Claude Cowork vs ChatGPT Work: One Weekly Marketing Report, Built in Both

ChatGPT Work and Claude Cowork are the modes in each assistant that carry out longer tasks with your connected apps. Both can repeat a task on a schedule, and the scheduled runs happen in the cloud, even when your computer is off.

For a weekly marketing report, both work. Add Windsor.ai to the assistant with its one-click install. Then describe the report and the schedule in one message, and the assistant creates the task.

ChatGPT created the task exactly as asked and ran it without asking anything. Claude showed me the task to approve first and asked before each data call in the test run. Its report explained more of what changed.

• Both reports were accurate. Spend, conversions and cost per conversion matched Windsor.ai to the cent for both weeks.
• ChatGPT was quicker to set up. It finished the first run in 1 minute 20 seconds and kept the 9:00 AM start I asked for.
• Claude explained what moved. It showed that one new brand campaign produced the whole rise in conversions. ChatGPT listed the campaign without that explanation.

Both reports come from the same prompt, run on 13 October 2026 on windsor.ai’s own Google Ads account.

Claude Cowork vs ChatGPT Work at a glance

	ChatGPT Work	Claude Cowork
Adding Windsor.ai	Plugins menu in Work, one click, then the Windsor.ai sign-in	Customize, Connectors, one click, then the Windsor.ai sign-in
Creating the schedule	Created from my message, Mondays at 9:00 AM as asked	Shown as a card to approve, set to 8:50 AM
Approvals in the test run	None	One per data call; the scheduled runs approve automatically
First run	1 minute 20 seconds	Longer, partly spent waiting for my approvals
Accuracy	Matched the source to the cent	Matched the source to the cent
Explanation	Listed the changes	Traced the rise to one campaign and gave a cause for each CPA increase
Schedule options	Once, recurring (up to hourly on paid plans), or on a Gmail, Slack or GitHub event	Hourly, daily, weekdays, weekly or manual
Plans	Paid ChatGPT plans except Free and Go	Paid Claude plans; web and mobile on Pro and Max, beta on Team and Enterprise

Anthropic is merging Cowork into the main Claude app. On Pro and Max plans the Chat and Cowork switch is being replaced by a single message box, and Claude decides when a request becomes a task. Scheduled tasks work the same way in both versions.

Which one to choose for marketing reports

• ChatGPT Work suits a team that already works in ChatGPT and wants the schedule set up in one message, at the exact time it asks for.
• Claude Cowork suits you if the report should explain the numbers and you want to approve the task before it runs.

Both are personal: the task runs under your account and your connections. For a report that a whole team can run and edit, ChatGPT has workspace agents, which I tested with the same instructions.

Before you start

• A paid ChatGPT plan with Work, or a paid Claude plan with Cowork. In company workspaces, admins can turn Work off for some roles, and Team and Enterprise admins control who can use Cowork.
• A Windsor.ai account with your sources connected. A report on Google Ads alone uses one data source, which the free forever plan covers, with no credit card. A report that combines Google Ads with GA4 or Meta Ads needs more sources after the first 30 days: paid plans start at $19 a month (billed annually) for three sources, or $99 a month for seven. See Windsor.ai pricing.

The prompt I used

I sent this message to both assistants. It is the instruction set from our workspace agent, with a first line that asks for the schedule:

Create a scheduled task that runs every Monday at 9:00 AM Eastern European Time, and run it once now.

Every run, use Windsor.ai to pull Google Ads data for the last full week (Monday to Sunday) and the week before, for every connected Google Ads account. First call the tool that lists connectors and accounts, and get valid field IDs before you request data.

Fields: date, account name, campaign, spend, impressions, clicks, conversions, conversion value.

Report in this order:
1. A headline with spend, conversions and CPA against the previous week, with % change.
2. The top 5 campaigns by spend, with the same metrics and their week-over-week change.
3. The 3 campaigns with the largest change in conversions, at any spend level.
4. Campaigns whose CPA rose more than 20% with at least 50 in spend, with one sentence on what changed.
5. Campaigns that spent with zero conversions.

Sum conversions from the raw rows before you calculate CPA. Leave ROAS out if conversion values are zero or look like placeholders. Use each account's currency and never add different currencies together. If a call fails or returns no rows, say so and stop. Use read actions only.

Keep the last sentence of the prompt. Windsor.ai can also make changes: its Execute Action tool can pause campaigns and change budgets on Google Ads, Meta Ads and other ad platforms.

Set up the report in ChatGPT Work

1. In ChatGPT, select Work at the top of the page.
2. Click Plugins under the message box and search for “Windsor”. Click + next to Windsor.ai and sign in to the Windsor.ai account that holds your data.
3. Open Plugins again and select Windsor.ai, so it is attached to the message. Paste the prompt and send it.

The Windsor.ai MCP entry in the screenshot is a custom plugin from our workspace agent test. For Work you need only Windsor.ai.

