8 Workflows worth stealing today

Half a day, twenty minutes

August 13, 2026

In partnership with

It was four in the afternoon and I was clicking through client folders, copying numbers into a document, reformatting a header for the third time. Ordinary work. The kind that feels productive because your hands are moving.

I had used AI six or seven times that same day. Asked it a question, took the answer, closed the tab, went back to the folders. Not once did it touch a single file I was actually working in.

That is the part that stings in hindsight. The tool was open the whole time. I simply had no system for it.

Blu Dot surpasses 2,000% ROAS with self-serve CTV ads

Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:

After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.

The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.

“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”

Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.

*Ad

The gap is not intelligence and it is not prompting

I read a post from someone who runs an AI newsletter, and the framing reorganized how I think about the category. The people quietly winning with AI are not smarter and they are not better at wording prompts. They have built an actual system around what AI does inside their business.

Repeatable workflows. An automation layer. Not random questions tossed into a chat window and then forgotten.

That distinction sounds small until you sit with it. Ad hoc questions never compound. You get an answer, and tomorrow you start from nothing again. A workflow you can run every Friday accumulates. The difference is not the model, it is whether anything you built survives the session.

What Cowork changes about the arrangement

The post's author says what unlocked this for them was building with Claude Cowork specifically. The distinction they draw is that you can run agents across your real files, emails, documents, and data. Not generate text about the work. Actually do the work.

I read that and thought about my Friday folders. The bottleneck was never that I could not write a status report. The bottleneck was that the AI had no access to the material the report was made of, so everything it produced was a template I then had to fill in by hand.

Give it the files and the job changes shape entirely. They shared twenty five prompts and flagged eight that save the most time immediately. Those eight are what I want to walk through, because grouped together they tell you what a real workflow looks like.

The reporting you are still doing by hand

The weekly status report generator reads your client files and drafts the full write up in minutes. This is the exact thing I was doing at four on a Friday, except the agent goes to the source material instead of making me dig for it.

The budget versus actuals tracker finds your files, compares the numbers line by line, and flags every variance over whatever threshold you set. Tedious, error prone, and completely mechanical, which is to say it is a terrible use of a human afternoon and a perfect use of an agent that does not get bored two thirds of the way through.

The end of day wrap report checks your files, calendar notes, and logged tasks, then writes the daily summary. A clean handoff to tomorrow you, written without the six pm mental scramble where you try to remember what actually happened.

See how AI delivers fast, secure, personalized experiences

Join Fin and Plaid on August 13th to see how AI can resolve issues directly in customer conversations. This includes everything from reducing bank-linking friction, enabling proactive support, and helping customers complete real financial tasks.

*Ad

The prep that decides how the meeting goes

The competitor research brief pulls the latest news on a competitor and compiles a tidy two page document. Useful before a strategy call when you need to sound current and you do not have an hour to spend on search results.

The sales call prep sheet combines company research, recent news, and your past conversation history into one page. You walk in warm instead of skimming old emails in the parking lot.

Notice that neither of these is a task you would have skipped if you had time. They are tasks you skip because you never have time, and the meeting happens anyway with you slightly less prepared than you wanted to be. That is a quieter kind of cost than a wasted Friday, and probably a more expensive one.

The admin that quietly eats your mornings

The inbox zero assistant sorts unread email by urgency and builds a prioritized task list. You open your morning already knowing what needs a reply first, instead of scrolling until something feels urgent enough to answer.

The performance review draft writer produces structured reviews with strengths, growth areas, and goals for the next period. You still supply the judgment, but the blank page problem disappears, and the blank page is the reason those reviews sit unwritten for two weeks.

The onboarding pack compiler grabs every file a new hire needs and builds a formatted document organized by section. Work that used to eat an afternoon, handed back to you in minutes.

See Why HubSpot Chose Mintlify for Docs

HubSpot switched to Mintlify and saw 3x faster builds with 50% fewer eng resources. Beautiful, AI-native documentation that scales with your product — no custom infrastructure required.

*Ad

Every one of the eight has the same skeleton

Look at them together and the pattern is impossible to miss. The agent locates real files or real data, reads them, does something useful with the contents, and returns a finished artifact. A report. A brief. A prioritized list. A document someone else can open.

That is the whole leap. The AI is not answering a question, it is completing a job.

Once you see the shape you can build your own version of any of them without needing the original wording. The prompt is not the asset. The shape is.

The test I run before calling something a workflow

Here is the filter I have started using. Can I hand this to someone else on my team and have it produce the same result without me in the room?

If the answer is no, it is a prompt I got lucky with. If the answer is yes, it is a system.

That word, delegatable, is doing most of the work in this whole idea. A prompt gets you one output. A system gets handed off, run on a cadence, and improved by whoever runs it next. Most prompt libraries fail this test, which is exactly why saving them changes nothing about your week.

The line I have not been able to shake

The author's closing argument is that founders and operators who build real systems around AI will compound their output in ways that feel almost unfair to everyone around them. The ones still using it as a fancy search engine will keep wondering why nothing changed.

I was in that second group for months. What bothers me is that nothing was blocking me. The tools were available, the prompts were public, and I simply never built anything I could run twice.

They also made a point I keep returning to. If someone on your team is still doing any of these eight by hand, this will land hard with them, because the manual version of compiling an onboarding pack or reconciling a budget is exactly the sort of work that drains a week without anyone noticing.

What I am doing tonight

I am not going to attempt all twenty five. That is how these experiments die in week two, buried under ambition.

I am picking the one that matches something already sitting on my plate, running it against my real files rather than a tidy test case, and seeing whether the time saved is real or imagined. If it works, it goes on a cadence. If it does not, I lost twenty minutes.

Pick the workflow that matches the chore you were dreading this week and run it once before you close your laptop. One thing you can repeat beats twenty five prompts you saved and never opened.