Name three categories of work
Where Claude saves real hours — not minutes, not novelty time, real hours that compound across a week.
Beginner training
Why Generative AI Matters for Knowledge Work
Make the business case in your own terms. Generative AI is a different kind of productivity wave, and the gain shows up where you don't expect it — in the boring middle of work, not the headline tasks.
Start here
What you'll walk away knowing
Reframe "do my job for me"
Explain why that's the wrong framing — and what the right framing replaces it with for an individual contributor at GMFS.
Identify one task
Pick one task in your week — specifically — that would change shape with Claude in the loop. You should be able to name it by the end of this module.
What changed and why now
Three shifts that make this wave different
Generation, not retrieval
Prior productivity tools surfaced things that already existed — search engines, dashboards, document repositories. Generative AI creates output. A draft, a summary, a plan, an explanation. That's a different muscle.
Judgment, not production
When the production step gets cheap, the judgment step gets more important. Anyone can generate ten options. The value is in picking the right one — and knowing why.
Tools you delegate to
Excel and Word are tools you operate. Claude is a tool you delegate to — you describe the outcome and let the tool work toward it. That's a meaningful shift in how you spend a workday.
Trainer note: The productivity gain shows up in the boring middle of work, not in the
headline tasks. The flashy demos look impressive; the real ROI is the email draft, the meeting prep, the
policy skim — the dozens of small things that take five to twenty minutes each.
Where the leverage sits
Three patterns to look for in your week
First-draft work
Anything where the blank page is most of the cost. Emails, memos, status updates, talking points. Once a draft exists, your real work — editing, deciding what to keep — is faster than starting cold.
Blank-page costDocument-heavy work
Anything where reading time dominates. Long PDFs, guideline updates, lender briefs, contracts. Claude reads end-to-end and surfaces the parts you actually need to act on.
Reading costDecision-prep work
Anything where assembling the inputs is most of the job. Pipeline reviews, account prep, manager check-ins. Claude does the gathering so you arrive at the decision with options, not just facts.
Assembly costFind your leverage
Five rules for picking where to start
Pick a task you do weekly
One-off tasks don't compound. Repeating tasks do. Five minutes saved on something you do every week is a working day a year.
Time it twice — before and after
Once without Claude, once with. Don't guess at the savings. Many "AI saves time" stories evaporate when you actually time them.
Count rounds of revision, not just first-draft time
A first draft that lands in two minutes but needs seven rounds of cleanup is not a win. The honest number is total time to ship.
Share what worked with a teammate
The compounding gains come from prompts and patterns spreading through the team. Hoarding a good prompt helps one person; sharing it helps twenty.
Resist using Claude on the wrong problem
Novelty is a bad reason to deploy a tool. If your honest answer is "this is faster with a keyboard shortcut," use the shortcut.
Prompt upgrade
Weak prompt
Help me be more productive this week.
Leverage-aware prompt
Work-ready prompt
Every Friday I write a pipeline update for my manager. Today it takes ~45 minutes — copying loan data, summarizing stuck files, writing notes. Here's last week's update (pasted below) and this week's loan export (attached). Draft this week's update in the same structure. Flag anything that changed materially from last week.
High-leverage patterns
Four reliable wins to start with
Notes → deliverables
Turn meeting notes, scratch ideas, or a voice memo transcript into a polished email, summary, or one-pager.
Long documents → decisions
Turn a 30-page guideline, a contract, or a long thread into the three things you actually have to do about it.
Blank pages → structured drafts
Turn a brief into a first draft you can edit. The hardest version of any deliverable is the empty one.
Vague asks → clear requests
Turn a four-line Slack request from a stakeholder into a properly scoped task with assumptions, questions, and a draft response.
Honesty checks
Five questions to ask before you declare a win
Employee rule: The point isn't to look productive with Claude. It's to free your time and
attention for the work that actually moves the business. Be honest about the math.
Find your leverage points
Six exercises for your week
Treat this as a personal audit. By the end you should be able to name one task you'll change permanently.
- List every recurring task you do in a typical week. Don't curate — write them all down.
- Rank them by hours. The biggest time sinks are usually the boring middle, not the headline tasks.
- Pick the top three candidates for a Claude trial. Apply the three workday-pattern lenses (first-draft, document-heavy, decision-prep) to decide which.
- Run a before-and-after timing on one. Note total time to ship, not just first-draft time.
- Share one finding with a teammate — what worked, what didn't, what you'd try differently.
- Decide which task you'll commit to running through Claude permanently. Write it down where you'll see it tomorrow.
Completion standard
You've finished this module when you can name one specific weekly task you've moved to a Claude-in-the-loop workflow, the time you spend on it now, and how you measured the change.