ChatGPT created a task called “Google Ads weekly report” for Mondays at 9:00 AM, Europe/Bucharest time. It ran the report straight away, with no questions and no approvals.

To edit, pause or delete the task, open Scheduled from the clock icon in the ChatGPT sidebar. The plan sets how many tasks can be active at once: 5 on Plus, 10 on Business and 15 on Pro and Enterprise. If a task tries to send a message or change data in a connected app, ChatGPT can pause it until you approve.

Set up the report in Claude Cowork

1. In Claude, open Customize, then Connectors. Search for “Windsor”, click Add next to Windsor.ai and sign in to your Windsor.ai account.
2. Start a new task in Cowork: select Cowork under the message box. If your message box has no Chat and Cowork switch, you have the merged Claude app, and any conversation works.
3. Paste the prompt and send it.

Claude did not create the task straight away. It showed a Schedule task card with the task name, the time, the instructions rewritten in its own words and a Schedule button. The rewritten instructions listed the four Windsor.ai read tools by name and told the task never to call a write tool. The card was marked Auto, which means the scheduled runs go ahead without asking. It also showed Cloud as the place where the task runs.

The time on the card was 8:50 AM. After the run Claude explained that tasks set on the hour can be delayed by server traffic, and it offered to switch to exactly 9:00. Read the card before you click Schedule, and ask Claude to change the time if it matters.

In the test run Claude asked me to approve each Windsor.ai call (get connectors, get fields, get data), with Allow once, Always allow and Deny. My session’s approval mode was Manual, shown under the message box. Claude then calculated the totals with Python on the raw rows.

The task appears under Scheduled in the Claude sidebar. Its page shows the next run, the instructions, the schedule and the permissions, with buttons to edit, delete or run it now. The permission was Automatically approve: Claude runs on its own and stops to ask if something looks unsafe.

What the two reports said

Spend, conversions and cost per conversion in both reports matched the two weeks I pulled from Windsor.ai myself, to the cent. Each report had all five sections and named the same campaign that spent with no conversions. Neither showed ROAS, because our account records conversion values as round placeholder amounts.

Our cost per conversion fell sharply that week. ChatGPT reported the drop and listed a new brand campaign as the largest change in conversions. Claude added that this campaign produced the entire rise, and that conversions from all other campaigns fell. For each campaign whose CPA rose, it named the cause in one sentence, such as fewer clicks, a lower conversion rate or a higher cost per click. Each of those statements matched the source data.

Google Ads kept adding conversions to past days while I tested. The previous week had one more conversion in the afternoon than in the morning. When you check a report, pull the source data at the same time, and expect the most recent days to change.

This was one test: one Google Ads account in one currency, one week compared with the week before, and one run in each assistant. A report’s wording changes from run to run. The first run of our workspace agent, with the same instructions, did not mention the brand campaign at all. Both assistants called the same Windsor.ai tools; the differences are in how each one created the task and wrote the report.

Other reports to schedule the same way

Each of these works as a scheduled task in either assistant. The first needs three data sources; the other two need one each.

Paid and organic traffic in one Monday table:

Create a scheduled task for every Monday at 9:00 AM. Through Windsor.ai, get last week's spend and clicks from Google Ads and Meta Ads, and sessions by source from GA4, with the week before for each. Put the three sources in one table with the change for each row, and name the source with the largest drop in sessions. Use read actions only.

Search Console queries that lost impressions:

Create a scheduled task for every Tuesday at 9:00 AM. Through Windsor.ai, compare Search Console impressions and clicks by query for the last 7 days with the 7 days before. List the 15 queries that lost the most impressions and the page each one sends traffic to. Search Console data arrives 2 to 3 days late, so give the last date in the data.

Daily Google Ads spend check:

Create a scheduled task for every weekday at 8:00 AM. Through Windsor.ai, get yesterday's Google Ads spend by campaign and compare it with each campaign's average daily spend over the previous 14 days. List campaigns that spent more than 30% above or below their average. Read the data only and change nothing.

Conclusion

ChatGPT Work and Claude Cowork both run a weekly marketing report from Windsor.ai data on a schedule, and in my test both got every total right. ChatGPT needed less setup. Claude’s setup took more approvals, and its report explained the change in conversions that ChatGPT only listed. Whichever you use, keep the task read-only and check the first few reports against the source.

The install steps for each assistant are in how to connect Google Ads to ChatGPT and how to connect Google Ads to Claude. For a report your whole team shares, see ChatGPT workspace agents with Windsor.ai. Our Claude vs ChatGPT study shows how marketing teams use each assistant, and ChatGPT vs Claude for social media runs both assistants on five social media accounts.

🚀 Connect Google Ads, GA4, Meta Ads and 350+ sources to ChatGPT and Claude with Windsor.ai. Free forever plan, no credit card. Try Windsor.ai.