An agent can translate a user’s preferences into explicit constraints such as price, location, and required characteristics. This threatens emotional advertising because products may be filtered out before brand affinity or creative campaigns can influence the decision.

User prompts become machine-readable purchasing criteria
Agents can filter advertisements against explicit requirements
Functional fit may outweigh emotional brand preference
Consumer software buying may resemble enterprise procurement

“You have said to that agent, I want the cheapest.”
Kieran Flanagan · 04:00

“it reduces everything to a pre-programmed list of functional specs.”
Kieran Flanagan · 05:30

#product specs#brand#procurement#advertising

Hot Take10:30

The Agent May Replace the Human Across the Whole Funnel

The envisioned agent consumes advertising, conducts organic research, signs up for software, and operates the product. Humans move above the interface while marketers and go-to-market teams increasingly design experiences for agents as the primary users.

One agent could span discovery, research, signup, and product use
Humans may stop visiting vendor websites and operating software directly
Go-to-market work may increasingly serve automated agents
Products and websites may need agent-first interfaces

“The agent has taken my place.”
Kieran Flanagan · 10:30

“in the future, a lot of the things that we may do may be to service agents.”
Kieran Flanagan · 10:30

#go-to-market#agent-first#software#marketing

Explainer· 4

Explainer01:30

AI Agents Could Become the New Advertising Audience

Advertising may shift from persuading humans to supplying information that attracts and satisfies AI agents. Merchants would compete for an agent’s attention while the human user remains insulated from the ads.

AI agents may consume ads on behalf of users
Brands could compete for agent attention instead of human attention
Users may receive purchasing outcomes without seeing advertising

“The human never consumes the ad in and of itself.”
Kieran Flanagan · 02:00

“They're competing for the agents' attention.”
Aravind · 02:00

#ai agents#advertising#attention#marketing

Explainer02:30

AI Agents May Run the Entire B2B Buying Process

Kieran predicts that agents will evaluate, purchase, and begin using business software for users. Advertising would therefore target the system conducting procurement rather than the person who ultimately benefits from the product.

Agents could select and purchase B2B software
Procurement may occur with little direct human involvement
Marketing information must become legible to automated buyers

“in the future I can see an agent going through the full B2B procurement process, going through how to choose software, buy software, and actually…”
Kieran Flanagan · 02:30

“the agent is the thing that is doing the procurement on behalf of the user.”
Kieran Flanagan · 03:00

#b2b#procurement#software#ai agents

Explainer05:30

Enterprise Procurement Offers a Preview of Agent Commerce

Large enterprises already screen vendors against mandatory functionality before considering softer brand attributes. Kieran argues that agent-mediated commerce could extend this procurement pattern across the broader software market.

Enterprise buyers use mandatory functionality lists
Brand sentiment cannot compensate for missing requirements
Agent commerce could apply enterprise-style screening to all software

“If you don't have this spec list, this kind of list of uh functionality that you need to have to be considered in that procurement…”
Kieran Flanagan · 06:00

“this is basically saying that all software is going to be bought that way”
Kieran Flanagan · 06:30

#enterprise#procurement#software buying#brand

Explainer09:00

Multimodal Agents Could Learn Software by Watching Users

Screen-aware agents can guide people through software while observing how they work. Over time, that observation may let the agent move from assisting with a task to executing the task through connected tools.

Screen-aware agents can provide contextual product guidance
Agents can learn workflows from browser activity
Tool connections allow agents to perform jobs inside applications
Assistance may evolve into autonomous execution

“It can see your screen.”
Kieran Flanagan · 08:30

“it's learning how I use the software.”
Kieran Flanagan · 10:00

#multimodal ai#software#automation#agents

Story· 1

Story07:00

Deep Research Can Eliminate the Marketer’s Website Touchpoint

Kieran demonstrates how Google’s deep-research capability can inspect dozens of websites and return a consolidated tutorial. Because the user may never visit the cited sources, marketers lose opportunities to persuade prospects during the research journey.

An agent researched 37 websites for one task
The user received a comprehensive tutorial without browsing sources
Fewer website visits mean fewer opportunities for conversion
Citations can build trust without generating source traffic

“It researched 37 websites, right?”
Kieran Flanagan · 07:30

“I never have to visit any of these websites.”
Kieran Flanagan · 08:00

#deep research#search#web traffic#conversion

Tool· 1

Tool09:30

Model Context Protocol Turns Claude Into a Tool-Using Agent

Kieran presents Claude’s Model Context Protocol as an early indication of agents becoming active software operators. Connecting Claude to services such as Brave Search gives it tools that can be invoked to complete jobs rather than merely answer questions.

Model Context Protocol gives Claude access to external tools
Connected search expands what the agent can do
Tool access lets users assign jobs inside integrated services

“you can actually give Claude tools so I can connect it to Brave Search.”
Kieran Flanagan · 09:30

“because you've given it access to tools, you can ask it to do different jobs within those tools.”
Kieran Flanagan · 10:00

#mcp#claude#tools#automation

Takeaway· 1

Takeaway08:00

AI Citations May Increase Trust Without Sending Traffic

AI products cite their sources to reassure users that generated answers are grounded rather than hallucinated. Yet a sufficiently complete answer can satisfy the user without producing clicks for the cited websites.

Citations help users assess whether an answer is grounded
Source attribution does not guarantee referral traffic
Comprehensive AI answers can end the research journey

“people feel much more comfortable if it cites the resources so it knows it's not hallucinating.”
Kieran Flanagan · 08:00

“But I'm never going to go and look at any of these because I actually have a pretty comprehensive tutorial and it works really well…”
Kieran Flanagan · 08:30

#citations#trust#search traffic#hallucinations

Frameworks in this episode

Agent Replacement Ladder

Agent-Spec Procurement

Insights & moments

◆Hot Take· 2

AI Buying Could Reduce Marketing to a Functional Spec Sheet