AI Is Becoming the Interface. Is your Business Ready?
The developments worth paying attention to — and what they tell us about where digital discovery is going.
Hillary Bassett Ross — Founder, Kinetic Pencil
Hillary Bassett Ross is an AI search and web optimization strategist with over 15 years of experience across content strategy, SEO, UX, analytics and digital optimization for companies including Microsoft, Cisco and Liberty Mutual. She is the founder of Kinetic Pencil, where she helps marketing leaders understand what changes when AI becomes a primary layer between their brand and their customer—and how content, websites and digital ecosystems need to change with it.
For years, digital strategy has centered on a relatively familiar relationship: organizations create websites, applications, and content experiences, and people navigate those environments to find what they need.
AI is beginning to change that relationship.
Instead of visiting a company’s website to compare products, a customer may encounter an AI-generated comparison built specifically around their needs. Instead of searching an organization’s knowledge base, an employee may ask an agent to retrieve information and complete a task. And instead of simply providing an answer, an AI system may eventually take action using an organization’s tools and services.
The implications extend well beyond search rankings.
This week’s developments suggest that the next competitive challenge isn’t simply making your business visible to AI. It’s making your organization understandable, useful, and trustworthy when AI becomes the intermediary.
01The big one
The intermediary is becoming the interface, the worker, and the execution environment
Three major announcements this week illustrate how quickly the role of AI intermediaries is expanding.
OpenAI: AI-generated experiences
On October 7, OpenAI introduced Intelligent UI with GPT-6, enabling ChatGPT to generate interactive experiences around a user’s specific question or task.
Instead of responding exclusively with text, ChatGPT can assemble comparisons, calculators, diagrams, and interactive elements that help people explore information.
- comparisons
- calculators
- diagrams
- interactive elements
OpenAI also introduced progressive answering, allowing ChatGPT to begin responding while continuing to investigate a question.
This introduces an interesting shift.
Traditionally, businesses design the interfaces through which customers encounter their products, services, and expertise.
But what happens when AI designs that experience instead?
A company’s information may be reorganized into an interface it never created, alongside competitors it didn’t select, using evaluation criteria determined by the user’s request.
The organization’s underlying information becomes the raw material for someone else’s experience.
AI Search IntelligenceKinetic Pencil
Google: AI as an organizational worker
On October 8, Google announced a universal enterprise agent designed to work across applications, retain organizational context, delegate tasks, and execute extended workflows.
The architecture brings together organizational knowledge, persistent memory, reusable skills, connected tools, and agent identities.
- organizational knowledge
- persistent memory
- reusable skills
- connected tools
- agent identities
An important distinction is that the agent is not synonymous with the model powering it. Different models can potentially support the same persistent organizational workflows.
The universal agent remains in private preview, but its proposed architecture points toward a future in which organizational information isn’t merely retrieved. It’s used to perform work.
Microsoft: Defining what AI is allowed to do
Microsoft’s Execution Containers, which became generally available October 7, address another part of this emerging architecture: authority.
These containers establish technical boundaries around the resources an agent can access, including files and networks.
Rather than relying exclusively on instructions telling an agent what it should or shouldn’t do, enforcement occurs outside the agent itself.
Agent-specific identity integration is still forthcoming, but the direction is significant.
As AI systems move from providing information to executing actions, organizations need reliable ways to determine what those systems are permitted to do.
Connecting the three announcements
Together, these developments reveal three increasingly important dimensions of AI readiness:
Experience: How is organizational information presented and evaluated?
Execution: How can AI use organizational knowledge and capabilities to accomplish work?
Authority: What actions is AI permitted to take, and under what conditions?
This expands the conversation considerably.
We’ve been talking about making content accessible and understandable to AI systems. But the emerging requirement may be much broader.
Organizations need to make their knowledge, capabilities, evidence, and operational constraints understandable to systems they don’t necessarily control.
The Kinetic Pencil perspective
This isn’t necessarily a new set of digital requirements.
Organizations have always needed coherent information architecture, accurate product data, accessible content, connected systems, and appropriate governance.
What’s changing is how that information gets used—and how visible the consequences become when those foundations are missing.
AI systems can now retrieve information, reorganize it into new experiences, and increasingly use it to perform work.
That puts greater pressure on the quality, consistency, and accessibility of the information organizations already manage.
One emerging question is whether information that can be accurately reassembled across different AI-generated experiences will create a competitive advantage.
An organization’s differentiation has traditionally been expressed partly through the interfaces and experiences it designs. But as AI increasingly determines how information is presented, that differentiation may need to survive outside those proprietary environments.
