OpenAI adds business-run agents to ChatGPT

Welcome to The Signal from Superpower Daily with Maya and Theo

Today

OpenAI is giving businesses verified profiles and their very own customer facing agents directly inside

ChatGPT We are looking at a fundamental shift in how you acquire customers and manage sales on the

internet Let us dive

in So we have massive structural changes to cover

today We really

do The internet ecosystem is fundamentally rewiring itself right

now Yeah we are basically watching AI step out of the chat box It is starting to take actual physical action in the real

world Let us start right at the

top

Right OpenAI just launched a major new beta

program They

did It's called Businesses in

ChatGPT And this

launch you know it completely changes the relationship between commercial brands and artificial

intelligence It is a major paradigm

shift I mean think about the past few

years

Right Businesses were just passive

subjects

Exactly They were just data scraped from the

web The AI would summarize their

websites It would just guess

things

Yes it would often guess their

pricing It would hallucinate their store

hours That lack of control was just a massive headache for

founders Oh

absolutely You could not control what ChatGPT told a customer about your

brand But now well that dynamic is entirely

flipped It really

is OpenAI has introduced the Business

Hub So what does that mean

practically Well eligible businesses can now claim official verified

profiles

Right They can do this in any market where ChatGPT is currently

live So you

are no longer just training data You

now

have an active verified presence inside the ecosystem And this

verified presence I mean

it comes with major new features It does

Businesses can

feature self service product feeds Oh

wow Yeah

and these feeds plug directly into the AI So

they include live pricing Yes

they include live pricing and real time inventory availability

That live pricing aspect is absolutely critical

It really is

I mean we all know the deep frustration of an AI

hallucinating a price Right A

customer

asks ChatGPT how much a product costs And the

AI confidently states a price from three years ago Which

is

a

disaster It is

The customer clicks through to the website They see

a much higher price They feel

cheated Exactly They

feel cheated and they abandon the cart It creates

immediate brand damage It destroys

the trust before the customer even reaches the checkout page This new

integration well it solves

that

problem Because it connects

the language model directly to current commercial reality The AI is

no

longer guessing It is pulling exact

current numbers from the merchant Now Shopify merchants have

a distinct advantage in this rollout They do They can

link their

stores directly to the business hub So it streamlines the

entire verification process Yeah it builds on

the existing Shopify catalog integration Because Shopify already had

a back end relationship with OpenAI Exactly Their eligible

products

were already being formatted for AI ingestion So now this

new merchant profile anchors that presence Right It gives

those

products a verified home base inside ChatGPT Now the profile

is incredibly important It is But

you

know

the

profile is just the foundation The truly disruptive

news is the agents Absolutely OpenAI is

introducing

business agents So what are

these exactly They are custom

built customer facing

AI representatives Wow

Yeah

They

live inside the ChatGPT interface They

talk directly to the consumer on behalf of the brand This

is a restricted launch for now though It

is It

is initially available only to eligible U S e commerce and online service

companies That phased rollout makes perfect sense to

me Oh for

sure Customer service is incredibly high

stakes

Right If an agent gives bad advice in a medical or legal

context the liability is

catastrophic

Exactly E commerce provides a relatively

constrained environment It is a safer place to test these interactions

at scale It

really is So let us walk through the actual customer

journey

here We really need to understand how a user actually encounters these

agents Well the journey starts with a standard

prompt Let us say a user asks ChatGPT a question about running

shoes

Right The AI generates a standard

answer and inside that

answer a specific business is

mentioned Got

it The user sees a verified badge next to the business

name

Okay so they click that

badge and then they are taken to the verified

profile

Yes They look at the live product

feed

And at that point they have a choice

What kind of choice

They must explicitly choose to start a conversation with the business agent

So they have to click a button to initiate that specific chat

Exactly

That explicit choice is a really interesting friction point

It is

The handoff is not automatic

Right

The AI does not just seamlessly turn into a salesperson

And that is a vital boundary for user trust

Yeah

that makes sense

OpenAI strictly labels this agent exchange They keep it

entirely separate from the original ChatGPT conversation So the interface

changes Yes the visual

cues change So you know

you are no longer talking to the general oracle Exactly You have

chosen

to step into a specific digital store I mean it

is like walking down the street Right You're talking

to

a smart friend That is the main chat GPT window Sure Then you decide

to

walk into a boutique The friend waits outside That is a great

way to put it You are now talking

to the store clerk That is a perfect

