Nauman Khalid

An agent comparing sellers needs more than product copy: it must know whether the correct variant is in stock, delivery will arrive on time, specifications match the request, return conditions are clear and the merchant can keep its promise.

• Much of that information sits outside the website, across catalogues, inventory systems, shipping rules, marketplace feeds, CRMs, support documents and, in smaller businesses, spreadsheets and WhatsApp threads.
• The recurring failure is disagreement between those systems.
• A human buyer can phone the company or lean on brand trust; an agent has no reason to fill in the gaps.
• Khalid's proposed fix is unglamorous: clean product attributes, dependable inventory, explicit policies, realistic delivery estimates and systems that agree.
• They point to Google's Universal Commerce Protocol, OpenAI's merchant feeds and Shopify's unified catalogue work as early signals, not settled standards.
• A companion post makes the same point from the shopper's side: discovery is shifting from searching to asking, and thin product data, stale reviews and slow pages can quietly remove a brand from the answer.

Khalid lists the inconsistencies an agent cannot resolve: the page says one thing while the feed carries another title; inventory is updated by hand; a delivery promise applies only to a few pincodes; and the return policy has exceptions no system understands.

The argument sharpens into a governance question.

• In a post on Shopify's agentic storefronts, Khalid notes that eligible stores are active by default and that the "Allow Shopify to manage for me" setting can go unowned.
• They lay out what to check: which channels can read the catalog, whether direct checkout is on, whether new channels are added automatically, and what product and order data are involved.
• Shopify says a direct-checkout channel receives an order's name, email, phone and delivery address, but not full order history or the customer database.
• Khalid adds that a catalog setting does not mean a product will rank, appear in an AI answer or produce a sale.

Read original · Original 2

One describes a client whose Meta leads were, in the client's own words, useless: the data showed auto-rickshaw drivers filling out forms about investing in Thailand property.

• Khalid diagnosed an optimisation problem, not a Meta problem, and the agency took on the account with a promise: no qualified leads, no invoice.
• The changes began before the submit button: high-friction forms to filter out the wrong people before they reached the CRM, creative built around investor psychology rather than mass reach, and audiences tuned for quality over volume.
• The second post explains the mechanism.
• If a business sends Meta only the event "form submitted," the platform learns to produce more form submissions; a serious prospect looks just like a casual click.
• Sending CRM outcomes back — contacted, qualified, converted, rejected — teaches the campaign what the business values.
• Khalid notes that since April, newly created campaigns cannot use Meta's qualified-leads goal without a Conversions API integration.
• They cite Meta's reported averages of a 15% reduction in cost per quality lead and a 44% improvement in the rate leads became qualified.
• The integration alone fixes nothing: sales needs a consistent definition of qualified, CRM stages need updating, and rejected leads need real reasons instead of being left in an open stage.

A fresh campaign was built in under 24 hours.

• The first event ran in Bangalore; for the second, the agency chose Ahmedabad, a smaller and less obvious market on paper.
• Results were strong enough that the client held another event there the following week.
• The reported outcome against the previous agency was 344% more reach at 91% lower CPM.

Read original · Original 2

A founder who manufactured a product for Rs 100 felt uncomfortable pricing it at Rs 400, even though that was the minimum needed to trade online.

• Khalid's reply: consumers pay for perceived value, brand trust and category benchmarks, not what a product cost to make; even Nike sells at roughly ten to twelve times manufacturing cost.
• On marketplaces, they offer a working rule — cost of goods times four equals survival — leaving only 7 to 8 percent net after ads, returns, fees, shipping and operations, not 30 or 50 percent.
• The conclusion: cost does not define price; brand, execution and the market do.

A portfolio post shows the same discipline at launch.

• For the brand Kardhini, the first conversation was not about a logo or Shopify theme, but what to launch.
• The founders had a limited budget and wanted a luxury position, so the discussion covered whether a price premium could cover product, packaging, delivery, operations and the cost of acquiring the first customer.
• The outcome: a chosen product direction, one logo mark with two lockups, brand guidelines, a wooden-box packaging direction, a pricing ladder and a live storefront.

Read original · Original 2

The causes they identify are mundane: hero images that take 16 seconds to load, JavaScript bundles that block rendering before a single product appears, and Total Blocking Time far above the benchmark.

• They stress that the comparison is about fundamentals, not a verdict on either brand's product or reputation.
• The most expensive platform and theme mean nothing if the basics are unhandled.
• On mobile, on slow 4G in India, those basics decide whether a visitor bounces or buys.

Khalid compares a well-known Indian D2C fragrance brand on Shopify Plus, with a mobile PageSpeed score of 34, against a brand they say they built on WordPress and shared hosting, which scored 82. The smaller store recorded a Largest Contentful Paint of 3.4 seconds and 0ms blocking time, against 970ms blocking time on the enterprise store.

Read original

Khalid points to what advertisers have reported: garments changing structure, products acquiring features they do not have, brand messages being altered, and teams finding AI enhancements inside live campaigns they did not expect.

• Their diagnosis is that the model produced something plausible because it did not know what the company considered true.
• Most brands, they say, have no usable source of product truth: descriptions sit on Shopify, approved claims in a PDF, packaging language with the design team, and unwritten rules in the heads of the founder and media buyer.
• The advertising system sees fragments, so it completes the picture.
• Khalid proposes a structured record they call a brand truth layer: immutable product attributes, approved and prohibited claims, visual details that must survive generation, current pricing and availability, regional language rules, and conditions that require human review.
• Large businesses will build sophisticated versions; a smaller brand can start with a clean catalogue, a claim library and a basic approval process.

Khalid illustrates the failure with a generated product image that looks convincing but shows a garment that has changed shape or a two-handled bicycle — plausible output that contradicts what the brand sells.