This is worth investigating, although we don’t yet have evidence that organizations investing in this capability consistently outperform those that don’t.
The opportunity isn’t necessarily to invent an entirely new digital infrastructure. It’s to make existing digital foundations more adaptable to an environment where organizations no longer control every interaction.
Sources:OpenAI: GPT-6 and Intelligent UI for Everyone · Google: Gemini at Work 2026 · Microsoft: Execution Containers
02The signal
1. AI visibility isn’t one ranking. It’s a collection of conditions.
Two new commercial studies provide additional evidence that AI visibility is considerably more complicated than whether a brand appears in a particular response.
Rezolve AI examined 122,820 AI-generated answers involving 2,047 U.S. companies.
One of the most interesting findings was how dramatically brand appearances varied depending on what the customer was trying to accomplish.
Companies appeared in:
In other words, being visible when someone is exploring a category doesn’t necessarily mean being visible when they’re evaluating alternatives or preparing to purchase.
A separate study from AI Syndicate examined 21,600 local-business questions across 600 markets.
- 22%
Only 22% of markets had the same leading business across three different question formulations.
- 52%
The research also reported that the leading business changed in 52% of 245 comparable markets from one month to the next.
However, its methodological limitations make it difficult to determine how much of that movement reflects genuine changes in visibility rather than sampling variability.
These aren’t controlled measurements of actual consumer journeys. Both datasets have important limitations, including researcher-generated prompts, platform-access differences, and sampling uncertainty.
But they reinforce a useful principle:
AI visibility is conditional.
It depends on the platform, the question, the customer’s objective, the information available, and the circumstances under which the answer is generated.
- the platform
- the question
- the customer’s objective
- the information available
- the circumstances
That changes how we should think about measurement.
A single visibility score may be useful as a high-level indicator, but it cannot adequately explain how a business is represented across the customer journey.
A more meaningful assessment asks where a business appears, where it disappears, what evidence supports its inclusion, and how accurately its offerings are described.
For organizations investing in AI search, understanding those conditions may be more actionable than chasing a universal ranking.
Sources:Rezolve AI: U.S. AI Visibility Study 2026 · AI Syndicate: October AI Local Report
2. Information architecture is becoming business infrastructure.
Two enterprise examples this week illustrate why information architecture is moving beyond the website and marketing department.
At L’Oréal, Chief Digital and Marketing Officer Asmita Dubey described a broader restructuring of marketing, product information, data, and technology.
According to Dubey, approximately 20% of the company’s marketing investments are now AI-led.
Part of that transformation involves standardizing product information and improving how AI systems understand the company’s brands, scientific evidence, and consumer needs.
- 16,000+
Meanwhile, Barclays reported that its Claude-powered knowledge assistant has been adopted by more than 16,000 employees and handled over one million searches.
- 120,000
Another AI-enabled system processes approximately 120,000 incoming emails daily, helping classify and route operational requests.
These are company-reported adoption and activity figures, not independently verified productivity gains.
Nevertheless, the two examples reveal a common dependency.
L’Oréal is working to make organizational knowledge more understandable to external AI systems.
Barclays is making organizational knowledge more accessible to internal AI systems.
Different applications. Similar underlying challenge.
Both require information that is structured, accessible, maintained, and sufficiently reliable for the systems using it.
This is why I believe information architecture is becoming an increasingly important organizational capability.
The same foundations that support effective content discovery can also support customer service, employee knowledge, operational automation, and future AI integrations.
Good information architecture isn’t just about helping people find things anymore. It’s about helping an organization function across an expanding ecosystem of human and machine interactions.
AI Search IntelligenceKinetic Pencil
Sources:Fortune: As AI Reshapes the Consumer Journey, L’Oréal Is Rethinking Its Marketing Engine · Anthropic: Barclays Scales Claude
3. How AI explains information may influence what people decide.
A new peer-reviewed experiment offers an interesting perspective on the relationship between AI-generated answers and human decision-making.
Researchers randomly assigned 417 participants to interact with either a standard chatbot or one that exposed structured reasoning.
The study focused on sustainable consumption decisions, where participants needed to evaluate information involving uncertainty and personal values.
Participants using the reasoning-enabled chatbot reported greater perceived knowledge and stronger behavioral intentions.
The analysis also identified relationships involving trust, perceived usefulness, personalization, and reduced cognitive load.
There are important boundaries here. The experiment measured intentions rather than actual purchasing behavior, and the proposed psychological mechanisms weren’t independently randomized.
