Q1 2026 report · Free PDF

The £4 Trillion Conversation

How AI is Making E-Commerce Feel Human Again

Published by Firney,

24 pages · No form, no email needed

Cover of the Firney report The £4 Trillion Conversation: How AI is Making E-Commerce Feel Human Again. A young man takes a selfie, surrounded by chat bubbles asking an assistant to book flights and a hotel to Lisbon under £500, whether a dog can go on the ferry, and to show brown hiking boots made from sustainable materials.
  • 90%

    of B2B buying will be AI-agent intermediated by 2028 (Gartner)

  • 3x

    higher conversion rate for users who engage with AI-powered search (Commercetools)

  • 43%

    of Gen Z start product searches on TikTok (Connect Media)

  • 14%

    trust AI for autonomous purchases

Key findings

As of January 2026, the traditional e-commerce playbook is obsolete. Here are the five takeaways from our report.

  1. The discovery model is broken

    Twenty years of forcing consumers to “speak machine” through keywords has hit diminishing returns. Incremental UX optimisation no longer creates real value.

  2. Gen Z demands conversation, not search

    43% start product searches on TikTok, 62% prefer visual search, and they’re “experience-loyal.” They ask “How do I fix it?” not “Best mountain bike chain lube.”

  3. AI agents deliver measurable ROI

    Klarna’s AI handles 2/3 of customer service chats, doing the work of 700 agents and driving £32M in profit improvement. Users engaging with AI-powered search convert at 3x higher rates.

  4. Trust is graduated, not binary

    65% trust AI for price comparison, 44% for product recommendations, but only 14% for autonomous purchases. Transparency (like Stitch Fix’s approach) can increase AOV by 40%.

  5. 2026 is the year of “zero-click” commerce

    By 2028, 90% of B2B buying will be AI-agent intermediated. Brands need machine-readable data and real-time inventory APIs, or their agents will “hallucinate.”

The shift from answer engines to action engines

We stand at the precipice of a £2.4 trillion to £4 trillion (~$3T to $5T) global shift in revenue orchestration (McKinsey). AI is no longer a tool for information retrieval; it is a workforce for execution.

  • The workforce redefined: In sectors like fintech, single AI assistants now handle the workload of 700 full-time human agents, slashing resolution times from 11 minutes to 2 minutes (Klarna).
  • The visual imperative: Traditional search volume is predicted to drop by 25% by 2026 as consumers migrate to multimodal interfaces (Gartner). “Camera Natives” now expect to search with images, not words.
  • The infrastructure of autonomy: 2026 is the year of AP2 (Agent Payments Protocol) and ADK (Agent Development Kit). We are moving toward “Zero-Click” shopping where agents both request and fulfil complex shopping requirements and replenishment loops autonomously.

The journey from “researchers” to “impatient discoverers”

  1. Phase 1: The indexical era (2000–2023)

    We taught humans to speak “machine.” We forced consumers to break their desires into keywords. We built SEO strategies to rank for fragments, and the burden of synthesis lay entirely on the user.

  2. Phase 2: The generative interlude (2023–2025)

    We introduced chatbots, but they were largely “polite librarians.” They could summarise text and retrieve information, but they were reactive. If the user stopped typing, the value stopped flowing.

  3. Phase 3: The agentic era (2026 and beyond)

    We are moving from “finding” to “fixing.” The new paradigm is “Agentic Commerce,” where systems reason, plan, and execute workflows to solve complex user needs.

Old playbook vs. new reality

A structural comparison of the old way (rigid) and the agentic way (fluid).

  • B2B qualification

    The old way (rigid): Static forms; Typical 42-hour wait.

    The agentic way (fluid): Instant agentic qualification.

  • B2B discovery

    The old way (rigid): Keyword guessing; paginated product grids.

    The agentic way (fluid): Understanding the “intent” or “vibe.”

  • Search mode

    The old way (rigid): “Rusty window track.” (Passive)

    The agentic way (fluid): “How do I fix a stuck window?” (Active)

  • Retention

    The old way (rigid): Fixed “set and forget” subscriptions.

    The agentic way (fluid): AI-driven Predictive Replenishment.

  • Lead generation

    The old way (rigid): Extracting data via gates.

    The agentic way (fluid): Building trust via consultative dialogue.

  • Outcome

    The old way (rigid): High abandonment; “0 results.”

