PRACTICAL WORK AIwith Alex ReedTry a workflow

PRACTICAL AI FOR MANAGERS

AI gave you
an answer.
CHECK THE
WORK.

A neat project plan. Three competing dates. A promise of “support included.” Turn them into evidence, open questions and decisions you can check.

Try a real workflow Get the free working kit →

Three fictional cases. Sources, prompts and editable records.

ONE LAUNCH.THREE DATES.
L2 / DRAFT SCHEDULEOCT 26DRAFT

CRM validation omitted.

L1 / CUSTOMER BASELINENOV 02APPROVED TARGET*

*Acceptance and CRM validation still apply.

L3 / SUPPLIER UPDATENOV 09UNSIGNED

Credential availability unknown.

Fictional practice case / Oct–Nov 2026
Newest does not mean approved.

Alex Reed, the established AI-generated fictional host

“My calendar would like a word.”ALEX REED / FICTIONAL AI HOST

A kit you can actually runThree staged prompt packs, 13 editable CSV templates, completed examples and a portable agent skill.

WORKING EDITION 0.2 / 05 OCT 2026

THE SAME CASE. A USEFUL NEXT STEP.

Make the problem obvious.
Make the answer checkable.

All three dates belong in the review. The approved target keeps its conditions. Drafts and proposals stay visible until the sponsor approves a change.

Open the three-date workflow →

Before you share the date

  1. SOURCEWhich document says it?
  2. STATUSApproved, draft or proposed?
  3. CONDITIONSWhat must happen first?
  4. OWNERWho can approve a change?

A tidy summary is not an approval.

START WITH A REAL JOB

Three messy jobs.
One way to check the work.

THE INCOMPLETE BRIEF

Build a defensible WBS

AI maps requirements to deliverables, proposed owners, acceptance evidence and calibrated role estimates. Draft additions and customer work stay visible.

Take away: a traceable WBS, coverage and uncertainty register.
THE CONFLICTING VERSIONS

Reconcile the documents

AI quotes each conflicting claim, preserves approval status and maps downstream impact. Open decisions stay open until an owner resolves them.

Take away: a conflict register, impact map and proposed decision log.
THE TOO-SMOOTH PROPOSAL

Review what’s actually promised

AI compares commitments to the buyer’s operating needs, separates optional costs and drafts specific questions for the supplier.

Take away: an obligation map, cost reconciliation and decision brief.

PUT YOUR AI TO WORK

Pick a job. Start a run.

Download the kit, attach the listed practice sources to a fresh chat, then paste the starter below. The full prompt pack takes you through building, checking and exporting the records.

FICTIONAL PRACTICE CASE

Harbor discovery pilot

Attach B1, B2, I1, E1_labeled.txt and reference_wbs_labeled.txt from practice/harbor. The last two are disclosed labeled copies; originals are included for checking.

Build a scoped plan for sponsor review. Preserve exclusions and CRM uncertainty.

Download full WBS prompt pack

Use the same text in ChatGPT, Claude or Gemini. File creation and calculation tools vary by account. Three small ChatGPT cases were reviewed; some needed corrections. Broader reliability, Claude and Gemini runs remain untested.

Start with source extraction. Inspect the result before moving to Build and Challenge in the full pack. Save each stage.

HOW THE WORK GETS BETTER

Let AI do the legwork.
Keep the decision.

The prompts extract, compare and draft. The skill saves records and runs an offline checker when tools are available. You review the evidence, assumptions and remaining decisions.

Setup notes for your assistant →
  1. 01

    Set the working basis

    Define the decision. Inventory sources, scope and approval.

  2. 02

    Build records you can inspect

    Preserve quoted evidence, proposed owners, calculations and gaps. Export to editable CSV.

  3. 03

    Check before committing

    Verify source quotes and arithmetic. A completed analysis can still leave a business commitment blocked.

WHAT THE EXAMPLES SHOW

Concrete checks.
Named limits.

WBS: 192 hoursThe supplied fictional role estimates reconcile. 2.4 capacity-weeks does not become a launch date.

Dates: three claims preserved2 November stays the approved conditional target. 26 October and 9 November remain open proposals.

Support: a $100 mismatchListed base items total $2,400; stated total is $2,500. Optional support and unknown tax stay separate.

Limited execution testing, with the outputs included. One fresh-session agent run per fictional case produced reviewed records and passed the mechanical checker. This is early evidence of execution, not proof of better professional outcomes or time savings. Three small ChatGPT workflows were also completed and reviewed; date and proposal variants needed corrections. Independent manager outcomes and broader ChatGPT, Claude and Gemini testing remain pending.

Read the validation record

YOUR WORKING PACKAGE

CHECK THE WORK.
The Managerial AI Playbook

Start with START-HERE. Use the staged prompts in a chat, or give the portable skill to a file-capable agent. Compare your independent practice run with the completed examples afterward.

  • Field guide with worked explanations in PDF and Markdown
  • Raw WBS, launch-date and support source packs
  • Three prompt packs plus check-and-resume instructions
  • 13 CSV templates, operating brief and completed records
  • Portable skill and dependency-free Python checker
Guide cover: Check the work. The Managerial AI Playbook.FREE WORKING EDITION 0.2

Start with the complete kit

Download complete kit · ZIPRead the field guide · PDFDownload the agent skill separately · ZIP

The kit includes the guide. No API key is needed. The checker requires Python 3.9+; ordinary chat prompts do not.

Use and adapt within your organization and client work. Sharing terms and correction details are included.

BEFORE YOU START

A few things to know.

What should I try first?

The three-launch-date case is a compact start. Attach its three labeled files, run extraction, inspect the quotes, then build the conflict register. Try the support review next, or use the WBS case for planning work.

Does this work with ChatGPT, Claude and Gemini?

The kit provides copy-and-paste prompt routes and setup notes based on official documentation. Actual capabilities depend on your account. The first three small ChatGPT workflows were reviewed, with corrections needed in date and proposal records. Revised prompts still need fresh runs. Claude, Gemini and a Claude skill import remain untested. If file export is unavailable, save the returned CSV text.

What does the checker verify?

Record structure, linked IDs, source hashes when supplied, exact quotes in labeled text, role sums and cost arithmetic. It cannot establish complete source coverage, approval authenticity, a realistic estimate or a feasible schedule.

Can I use real workplace documents?

Use an AI service and account your organization permits for those documents. Start with a scoped task and an explicit decision owner. Keep unknowns open and verify the original evidence before using the output.

Is Alex a real person?

Alex Reed is an original AI-generated fictional character. The cases illustrate workplace situations; they are not testimonials or live delivery benchmarks.