analogy OpenAI wants to

maintain the integrity of that main conversation Furthermore businesses can

actually

sponsor their business agent Oh wow They

can pay

to make their agent the destination for chat GPT ads So the monetization engine

is really spinning up here It is Sponsored placements

are clearly

labeled But the underlying intent

is pretty obvious They are building an

advertising network We definitely need to

connect this to the bigger picture Yeah Because this transforms

chat

GPT entirely It does It moves

from being

a neutral research tool into a direct customer acquisition channel Exactly It becomes a

customer

service channel OpenAI is essentially building

a walled garden They absolutely are A

walled garden is

a closed ecosystem Right The platform provider

controls

all the activity inside So product discovery happens

inside chat GPT Yes Product comparison

happens

inside chat GPT Right And

customer

service happens inside chat GPT Exactly

The

user never has to leave They

never have to visit the open web No They

never have to use a traditional search engine This

is a direct assault on the traditional internet model It

really is It

threatens Google search It

threatens traditional e commerce storefronts

Because everything takes place inside OpenAI's ecosystem

Exactly Now that is a massive paradigm shift for the

internet But I have to push back a bit on the value proposition for the merchant

here

Okay Because I look at the system and I see a glaring

issue What is

that This feels like the yellow pages of the AI

era But it is a yellow pages that talks back to your

customers

Ah you are looking at the data

boundary

This is the most important limitation of the entire announcement

It is

It is a massive caveat for any founder listening right now

Businesses only receive aggregated insights from these interactions

Right They do not get the raw

data They do not get the raw

data They just get high

level

metrics Yes They get

mention counts They get website

click through rates And they get

recurring discovery topics But there is a catch They only get these topics when certain interaction

thresholds

are met Right They do not get access to

individual customer transcripts That is

the core limitation OpenAI explicitly states they do not

share user conversations So they do not share the business agent chats with

the

business sales No The merchant cannot read

the chat logs Is

aggregated data enough I mean I have to question the return

on investment here That is

a fair question I think about a founder trying to

optimize their sales funnel Well

founders are deeply analytical They

rely

on granular data Exactly You are asking a founder to invest time and money

into

a new channel Right But you are blinding them to

the actual customer interactions Yes They cannot read the chat to see

why a sale dropped off They do not know what the AI said that

caused the customer to leave

It is a very valid concern Digital marketers have spent 20 years

getting

used to pixel perfect tracking They

want to read the transcript They want to analyze the sentiment

of every single lost lead They want to know exactly which adjective caused the user

to

abandon the shopping cart Exactly But

they

cannot do that here No The data is

locked inside OpenAI's servers Because OpenAI is prioritizing user

privacy

over merchant visibility Right They have

to make this choice They want users

to trust the platform Because if users think every single chat is being handed over to

a corporate sales team they will just

stop using the agents They will

fear the spam They will fear

the retargeting ads I understand

the privacy angle

I really do But we also do not know what

those interaction

thresholds

are Right OpenAI has not disclosed the specific volume required before a business sees those recurring

discovery topics We do not know if it is 10 chats or

10

000 chats Exactly A small boutique might never

hit the threshold They might set up this agent sponsor it and just see a

blank dashboard forever Because the metrics are aggregated

to

protect privacy Yeah If a store only has three

chats a day showing the topics might identify

the specific users So the dashboard remains dark

until

volume increases Exactly And this makes early optimization incredibly difficult

for the merchant It creates a customer service channel with defined frustrating

limits on measurement

It really does You can

bill the agent You can pay

to promote it But you cannot monitor

its individual conversations You cannot quality control

the

specific interactions Precisely The dashboard helps businesses understand how they

are

broadly represented It highlights high level opportunities to improve

the product feats But the underlying specific customer conversations remain a

total black

box They do This is a

major tension point So what listeners should watch next is

the expansion timeline Yes We need to see when OpenAI will

expand agent access Right They need to move beyond U S

e commerce and online services We must watch for

expansion into other global markets We also need to watch for expansion

into more complex business categories Like

will they allow financial services

Will

they allow travel booking Right And we should also watch if they develop

better analytic

tools for organizations Because businesses will definitely demand more

complex data access OpenAI did

promise future tools but they attached no

timetable to that Right The tension between user privacy and merchant data needs will really define