Still, the finding raises a significant question for content strategy.
What if visibility is only part of the equation?
A business might appear in an AI-generated recommendation, but how the system explains its relevance could also influence how people understand and evaluate that recommendation.
This suggests a broader role for content architecture.
It’s not enough for information to be discoverable. The evidence, claims, distinctions, consequences, and trade-offs need to be understandable as well.
- evidence
- claims
- distinctions
- consequences
- trade-offs
Content should be structured not only to be found, but to support accurate reasoning.
AI Search IntelligenceKinetic Pencil
Source:International Journal of Information Management: Opening the Black Box
03The data point
90% want autonomous IT. 77% hesitate to surrender control.
A new Riverbed survey of 1,200 business and technology professionals across seven countries reveals a striking tension in enterprise AI adoption.
- 90%
expect to pursue human-supervised or highly autonomous IT operations within two years.
- 77%
report organizational hesitation about allowing AI to make operational decisions without human approval.
- 17%
Only 17% report fully unified visibility across their technology environments.
- 9%
And just 9% report AI initiatives fully deployed enterprise-wide.
The ambition is substantial.
The readiness is less certain.
The findings are self-reported and come from vendor-sponsored research, so they shouldn’t be interpreted as independent evidence of operational maturity.
But they highlight an important distinction between wanting AI capabilities and being prepared to delegate authority.
A separate AP-NORC poll published October 8 found that 64% of American adults believe AI is developing too quickly.
That probability-based survey interviewed 2,140 adults between September 24 and 28.
These surveys represent different populations and cannot be directly compared. Together, however, they suggest that questions about control remain central to the broader adoption conversation.
For organizations, that means measuring AI adoption requires more than tracking usage.
It also requires understanding where people are comfortable accepting recommendations, where they require approval, and where autonomous action is appropriate.
- accepting recommendations
- require approval
- autonomous action
Sources:Riverbed: Global Survey Finds Agentic AI Reshaping IT · AP-NORC: Most Adults Think AI Is Developing Too Fast
04Connecting the dots
The next frontier of AI readiness is organizational understanding.
There’s a common thread running through this week’s research and announcements.
AI systems are increasingly positioned between organizations and the people who need their information, products, services, and capabilities.
That intermediary role introduces new questions.
First: Who shapes your reputation?
The AI-visibility studies suggest that organizational representation changes with context.
Your company may appear prominently in one type of question and disappear in another.
And much of the information AI systems use may come from outside your own digital properties.
- 1.5%
Rezolve’s research found that just 1.5% of attributable citations came from the specific company website being tested.
- 11%
AI Syndicate found an 11% identifiable owned-website share in its local-business dataset, although the figures aren’t directly comparable because the studies used different methods.
This doesn’t mean only 1.5% of citations came from company-owned sources overall, and Perplexity accounted for a disproportionate share of the study’s attributable citations.
Nevertheless, they raise an important possibility.
An organization’s machine reputation may be constructed from a distributed evidence ecosystem, with different information becoming relevant under different circumstances.
That suggests reputation management in AI search may require a broader perspective than optimizing owned digital properties alone.
It also creates an interesting research opportunity: determining whether stronger, independently verifiable evidence across multiple sources leads to more consistent and accurate organizational representation.
Second: Who shapes the customer experience?
With generated interfaces, AI can potentially decide how information is organized, compared, and explained.
The behavioral research suggests that explanation itself can influence users’ perceptions.
That makes the quality and clarity of underlying evidence increasingly important.
An organization may not control the interface through which someone evaluates its products or services.
But it can influence the quality of the information available for that evaluation.
This is another reason to revisit content fundamentals: clear differentiation, credible claims, supporting evidence, accurate product information, and accessible explanations of value.
Third: Who performs the work?
Enterprise agents are beginning to connect organizational knowledge with tools, applications, and workflows.
But making information accessible isn’t the same as making actions safe.
Organizations also need to establish permissions, constraints, and governance.
Together, these developments suggest an emerging sequence:
- Organizational reality
- Structured knowledge
- Accessible evidence
- AI reasoning
- Generated experience
- Authorized action
Each transition creates an opportunity for meaning to be preserved, distorted, or lost.
This is a useful way to expand our definition of digital readiness.
The question is no longer exclusively whether machines can find your organization.
It’s whether they can understand it accurately enough to explain it, represent it, and potentially work with it.
AI Search IntelligenceKinetic Pencil
An important economic distinction
Two recent experiments add another dimension to this conversation.
