Practical AI · Episode 54 · Show Control Panel

What To Do When Nobody Wants What You Built

Friday, August 28, 2026 · 11:00 AM CT · Olga + Chris · title LOCKED by Olga 8/27

🎙️ Live Flow · read the news from this ♟️ Deep dive · the slides 💪 Your segment · nobody wants the gym 💰 Funding · Week 39 📰 News · review page 🔎 Deep dive · the research

Read this first. One of these changes the whole hour.

✅ Title locked: "What To Do When Nobody Wants What You Built." It pays off in your gym segment at 11:38, which is minute 38 of the show — so the open now carries a one-line tease ("something I built that nobody wanted, and what I did about it in forty-three minutes") so anyone who clicked for that hears it is coming inside the first minute. The title also covers Salesforce, HubSpot and Meta, which is what makes it an episode title and not a segment title. "Chess, Or Scrambling?" stays as the deep dive's internal name only.

✅ The news and the deep dive are the same story, and it is handled. News story two IS Claudeforce. DECIDED by Olga, 8/27: story two shrinks to a 30 second tee-up and the deep dive carries it. The Live Flow is already rewritten this way — beat 2 is 30 seconds and says so on the page. Do not run both at full length.

✅ The Live Flow is built and live. Read the news from it — it opens with your own standing open, and each story is one continuous read that starts with a question to Chris. Do not read an open off this page; there is only one and it is over there.

✅ Both middle segments are running, CRM first. Deep dive at 11:22, your gym segment at 11:38, funding closes at 11:54. That puts the show at about 60 minutes of content against a 83 to 108 minute norm, so there is room, but the funding hand-off is where it will get squeezed if anything runs long.

What is live vs local. The Live Flow, this panel, the funding page and the tracker are on PageMotor and verified 200. The deep dive slides, your gym deck, the news review page and the research doc are local files on your Mac, which is fine because you are hosting from your machine. Chris cannot reach those four.

1 · Cold open · straight in, 40 sec

No welcome, no name, no housekeeping.

You write this in your own voice. Below are the verified ingredients, not a script. The ingredients

In 24 hours: Salesforce put its entire CRM inside Claude, and HubSpot's co-founder launched a one dollar CRM built for one person.
That is not two product launches. That is the industry admitting the CRM app is optional.
Salesforce stock had its second best day in history the next morning.

Do not open with Anything that starts with what you predicted, what you have been reading, or catching up. The callback comes later and it has to be framed as the viewer's tool, not your scoreboard.

2 · Who we are · 20 sec, then move

Say it once at normal speed and get to the news.

Names, what the show is, that you pull out the part usable on Monday, and the episode number. This is episode fifty-four. Started August last year.

3 · The callback · 90 sec · Chris leads it

Two predictions landed. One of them was yours.

Episode 47, June. Olga said if she had to bet between Salesforce and Microsoft on which one actually becomes AI-native, her bet was Salesforce. On August 26 they announced Claudeforce. Salesforce did not build a chatbot on the side, it moved itself inside Claude.

Episode 50. When OpenAI cut prices 80%, Chris said that could not be the models getting better and something else was going on. On August 25 OpenAI published the first benchmarks for its own chip. The framing trap Chris says the Olga one, Olga does not say it about herself. A callback to your own prediction read in your own voice is a scoreboard. Read by him it is evidence the show calls things early, which is the viewer's reason to keep watching.

4 · News · 7 stories + rapid fire · about 17 min

Read straight down the Live Flow. Do not assemble it here.

1 · Meta tried to replace its own people with agents, and the code broke.
2 · Salesforce moved into Claude. 30 second tee-up only. The deep dive owns this.
3 · A million people hit LinkedIn's slop button, reach dropped 40%. The one story that changes what someone does Monday. This is the Short.
4 · Agents that run without you, in both directions, in 48 hours.
5 · Google will book your travel without you leaving the chat.
6 · Five years and six million dollars of work, done in two weeks for twelve thousand.
7 · The jobs numbers nobody is quoting.
Close of the news block: two scientists said no to Jeff Bezos. Then rapid fire.

5 · Deep dive · the CRM segment · about 8 min

Slides only. There is no script and that is deliberate.

Share the SLIDES tab, not your desktop. Scroll one slide at a time. 16 slides. The spine, if you lose the thread The AI is becoming the brain. Salesforce and HubSpot are fighting to be its memory. The open question is whether that is a plan or a scramble. The four slides that must land

Slide 3, the timeline. Eleven moves in seven months. None of it existed before February.
Slides 4 and 5, the two readings. Chess versus staying relevant. This is the argument. Do not resolve it.
Slide 13, the Anthropic stake. Salesforce owns a piece of Anthropic and booked 2.7 billion in paper gains on it the same quarter. It is in the filing, not the press release, and no coverage mentioned it.
Slide 14, the 6 percent. Your own CRM. The only thing in the segment with proof behind it.

Do not say

No precise Salesforce year-to-date figure. It swung enormously Wednesday. Say "roughly flat on the year."
No record counts from your CRM. Percentages only. Six percent, ninety four percent.
Not "70% of salespeople don't log their notes." No study exists. Use 76% say less than half their CRM data is accurate.
Not "HubSpot missed." Rangan called the slowdown deliberate, on the call, in advance.
Not the fivefold Anthropic usage stat, the Parker Harris quote, or the "fourth monetization vector" line. All secondhand.

6 · Funding · Week 39 · closes the show · 3 to 8 min

More companies, less money. That is the story.

191 companies funded, $8.07B total, the smallest week since April 23. AI took 80 of those companies and $4.57B, which is 56.6% of every venture dollar. The line that makes it land Last week had exactly one round over a billion. This week has exactly one. Strip the billion-dollar round out of each and the market went from $3.59B to $3.57B. Flat. Nothing actually changed except what the money bought. The median AI round rose to $13.0M from $10.8M and the top five took 56.9% instead of 78.5%. Broad week, not a big one. Top five

Poolside, $1B · SF, $12B pre. Half of a roughly $7B Nvidia deal. Press-reported only, say "reported" out loud.
XPENG Robotics, $900M · Shenzhen. 82% of China's entire AI total this week.
Starcloud, $250M · Data centers in orbit. Nvidia and Cisco in.
Instinct, $250M · ⚑ Most show-friendly. A consumer AI assistant your audience could actually download. Still private beta at $2.5B.
Gatik, $200M · Driverless box trucks, warehouse to store.

7 · The close · its own card · last position

ONE ask. Nothing to scroll past.

The close, in your own words — this is the line the whole hour has been building to:
AI made building cheap. It did not make it any easier to find out whether anyone wants what you built. That part is still marketing and sales, and it is now the whole job.

You, Salesforce, HubSpot and Meta all built something the market did not want. The only difference between the four of you is how fast and how cheaply each one found out.

The ask is a pick and it is yours. Recommendation: the email list. It is the number that actually grew this year, and pagemotor.com still has no signup in the footer. If you would rather ask for subscribes, say so. Do not do both.

Do not Do not stack a second ask. Do not summarize the segment again. Do not tease next week unless it is real. Ep53 taught this: a scripted close at minute 58 does not reach the mouth, so this card is short on purpose.