this product's ultimate success Because if merchants cannot measure

the return on investment they will just

stop paying for sponsorships It is

a delicate

balancing act It is We will keep a close eye

on those data thresholds In other news we are seeing massive financial scale ups

across the industry

right now We really are OpenAI is scaling its product ecosystem Meanwhile a major

rival is scaling its capital structure

to unprecedented heights Let us

talk about Anthropic

Yes let us Because this is a

story of staggering

almost incomprehensible

numbers It is Anthropic just set its proposed

IPO price range The filing dropped

on

November 10th Okay So what

are they targeting They are targeting 70 to

78

per share Wow The scale of this

offering is dizzying

It really is We need to break down

exactly what happened Because this is not just a

standard public offering No

it is not

Alongside the IPO there is a massive private

purchase

happening simultaneously Right Tech giants

are stepping in Yes They are buying 26 billion worth of non

voting Class

N shares 26 billion And they

are executing this purchase exactly

at the IPO price Let us look at

the roster of buyers It is basically a who's who

of global technology

infrastructure It really

is We have

NVIDIA We have

Microsoft We have

Google

We

have TSMC

Broadcom AMD and Micron These are the companies that literally

build

the physical internet Right

They manufacture the chips They

run the data centers They supply

the

global compute capacity Exactly So why does this matter

beyond

the obvious headline Well

it is the valuation The sheer scale is almost hard to comprehend for

a company this young Let us

run the math carefully On a

simple outstanding share basis

what is the valuation It is about 1 03

trillion A trillion

dollars

Yes And that assumes the underwriters exercise their full option to buy additional

shares Right Axios estimates roughly 13 09 billion shares would be outstanding across all classes at the top price of 78 Which pushes them

cleanly over the trillion dollar mark But that is

not even the fully diluted number No it is

not

The fully diluted

number is much much higher Because

you have

to include employee stop options Yes You have

to

include restricted stock units And you have

to include convertible securities Exactly When you

factor

in all those potential shares the valuation just

explodes It really does

Yahoo Finance puts

the fully diluted estimate as high as 1 45 trillion A valuation

of 1 45 trillion requires incredible underlying business metrics It does

You cannot

justify that with just hype No But

the

financial reporting does show incredible growth The business

metrics look very strong to institutional marketers They do

Anthropic reported 33 2 billion in revenue for the first nine months of

2026 33 billion in nine

months That is astonishing

scale It represents massive

acceleration

They are scaling revenue faster than almost any enterprise software company in history

And they are doing it with incredible efficiency

Yes They are boasting gross margins above 85

Those margins are fascinating

I mean 85 is a classic software margin

It is What do you expect from a company selling lightweight cloud applications But

AI is incredibly hardware intensive The

compute costs are astronomical Right

Achieving

software like margins while paying for AI compute implies extreme pricing power It

implies they are charging a massive premium for their models Absolutely

Let

us talk about the people behind this The

founders are sitting on a gold mine They

are The

seven co founders plan to sell roughly 250 million worth of shares each during this offering

Wow

And this calculation is based on the midpoint price of 74

So they are taking significant chips off the table They are securing generational wealth

But they are not giving up control of the company

This is a crucial structural detail Right

The founders will retain a collective 51 1 voting

control And they achieve this absolute majority via super voting class

F

shares Yes Super voting shares give them multiple votes per

single share Right It separates economic ownership from

voting power You can

own it tiny fraction of the

actual equity but still dictate the direction of

the board And no individual

founder

dominates The controlling position belongs to the group of seven acting

together

Exactly This structure completely isolates them from activist

investors The public markets cannot force a change in

leadership Right Now here's where it gets

highly

concerning Okay We really need to discuss the caveats There is a completely unexplained share count discrepancy in the

financial

reporting Yes it is highly unusual for an IPO of

this magnitude The financial press

is

confused Yeah The filings are

extremely dense

They are Yaku Finance counts 825 million shares

being

offered Okay And that implies 64 35 billion

raised in the public market

But other analysts see very different numbers

Right

Axios counts 801 4 million shares And that implies 62

5 billion raised Neither source explains

the difference They are reading

the exact same SEC documents They are And

the discrepancy

is over 23 million shares That is a

nearly 2 billion rounding error It is massive You

do not just

lose track of 2 billion in a public filing No you do not It

suggests a fundamental misreading

of a

specific share class by one of the reporting outlets Or it suggests a genuine error in

