AI is quietly ripping up the online platform rulebook, shifting power from aggregators and marketplaces to AI agents that sit on the customer’s side of the table. What began as “better search” is fast becoming “do it for me commerce” across travel, ecommerce, automotive, real estate, insurance, health, NBFC lending, and more, putting the classic portal and marketplace model under real pressure.
The big shift: from platforms to personal agents
Two structural shifts define this moment.
- Discovery is moving from category platforms and comparison sites to horizontal AI assistants that understand intent, history, and constraints across a user’s entire life.
- Transactions are moving from app-centric flows to delegated, autonomous agents that shortlist, compare, negotiate, and execute on behalf of the customer.
The “front door to the internet” is no longer a portal homepage or app icon but an AI layer that decides which platforms even get a chance to participate. In that world, owning traffic matters less than being indispensable to the agent.
Automotive: search, configure, negotiate, close
Automotive journeys have long been fragmented across research sites, OEM configurators, dealer platforms, and financiers. AI is stitching this into one continuous, agent-led experience.
- Unified research-to-decision: AI can synthesise reviews, spec sheets, incentives, TCO models, charging/fuel availability, and ownership risks into personalised recommendations for a specific driver profile and use-case.
- Cross-platform negotiation: The same agent can ping multiple dealer inventories, compare on-road cost, and even negotiate within pre-set boundaries before presenting a shortlist.
- Embedded post-purchase: After purchase, agents can manage insurance, servicing schedules, warranties, and resale timing, keeping the user in an agent-led loop rather than any one platform’s app.
Old-school automotive platforms that treat each funnel step as a separate property are being collapsed into “data feeds” for multi-brand AI advisors that run the full journey.
Insurance: agents for risk, not just policies
Insurance platforms historically monetised complexity and opacity. AI is compressing both.
- AI-first underwriting: AI underwriting engines are automating a large share of risk assessment, often cutting manual effort by more than half and shrinking turnaround from days to minutes.
- Agentic servicing and claims: Agentic AI can ingest proposals, declarations, and claims documents, auto-validate cover, flag anomalies, and route cases, with carriers reporting faster settlement and lower fraud in live pilots.
- Invisible platforms: As customers delegate “keep my risk optimally covered at the best price” to AI, the primary relationship shifts from any one insurer portal to the agent that continuously quotes, switches, and rebalances cover across carriers.
Insurers that behave like clean, agent-friendly infrastructure win more than those behaving like walled-garden portals.
Health insurance: from portals to health coverage copilots
Health insurance is where the stakes of this shift are most visible: the interface is no longer just about money, but about care.
- Smarter underwriting and pricing: AI is enabling much more granular health risk models, using medical histories, lifestyle data, claims patterns, and sometimes wearable data, to produce more precise pricing and product design.
- Agentic claims and navigation: Agents can read prescriptions, diagnostic reports, and hospital invoices, validate coverage, estimate out-of-pocket costs, and route claims, with early deployments showing significant reductions in claim cycle times and leakage.
- Proactive optimisation: Generative AI systems push personalised wellness nudges, preventive care reminders, and plan optimisation suggestions, nudging members to care pathways that reduce long-term risk and cost.
When AI “shops” for your care, it is effectively managing a dynamic portfolio of policies, hospitals, and treatment pathways on your behalf. That makes health insurers’ real customer—the entity evaluating trade-offs—an algorithm, not a human logging into a portal.
NBFCs and digital lending: AI-native credit rails
NBFCs, especially in India, are becoming AI-native lenders, and agentic finance intensifies that trajectory.
- AI-native underwriting: Digital NBFC stacks increasingly use ML across bank statements, GST data, bureau files, cash-flow signals, and alternative data to automate a majority of decisions and drastically cut underwriting times, especially for MSMEs and thin-file borrowers.
- Continuous risk and collections: Agentic AI monitors portfolios in real time, flags anomalies, and automates collections workflows, moving portfolios closer to “self-healing” with early, tailored interventions.
- Embedded, agent-led credit: Personal finance agents can automatically shop credit lines across NBFCs, refinance high-cost loans, or trigger top-ups based on cash-flow stress, without the user visiting any particular lender app.
Here again, the power shift is from “who owns the lending app” to “whose products and risk models are easiest for third-party agents to integrate and trust.”
Ecommerce: marketplaces vs agentic commerce
Marketplaces assumed consumers would keep searching, comparing, and filling carts. Agentic commerce assumes the opposite: consumers will outsource that work to AI.
