving, relationship-building that only you can do
- Necessary Coordination: Admin that’s genuinely essential for operations
- Drudgery: Repetitive coordination that creates the illusion of productivity
Most leaders are shocked. They assume they spend 60-70% of time on strategic work. The reality is usually 20-30%.
Step 2: The Action Inventory
List every recurring task in the “Drudgery” category. For each one, ask:
- Does this require my judgement, or just my action?
- Could I explain the rules for doing this to someone else in under 10 minutes?
- If this didn’t happen for a week, would anything actually break?
If it’s rules-based, teachable, and non-critical? It’s an automation candidate.
Step 3: The Liberation Plan
Calculate the dividend. If you reclaimed 40% of coordination time, what becomes possible?
For that professional services firm, it meant partners could:
- Mentor two junior consultants properly (currently an afterthought)
- Write thought leadership that actually generated inbound leads (currently non-existent)
- Develop one new service offering per quarter (currently “we’ll get to it next year”)
That’s transformation, and goes way beyond efficiency.
Step 4: Agent Design
Now you’re ready to deploy agentic AI strategically, not randomly. You’re asking it to reclaim attention, not just save time. The professional services firm started with calendar orchestration and client update automation. The agents didn’t just “do tasks faster.” They gave partners their evenings back and their strategic thinking space back.
Six months in, the firm had launched two new practice areas and increased profit per partner (and also job satisfaction) Not because they worked more. Because they worked on what mattered.
The Trust Challenge (and How to Solve It)
I’ll be honest and admit that letting software act on your behalf feels weird at first. When my calendar rearranged itself, my immediate thought wasn’t “brilliant.” It was “what if it got it wrong?” This is the trust paradox of agentic AI. We want it to free us from drudgery, but we’re terrified it’ll make mistakes that create more work.
The solution shouldn’t be to slow down adoption, instead we should look to design automation for transparency. The agents that earn trust do three things:
- Show Their Working
My calendar agent doesn’t just move meetings, it tells me why. “You have 90 minutes of prep materials to review. Moved meeting to create space.” I can see the logic. I can override it. But mostly, I trust it because I understand it.
- Start with Boundaries
Define what the agent can and cannot do. Early on, my agent could suggest meeting changes but not confirm them. Once I trusted its judgement, I expanded permissions. Start narrow. Earn trust. Expand gradually.
- Maintain Human-in-the-Loop for High Stakes
Agentic AI should handle the drudgery, not the decisions that define your business. One client uses agents to draft contract amendments but never to approve them. The agent does the tedious formatting and clause checking. The lawyer does the strategic judgement call.
Clear boundaries. Transparent reasoning. Human oversight on what matters. That’s how you capture the dividend without the risk.
What This Looks Like in Practice
Let me paint three scenarios, one business, one education, one persona, where the drudgery dividend changes everything.
Scenario 1: The Marketing Director
Currently spends 12 hours/week coordinating campaign assets, chasing designers, updating project trackers, sending status reports, scheduling reviews.
With agentic AI: Agents handle coordination. She spends those 12 hours analysing why last quarter’s campaign underperformed and designing a test-and-learn framework for next quarter. Revenue impact: measurable. Attention impact: transformational.
Scenario 2: The University Administrator
Currently spends 15 hours/week on accreditation reporting, collating data from six systems, formatting documents, chasing faculty for inputs, updating compliance trackers.
With agentic AI: Agents pull data, format reports, send automated reminders. She spends those 15 hours redesigning the student feedback system to actually improve teaching, not just tick boxes. Student satisfaction impact: significant. Administrator burnout: reduced.
Scenario 3: The Small Business Owner
Currently spends 8 hours/week on invoicing, expense tracking, client onboarding admin, and social media scheduling.
With agentic AI: Agents handle it. He spends those 8 hours building relationships with his three biggest clients and developing a new service offering. Business development impact: game-changing. Work-life balance: restored.
Notice the pattern? The dividend isn’t “do more admin.” It’s “do less admin, more strategy.”
The Multi-Agent Future (It’s Closer Than You Think)
Right now, most of us are experimenting with single agents, one tool, one task.
The next 18 months will bring multi-agent orchestration. Multiple AI agents working together, coordinating amongst themselves, handing off tasks seamlessly. Imagine: one agent monitors your project pipeline, another manages your calendar, another handles client communications. They talk to each other. When a client asks for a meeting, the communications agent checks with the calendar agent, who checks with the project agent to see what’s urgent, then proposes optimal times and sends the invite.
You? You’re writing the strategy document that actually moves the business forward. Is it science fiction? No, it’s what Google, Microsoft, and a dozen startups are building right now.
The Real Choice
Here’s an uncomfortable truth: agentic AI is coming whether you’re ready or not.
And the question won’t be whether software will start acting autonomously in your workflows. It’ll be whether you’ll use it to do more drudgery faster, or even escape drudgery entirely. Many organisations will probably choose the former. They’ll automate the busy work and immediately fill the space with… more busy work. Because that’s what organisations do. Parkinson’s Law is undefeated. But the organisations that really stride forward will choose differently.
They’ll ask: “What would we do if we had our attention back?” They’ll reclaim cognitive capacity for innovation, strategy, and the distinctly human work that AI can’t touch: creative problem-solving, relationship-building, moral judgement, and envisioning futures that don’t exist yet.
They’ll collect the drudgery dividend.
Your Action Step
Pick one task you do every week that’s pure coordination drudgery. You know the one, the thing that makes you think “I can’t believe I’m paid to do this.” Now ask: Could I teach the rules for this task to someone in under 10 minutes?
If yes, that’s your first automation candidate. Test an agentic tool. Monitor what happens. Refine the boundaries. And pay attention to what you do with the time you get back. Because that’s where the real ROI lives. Not in the tasks you stopped doing.
In the work you finally have space to start. Let me know how you go, I’m keen to follow this up with you.