the prospectus tables Right Furthermore there is another massive number

that

requires

precise definition here What is that Anthropic disclosed 895 4

billion in committed

spending Almost 900 billion That is approaching

the GDP of

a small country It is This spending is primarily

dedicated to

computing and infrastructure But we must be very precise

with the language here This is committed spending Right It

is a contractual promise

for

future payments It is not cash that was

already spent or incurred during those first nine months No it is a massive liability

hanging over the balance sheet They have promised to buy that

much compute over the coming years Exactly Let us go back to

that

26 billion private purchase I have some serious pushback here

Okay Regarding the tech giants buying

the

non voting shares Yes These giants are buying

in

at the IPO price But they are doing it

without the standard six month resale lockup Right Yahoo Finance explicitly

reported

that detail And it means the

terms for the tech giants wildly differ from the terms for everyday public investors They really do We

need to explain

why this is so aggressive A lockup period is

standard practice It is It prevents

insiders from

dumping their shares on day one and crashing the stock price It protects the retail

investors But these giants are

exempt Are they essentially hedging

their bets on the AI race It certainly looks like

a strategic hedge Right They are funding

Anthropic

to ensure the ecosystem grows But they retain a

ripcord Exactly They can dump

those

shares immediately if the market turns sour While public

investors are basically

trapped Right If the AI bubble

pops

on day two retail buyers cannot sell for

six months But the giants can liquidate

their positions immediately It raises a very important

question about market stability It does Look at the

buyers again

NVIDIA Microsoft They are securing

their

positions

in

the broader AI ecosystem Right They either supply the compute

to

Anthropic or they compete with Anthropic directly Buying non voting shares keeps them

involved financially It gives them a piece of the upside But the lack of voting

rights prevents antitrust regulators from blocking the deal based on governance control It is a brilliant legal

maneuver It is But the lack

of a lockup is a massive structural advantage Right It changes the risk

profile

completely for these specific buyers It feels like they are

playing a completely different game on a completely different board They are They are dictating

terms because

Anthropic desperately needs their hardware So what listeners should watch

next is the confirmed NASDAQ trading date Right The company will trade

under

the ticker symbol A N T We also desperately

need to watch for any SEC amendment Yes Anthropic must

resolve

that missing 23 plus million shares discrepancy before trading actually begins Because the market

will demand absolute clarity You cannot launch

a trillion dollar IPO with a two billion dollar typo in the cap table Institutional investors will

halt their orders until the math is perfect We

will watch for that

amendment closely Next up we are moving

away from the massive enterprise scale of trillion dollar models We are pivoting down to

the

hyper

personal scale of AI We are looking at intelligence sitting

right on your wrist We are looking at meta Yes

On November 9th 2026 Meta

aggressively

expanded its personal

AI

ecosystem They brought their Muse personal agent

to smartwatches It is now available on both

the Apple Watch and Android's Wear OS platforms Let us

outline exactly what happened here

Well users can now interact with the

Muse agent directly from their wrist They can use text input Right And

they can use voice dictation

They

can even initiate full voice calls

with the AI The Apple Watch version arrives smoothly through

a standard update

to the existing Muse iOS app And Apple Watch users also get a dedicated

watch face complication Right A complication is a small widget on

the main clock screen It acts as a quick access shortcut

One tap and the AI is listening

So why does this matter beyond the

simple headline of a new app It is fundamentally about the nature of

this specific agent Exactly We really need to distinguish Muse

from

older technologies Muse is not just a simple chat

bot No it does not just answer

trivia questions or set cooking timers like the original Siri Muse is an action oriented agent

It executes complex workflows It

is deeply integrated across

Meta's entire platform Right It talks to Instagram

It

talks to Facebook It connects to external services

It manages your emails It

books restaurant reservations It

orders groceries for delivery

It sets fitness goals and

makes actual financial purchases Moving those specific complex capabilities

to the wrist is a massive behavioral shift

It really is It radically

reduces the friction

of initiating these tasks Think about the current workflow

for ordering groceries Right You have to locate