# Presentation decision worksheet

Use AI for X · Worksheet version 1 · 27 September 2026

**Proposed evaluation, not performed:** use this worksheet to compare two tools on your own accounts. No vendor results, timings or costs are supplied. Blank cells are for your observations; an empty cell is not a pass.

Start with the [presentation task](https://useaiforx.com/for/ai-presentation/). This worksheet turns its Q3 support outline into a decision record. Allow a provisional 30–45 minutes for two draft checks, excluding signup and optional export checks; this is a planning allowance, not a measured completion time. Save a copy for each comparison. The companion CSV contains the same checks with separate Tool A and Tool B columns.

## 1. Fix the brief before opening either tool

**Decision:** Which account can turn this outline into an editable draft that preserves the facts, within my limits?

| Your requirement | Fill in before testing |
|---|---|
| Audience and purpose | |
| Tool A / Tool B | |
| Draft deadline and maximum hands-on minutes per tool | |
| Maximum additional spend, currency and tax assumption | |
| Required delivery format, if any: PPTX / PDF / link / other | |
| Must the recipient edit individual text elements? | |
| Is visible vendor branding acceptable? | |
| Required recipient app and version, if relevant | |
| Allowed extra generation attempts per tool | |
| Brief version and saved input filename | |

Copy this fixed outline into a plain text file:

```text
1 Title: Q3 support update
2 Agenda
3 Goal: cut first-reply time
4 Result: median first reply 4h 10m against a 3h goal
5 Goal: publish 20 help articles
6 Result: 14 published
7 What slowed us
8 Q4 plan in three bullets
9 Risks
10 Ask: one extra hire
```

Use the same instruction in both tools: **“One slide per line, keep the order, do not add figures.”** Save any additional instructions with the brief and use them unchanged in both tools.

This is a practice brief, not a report of a real company's results. It does not supply the causes on slide 7, the three actions on slide 8 or the risks on slide 9. Either add your own approved details to both copies before starting, or record those gaps as “needs input.” Do not accept plausible AI inventions as evidence. Unresolved placeholders prevent a delivery-ready result. Use only material you are allowed to upload.

## 2. Record access, then preserve the first draft

| Account and run record | Tool A | Tool B |
|---|---|---|
| Product, plan and feature used | | |
| Run date, region and relevant account restrictions | | |
| Free / trial / paid; trial end and renewal obligation | | |
| Drafting feature available on this account? | | |
| Official plan/source URL and date checked | | |
| Starting credit or usage display | | |
| First draft saved filename or private link | | |
| First draft slide count and outline order | | |
| Generation attempts, including failed attempts | | |
| Ending credit or usage display | | |
| Observed usage difference, or “not exposed” | | |
| Active work minutes / waiting minutes | | |

If a required feature is unavailable, record **access limit** and stop that part. Do not silently upgrade. Treat a separately authorized paid rerun as a new run with its own plan, time and cost. Do not infer credits from the slide count or equate different vendors' credits.

Save the untouched draft before correcting it. For each row below, record: **actual slide number; observed wording/figure; editable yes/no; correction needed**. Use “not run” if blocked. Add an output filename or screenshot reference when it helps another person check your observation.

| Outline reference | Tool A: first-draft observation | Tool B: first-draft observation |
|---|---|---|
| 1 — Q3 support update | | |
| 2 — Agenda | | |
| 3 — Cut first-reply time | | |
| 4 — Median 4h 10m; goal 3h | | |
| 5 — Goal: 20 help articles | | |
| 6 — Result: 14 published; compare with goal 20 | | |
| 7 — What slowed us; approved detail or needs input | | |
| 8 — Q4 plan: three approved bullets or needs input | | |
| 9 — Risks; approved detail or needs input | | |
| 10 — Ask: one extra hire | | |

Check every added number, name, quote and factual assertion against the brief or an approved source. “Sounds reasonable” is not support. Record each unsupported assertion below; duplicate rows as needed.

| Tool / slide | Assertion | Supporting brief line or source passage | Keep / correct / remove / needs input |
|---|---|---|---|
| | | | |
| | | | |

Now change the layout of one slide without changing its text. Record the slide number, whether its text and figures survived, manual corrections and correction time. Save the corrected version under a different name.

**Draft acceptance:** ten slides in outline order; all supplied figures preserved; no invented statistics or unsupported factual additions; text editable inside the tool; no unresolved placeholders. Merged/reordered slides, changed figures and corrections requiring a whole-deck regeneration are failures against this brief. Keep the first-draft result separate from the corrected result: a repaired draft does not erase the original error.

## 3. Optional delivery check — separate from draft acceptance

The task above evaluates outline-to-draft fidelity. Complete this section only if your decision also requires a deliverable outside the editor. A successful draft does not prove a successful export, and a blocked export does not erase the draft observations.

| Delivery evidence | Tool A | Tool B |
|---|---|---|
| Export format and required plan; official source/date | | |
| Export allowed on the tested account, or access limit? | | |
| Export filename and corrected draft version | | |
| Vendor watermark/branding: none / present / unknown; where? | | |
| Opened in recipient app/version; date | | |
| Ten slides and order preserved in the opened file? | | |
| Slide 4: 4h 10m and 3h still exact? | | |
| Slides 5–6: goal 20 and result 14 still exact? | | |
| Text clipped, font substituted, chart/image misplaced? | | |
| If editable delivery required: changed one text element and saved? | | |
| Reopened saved file; edit and layout intact? | | |
| Export/rework active minutes; waiting minutes | | |
| Delivery result: pass / fail / access limit / not run | | |

A PDF or flattened image can satisfy a viewing requirement; it does not establish editable text delivery. Inspect the actual downloaded file for branding. Do not remove a watermark to make the result pass; record the account condition and whether it meets your requirement.

## 4. Make the decision with explicit assumptions

| Decision record | Tool A | Tool B |
|---|---|---|
| First draft: pass / fail / access limit / not run | | |
| Corrected draft: pass / fail / access limit / not run | | |
| Layout-edit slide and preservation result | | |
| Corrections made and active correction minutes | | |
| Corrected version filename | | |
| Total active minutes / total waiting minutes | | |
| Actual additional charge, currency and tax treatment | | |
| Existing subscription cost included or excluded, and why | | |
| Any proposed upgrade: quoted amount, billing term, seats, region | | |
| Upfront commitment versus monthly equivalent | | |
| Planned decks/month, attempts/deck and quota reset assumption | | |
| Unknowns that could change the choice | | |

Add active setup, generation supervision, review, correction and optional delivery work to obtain total active time. Keep waiting time separate. Record an unavailable charge as **unknown**, not zero. A price displayed by a vendor is a quote, not an amount you paid. Do not derive an exact yearly bill from a rounded monthly equivalent, compare different currencies without a documented conversion, or assume a one-time trial allowance renews.

**Selected tool and plan:** ____________________

**Evidence that it meets the required draft and delivery conditions:** ____________________

**Compromise accepted / remaining blocker:** ____________________

**Next action and recheck date:** ____________________

Choose from the conditions that passed, then compare observed correction effort and relevant cost. If neither tool meets a must-have requirement, record **neither yet** and identify the missing input, access or delivery capability. This worksheet gives no vendor score or ranking.