- 37,561
A September 30 randomized field experiment involving 37,561 Washington Post readers found that AI-assisted search within the publisher’s own environment increased search frequency and modestly increased article consumption per reader, even as consumption per search declined.
- 1,100
Meanwhile, an August preregistered experiment involving 1,100 Google Search users found that removing AI features increased publisher click-through, while an AI Mode-only experience reduced referrals and worsened reported user experience and trust.
The studies examined different environments and shouldn’t be treated as directly comparable.
But they raise an important possibility.
AI may encourage certain forms of information exploration while changing where that exploration happens—and who captures its economic value.
For businesses and publishers, the distinction between AI operating inside their own environments and AI operating as an external intermediary deserves much closer attention.
The strategic implications could be substantial.
If an organization can provide AI-assisted exploration within its own environment, it may retain more control over the customer experience and the value created by that interaction.
When exploration happens through a third-party intermediary, the organization may become an information source without receiving the same level of direct engagement.
That doesn’t mean organizations should avoid third-party AI platforms. It means the relationship between discovery, engagement, and value capture needs to be examined more carefully.
Sources:Rezolve AI: U.S. AI Visibility Study 2026 · AI Syndicate: October AI Local Report · Breaking News Out of the Filter Bubble: Washington Post AI Search Experiment · AI in Search Reduces Publisher Referrals Without Improving User Experience
05On my radar
AI-generated interfaces and brand influence.
As generative interfaces become more capable, I’ll be watching for controlled research examining how they influence product comparisons, brand recall, trust, choice quality, and actual purchasing decisions.
Engagement alone won’t tell us whether these experiences improve decision-making.
The next generation of AI measurement.
Current visibility datasets are useful, but we need stronger longitudinal research using real consumer questions, repeated measurements, retrieval-method distinctions, and business outcomes.
Understanding where and why visibility changes is more valuable than simply counting appearances.
Agent permissions and operational readiness.
The enterprise agent conversation is moving beyond what AI can accomplish toward what it should be authorized to do.
The distinction between demonstrated capabilities and promised future functionality will be particularly important, especially as organizations begin delegating consequential tasks.
Organizational knowledge as infrastructure.
Product information, content, internal knowledge, evidence, permissions, and operational capabilities have traditionally lived in different systems.
I’ll be watching whether AI accelerates their convergence into a more coherent architecture—or whether organizations continue managing fragmented information across disconnected platforms.
The human delegation boundary.
Where do people want AI assistance, and where do they insist on retaining control?
Understanding that distinction may prove just as important as improving technical capabilities.
Research that separates willingness to receive recommendations from willingness to authorize actions will be especially valuable.
06The bigger picture
From digital presence to machine-readable organizations
For much of the internet’s history, organizations have invested in creating destinations.
Websites, applications, portals, content libraries, and digital experiences were designed to attract people and help them accomplish specific tasks.
Those destinations aren’t disappearing.
But they are becoming part of a larger information ecosystem in which AI systems can retrieve, interpret, reorganize, and increasingly act upon organizational information.
That requires us to rethink what makes an organization digitally ready.
Discoverability remains important. So does content quality, technical accessibility, and a coherent digital presence.
But the emerging requirements extend further.
Organizations need to be understandable enough for AI systems to accurately represent them, structured enough for those systems to work with their information, and governed enough to support appropriate action.
- understandable enough for AI systems to accurately represent them
- structured enough for those systems to work with their information
- governed enough to support appropriate action
The fundamentals of good digital strategy haven’t disappeared. If anything, they’re becoming more important.
Clear content, thoughtful information architecture, reliable data, connected systems, and sound governance remain essential.
The difference is that these foundations now need to support a growing range of human and machine interactions.
Organizations that have invested in getting the fundamentals right may be better positioned to adapt. Those that haven’t may find AI exposing weaknesses that were already there.
We don’t yet know which architectures will deliver the greatest competitive advantage. Nor do we know how quickly consumers and enterprises will embrace the more autonomous capabilities now being introduced.
But the direction of platform development is becoming clearer.
The next generation of digital infrastructure won’t simply help people navigate organizations. It will help machines understand how those organizations work.
And that may be one of the most consequential shifts in digital strategy since the emergence of search itself.
Kinetic Pencil Intelligence examines the research, platform developments, and behavioral changes shaping AI search, digital strategy, and organizational adaptability. We distinguish verified capabilities from emerging hypotheses—and focus on what those developments mean for building better digital foundations.