The Tasks You Should Stop Doing Yourself (And How to Transfer Them)
There is a type of work that experienced professionals are particularly bad at letting go of.
It is not the hard work — the judgment calls, the complex analysis, the client conversations that require their full attention. Most professionals have learned to protect that work.
It is the medium-difficulty work. The tasks that take real skill but not necessarily their level of skill. The tasks where they feel, vaguely, that they should be the one doing it — because they have always been the one doing it, because they are faster than explaining it to someone else, because they care about the quality.
That category is where most senior professionals spend hours every week that they could be spending differently. And it is the category where AI creates the biggest practical opportunity right now.
The Trap of Competence
The most experienced professionals are often the most reluctant to transfer work. They got where they are precisely because they did things well. They have standards. They have done it enough times to do it efficiently. They know exactly how it should come out.
This is understandable. It is also the logic that keeps senior people doing $50-an-hour work while their $500-an-hour judgment sits idle waiting for space on the calendar.
The question is not "can I do this better than AI?" The answer is almost certainly yes, for most tasks. The question is whether the differential between your output and AI-assisted output justifies the cost of your time. For most routine knowledge work, it does not.
A Framework for Identifying What to Transfer
Run through your recurring tasks with these three questions:
Does this task require my specific expertise and judgment, or does it require general professional competence?
Writing a first-draft client update requires professional writing ability. It does not require thirty years of experience in your specific field. Your expertise adds value in the review — catching what the draft missed, adding the nuanced observation, adjusting tone for the specific relationship. But the initiation of the draft does not require you.
Would I feel comfortable having a skilled associate do this with my oversight?
If the answer is yes, AI can probably do the same task. The question is whether you are currently doing the work of both the associate and the senior reviewer — because there is no associate.
How much time does this take, and how often does it recur?
The high-payoff transfers are not one-time tasks. They are the things that happen every week: status reports, meeting prep, research summaries, first drafts, follow-up emails. A task that takes two hours and happens three times a week is worth twenty times more attention than a two-hour task that happens once a quarter.
The Transfer Categories
First drafts of any written communication. Client emails, proposals, memos, reports, case summaries, project updates — AI can produce a working first draft from a set of bullet points or a voice note. Your job becomes review and refinement, not initiation.
Research and synthesis. Background research on a company, a regulation, a case precedent, a market dynamic. Claude can pull together a structured synthesis of publicly available information quickly. Your expertise then applies to evaluating the synthesis and knowing what it missed.
Meeting preparation. Agenda creation, background briefs, question development, pre-read summaries. Most professionals enter meetings having done none of this because there was no time. With AI, the prep takes fifteen minutes instead of an hour.
Post-meeting documentation. Meeting summaries, action item lists, follow-up emails. This is work that falls to whoever has the most professional responsibility and takes time that produces little value relative to the effort.
Template-driven documents. Any document that follows a pattern — proposals, engagement letters, SOW structures, reporting templates — can be populated from a set of inputs. The professional reviews the output, not builds the structure from scratch.
Data analysis and pattern identification. Not complex statistical modeling, but the initial scan of a spreadsheet, a set of financial statements, or a comparative report. AI can identify the top-line observations quickly, and the professional can then dig into what matters.
How to Actually Make the Transfer
Transferring a task to AI is not as simple as asking Claude to "write my weekly client update." The first few attempts will produce output that is too generic or misses the point. This is normal.
The process has three steps:
Step one: Describe the task in detail the first time. What is this document? Who reads it? What do they need to know? What tone is appropriate? What should it never include? What format works best? Spend twenty minutes building a detailed prompt the first time you hand off a recurring task.
Step two: Evaluate the output and give specific feedback. Not "this isn't quite right" but "the summary is too long, the third paragraph should lead with the risk item not the mitigation, and the tone is too formal for this client relationship." That feedback teaches you what the prompt needs.
Step three: Build a template prompt for the task. After two or three iterations, you will know how to get useful output reliably. Save that prompt. Use it every time. The task is now transferred.
Most professionals who fail to get value from AI are stuck in step one — evaluating each attempt as if it is supposed to be perfect immediately. The transfer requires a short investment in refinement before the payoff arrives.
What Happens to the Time You Recover
This is the question that actually matters, and most guides on AI productivity skip it.
If you transfer two hours a day of medium-difficulty work to AI, you have two options: add two hours of high-leverage work, or stop working two hours earlier.
Both are legitimate. The point is that the choice should be intentional.
For some professionals, the recovered time goes to client relationships that were being crowded out. For others, it goes to the strategic work they keep pushing to "when things slow down." For others, it is simply time to think without the constant pressure of the task queue.
What tends to happen without intention: the recovered time fills back up with more reactive work. The inbox expands to fill the available space. The time saved by AI gets absorbed without a trace.
The redesign is not just about the tasks you transfer. It is about what you deliberately do with the capacity that frees up.
Practical Examples by Profession
Senior HR executive, 18 years: Transferred all first drafts of employee communications, policy updates, and onboarding materials to AI. Estimates fifteen hours a week recovered. Now uses that time for one-on-one conversations with managers — relationship work that was consistently being deferred.
Tax partner at a mid-size firm, 24 years: Stopped writing first drafts of client letters explaining tax positions. Provides key points; Claude writes the explanation. Partners review for accuracy and tone. Firm-wide, this has changed the capacity math significantly.
Operations director, 16 years: Transfers all weekly status reports to Claude based on bullet-point inputs from team leads. Reviews take twenty minutes instead of three hours. Now uses Tuesday afternoons — previously consumed by report writing — for process improvement work that produces durable results.
Frequently Asked Questions
How do I know what quality level is acceptable for AI output in my field?
Apply the same standard you would for work from a capable associate: is the output accurate, clear, and professional enough that your review and approval makes it yours? If yes, the transfer works.
Won't the quality suffer?
In the first few attempts, yes, sometimes. After you have tuned the prompt, typically no. For most routine professional writing and analysis, the quality of well-prompted AI output is competitive with the quality of a solid junior-to-mid-level professional.
What about confidential client information?
Use judgment about what you put into AI systems, and follow your organization's policies. Many tasks can be handled by describing situations in general terms, anonymizing specifics, or using AI to work on structure and language while keeping the substance in your own hands.
I've tried delegating before and it created more work.
AI is different from delegation because the overhead of instruction, review, and revision is lower. There is no relationship to manage, no training curve beyond the first few prompts, and no delay from waiting for someone's availability.
The Leveraged Associate course ($395) is built around this transition — identifying your highest-value work, mapping what surrounds it, and systematically transferring the supporting tasks to AI-assisted workflows. The Leveraged Executive ($1,495) covers this at leadership scale.
Where this goes next
Want the guided, build-it-this-week version of this? See The Leverage Starter — or Turn Experience Into Income with Claude if you want the broader path.
Related reading from The Briefing
- Why I Ignored AI for Two Years and What Changed
- How to Learn a New AI Tool Without Losing Three Days to Tutorials
- The Tool Stack.
Not sure which program fits where you are? take the 2-minute course-fit quiz, or browse the full TLY course catalog.