- End-to-end shopping agents: AI agents can now discover products, compare across retailers, check reviews and returns, negotiate discounts, and complete checkout using stored preferences and payment details, often while the consumer is offline.
- Always-on optimizers: On the supply side, AI agents are running campaigns, reallocating ad spend, tuning pricing, and managing catalog content in real time, tightening feedback loops far beyond what human teams can manage.
- Marketplaces in the background: Analysts warn that if marketplaces do not respond, they risk becoming low-margin fulfilment utilities living “behind” AI interfaces, rather than destinations consumers consciously visit.
When AI shops for you, marketplaces risk becoming invisible infrastructure unless they learn to compete for the agent’s trust, not just the user’s clicks.
Real estate: portals under pressure
Property portals grew by aggregating listings and monetising leads. AI is attacking both search and brokerage workflows.
- Intent-level property search: AI can infer latent preferences (light, noise, commute patterns, school choices) from behaviour and context, then surface properties that match the true need, not just the filters people remember to click.
- AI agents for agents: On the supply side, AI is qualifying leads, scheduling visits, writing listings, and predicting pricing and rental risk, cutting mundane workload by more than half for many real estate teams.
- Owning the relationship: If the buyer’s primary interface is an AI advisor that spans portals, brokers, and lenders, portals become interchangeable pipes unless they differentiate through richer, more accurate, more structured data.
The gravitational centre moves from “largest listing pool” to “richest machine-readable signals that AI advisors can understand and act on reliably.”
Travel: OTAs’ moat is leaking
Online travel agencies (OTAs) were built on aggregating inventory, arbitraging performance media, and owning search and discovery. That stack is being scrambled fast.
- AI discovery over links: Generative AI now answers rich, multi-constraint queries (“five days in Japan with kids, no red-eyes, under a fixed budget”) in a single conversational response, compressing the research phase that used to span dozens of OTA and airline tabs.
- Agentic booking flows: New AI travel agents can check live availability, compare prices across OTAs and direct channels, reconcile with calendars, and complete bookings in one flow, shrinking decision cycles from weeks to minutes.
- Intermediary risk: Hotels and airlines that expose clean APIs, rich content, and transparent rules become preferred rails for agents, weakening the classic OTA toll-booth model.
In an agent-first travel world, the winning “platform” is the one whose data and APIs are easiest for third-party agents to consume and trust, not the one that shouts loudest in paid search.
How AI is rewiring platform power
Across categories, the same structural pattern is playing out.
| Dimension | Old platform logic | AI-scrambled reality |
| Discovery | Users search inside category platforms. | Users ask horizontal AI agents that decide which platforms get surfaced. |
| Differentiation | Brand, UX, and breadth of inventory. | Quality, richness, and structure of machine-readable data and APIs. |
| Monetisation | Ads, commissions, lead-gen, traffic rents. | Revenue via agent integrations, data access, and execution rights. |
| Competition | Platform vs platform in a vertical. | Platforms vs AI agents for customer relationship ownership. |
| User behaviour | Multi-tab comparison, manual filtering. | Single conversational thread where the agent does comparison and execution. |
This changes what it even means to “build a platform.” Being the place users go to may matter less than being the infrastructure agents must go through.
Strategic moves for CX, product, and growth leaders
For leaders building next-gen CX and commerce models, the critical question is no longer “Should we add AI?” but “How do we redesign our platform so external agents cannot ignore us?”
- Instrument for agents, not only humans: Products, prices, inventory, rules, and policies must exist as clean, consistent, machine-readable objects that an AI can parse and act on, without scraping brittle UIs.
- Design for GEO, not just SEO: Shift from classic search engine optimisation to “generative engine optimisation” so your brand shows up inside AI answers, itineraries, and plans, not just in blue-link rankings.
- Make peace with disintermediation: Assume some direct user touchpoints will be replaced by agents; focus on value in data, fulfilment, risk, and post-purchase experiences that are hard to commoditise.
- Build your own trusted copilots: Embed first-party copilots into apps and products, so when external agents negotiate or orchestrate, your own AI is negotiating on your behalf with full context and governance.
For old-school platforms, the uncomfortable truth is this: when AI shops for your customer, the UI is no longer your homepage—it is an agent deciding, in milliseconds, whether you deserve a seat at the table. The winners in this new era will be the brands and platforms that stop optimising only for human eyeballs and start architecting for machine understanding, trust, and autonomous action—before someone else’s AI quietly shops them out of the journey.