    The agentic way (fluid): +30% to +40% CVR lift via Visual Search.

Redefining performance benchmarks

In the conversational economy, traditional metrics like click-through rates are becoming vanity metrics. The new benchmarks are rooted in interaction quality and outcome efficiency.

  • 340%

    The increase in repeat purchase revenue achieved by brands switching from generic reminders to AI-backed predictive schedules.

    Source: Replenit

  • 217%

    The increase in engagement when personalisation is driven by AI that understands a customer’s stated needs.

    Source: Singlegrain

  • 3x

    The higher conversion rate for users who engage with AI-powered search compared to those who do not.

    Source: Commercetools

  • 60%

    How much more likely users of generative shopping assistants (such as Amazon Rufus) are to purchase compared to non-users.

    Source: Core Blueprint

  • £32 Million

    The estimated annual profit improvement driven by a single integrated AI assistant in the fintech sector.

    Source: Klarna

  • 78%

    The reduction in return rates achieved by guiding B2C customers to the right product the first time via conversational consultation.

    Source: Retainful

The Gen Z paradigm: winning the “impatient discoverer”

To win the fastest-growing customer segment, we must recognise their unique psychology. They don’t want to describe a look; they want to show it.

  • Social-first discovery

    43% of Gen Z start product searches on TikTok rather than Google (Connect Media). They value “Social Listening” and cultural relevance over statistical popularity.

  • The camera is the keyboard

    62% of Gen Z and Millennials prefer visual search over any other retail technology (Visual Search vs. Text Search Strategy).

  • The search-to-spend ratio

    Users who interact with site search spend 2.6x more than those who simply browse (Algolia).

  • The trust paradox

    They are the most willing to use AI (70% use AI tools for shopping), yet the most demanding of transparency. While 32% of Gen Z are comfortable with agents making purchases, only 12% fully trust it without oversight (GetCoAI).

Navigating the trust gap

Adoption is a gradient that shifts based on the risk profile of the task.

  • High trustAnalytical support: Tasks like price comparison across stores are viewed as objective and verifiable.

    65%

  • Moderate trustAdvisory services: Subjective territory like product recommendations where “hallucination anxiety” lingers.

    44%

  • Low trustExecutive agency: Authorising an agent to place an order without human supervision.

    14%

The trust gap: the share who trust AI with each kind of task, from high trust to low trust

Want to know how AI assistants see your site?

Results in a few minutes

Free GEO audit

2026 forecasts

By 2027, the primary interface for many brands will be an API called by a bot, not a screen viewed by a human.

  • 90% AI intermediated: 90% of B2B buying will be AI-agent intermediated by 2028 (Gartner).
  • Zero-click shopping: Predictive replenishment models will manage recurring household and business needs without manual input, driven by cryptographic AP2 Mandates.
  • Death by AI liability: We expect over 2,000 legal claims regarding AI “hallucinations” (like the Air Canada precedent) by the end of the year (Gartner). Governance is now a marketing issue.
  • The service “dip”: Forrester predicts that overall customer service quality will paradoxically dip in 2026. “Faithfulness”, the AI’s ability to ground responses in verified source documents, will become a primary CMO KPI. Target a score of 0.95+ to prevent brand damage (Ragas).

Your 2026 investment checklist

  • Understand the intent of the search

    Capture the intent of customer searches through natural language querying and conversation on-site.

  • Clean your data lake

    You cannot build a concierge on fragmented data. If your inventory APIs aren’t real-time, your agent will hallucinate.

  • Invest in visual asset transformation

    Shift budget from copywriting to visual creation. Your images are your new keywords. Optimise them for machine vision.

  • Implement “zero copy” architecture

    Enable AI to access live data without moving or storing it in a new location (Salesforce).

  • Agentic optimisation (AO)

    Ensure your pricing and specs are machine-readable so personal agents can “index” and buy from you.

  • Response grounding / faithfulness

    Ground responses in verified source documents. Target a score of 0.95+ as a CMO KPI to prevent brand damage.

Read the full report

The full report covers the evidence and sources behind every figure here, with sector spotlights on retail, travel, finance and B2B, and executive perspectives from Klarna and Stitch Fix.

Thinking about what this means for your site?

We’re happy to talk through where conversation, search and recommendations could help your customers find what they need.