Read original

Most agencies, they note, would have quoted by then.

• Their reasoning is that running ads is the easy part — any competent media buyer can spend a budget — while the factors that decide whether a business makes money surround the ad.
• One example: Meta's algorithm can serve a brand's existing customers at the same acquisition cost as a first-time buyer.
• CAC looks identical on the dashboard, even though one conversion is profitable and the other is a retention failure the brand has paid to repeat.
• Profit in D2C, Khalid argues, comes from the second, third and fourth order, and from WhatsApp and email sequences that bring warm audiences back without putting them into cold acquisition campaigns.
• A brand whose ROAS looks fine but whose margins do not is the stated fit.

Khalid draws on their own ventures — Forever Muslim, started in 2018, and Sirat Perfumes after that — to make the case. Years of learning with their own money taught them that ROAS on a dashboard and profitability in the business are two different numbers that do not always move together.

Read original

Researchers examined the rollout of Claude Code and GitHub Copilot CLI across tens of thousands of Microsoft engineers.

• Adoption was driven less by training than by seeing colleagues use the tools: where more than a quarter of an engineer's wider peer group had already used Copilot CLI, their own likelihood of trying it rose 216%.
• Continued use followed a different pattern: engineers stayed when the tools fitted the coding they were already doing.
• The researchers estimated that adopters merged around 24% more pull requests during the study period, while cautioning that a merged pull request measures output, not necessarily value.
• Khalid's argument: most corporate AI programmes begin with procurement, training sessions and an adoption dashboard, then return employees to a workflow designed before the tool existed.
• A better rollout starts with a few credible practitioners, a small number of recurring jobs, public use and honest reporting of what was completed, what had to be corrected, how much review it created and where it failed.
• They extend the reasoning to sales, operations and marketing teams.

Khalid's examples beyond software are concrete: a sales team adopts AI research after seeing a colleague walk into a better-prepared meeting; an operations team adopts automation when a weekly report arrives without three hours of spreadsheet copying; and a marketing team adopts an agent when it removes an actual production bottleneck.

Read original

The self-serve Ads Manager is live, the US pilot crossed $100 million in annualised revenue within six weeks, and more than 600 advertisers had joined by late March.

• Yet advertisers testing it still ask for targeting, experimentation and reporting that Meta and Google made ordinary years ago.
• Khalid reads that gap as a possible opening, not just a weakness.
• In a chat, users state their problem in a sentence — payroll software for a 40-person company, a CRM for a small real-estate team, an agency to rebuild a website — instead of being interrupted between photos or reels.
• The test they propose is narrow: can a brand appear during a high-intent conversation, make a credible promise and carry that intent into a landing page written for the question asked?
• Khalid warns against treating it as a mature performance channel and judging it only by whether it beats Meta's cost per lead next week.

Khalid says they have spent their own money across enough performance campaigns to be cautious of every new platform's launch narrative: early reach is rarely the same thing as repeatable acquisition.

Read original

Start here to understand why the work is operational, not promotional.

Khalid argues that an AI agent recommending a product reads more than the product page.

• It needs to know whether the correct variant is in stock, whether it can arrive on time, whether the specifications match the request, what the return conditions are and whether the merchant can keep the promise.
• Much of that information sits outside the website, across the catalogue, inventory system, shipping rules, marketplace feeds, CRM, support documents and, in smaller businesses, spreadsheets and WhatsApp conversations.
• The failure mode they describe is inconsistency: the page says one thing, the feed gives another title, inventory is updated by hand, a delivery promise covers only a few pincodes, and the return policy has exceptions no system understands.
• A human buyer can navigate around those gaps; an agent comparing twenty sellers has no reason to.
• They point to Google's Universal Commerce Protocol, OpenAI's merchant feeds and Shopify's unified catalogue work as early signals, and say the advantage may come from making the business unusually easy for someone else's assistant to understand.

For brands planning a campaign

Use this as the framing document for a data-cleanup audit. The implied scope: clean product attributes, dependable inventory, explicit policies, realistic delivery estimates and systems that agree with one another.

Khalid points out that Shopify says agentic storefronts are active by default for eligible stores.

• With the "Allow Shopify to manage for me" setting on, available AI channels can reach product data through Shopify Catalog, direct checkout is active where relevant, and new agentic channels can be added automatically.
• They recommend that anyone running a D2C store open Sales channels, then Agentic, and check four things: which channels can access the catalog, whether direct checkout is available and switched on, whether new channels should be added automatically, and what product and order information is involved.
• Shopify says a direct-checkout AI channel can receive an order's name, email, phone number and delivery address, but not the store's full order history or general customer database.
• The same area carries channel-level performance reporting, so growth, ecommerce and whoever owns customer data should make the choice together.
• Eligibility and direct-checkout availability vary by channel and market, and a catalog setting does not mean a product will rank, appear in an AI answer or generate a sale.

For brands planning a campaign

Start the operational audit with this settings screen, and assign a named owner to the default. The guardrail: do not promise AI-channel visibility as the outcome of a settings change.

Khalid describes what advertisers have reported when platforms generate creative automatically: garments changing structure, products acquiring features they do not have, brand messages being altered, and teams discovering AI enhancements inside live campaigns they did not expect.

• Their diagnosis: the model made something plausible because it did not know what the company considered true.
• Most brands have no usable source of product truth, they point out, because the description sits on Shopify, approved claims sit in a PDF, packaging language sits with the design team, and the founder and media buyer each hold rules that were never written down.
• The advertising system sees fragments.
• Their proposed fix is a structured record they call a brand truth layer: immutable product attributes, approved and prohibited claims, visual details that must stay intact, current pricing and availability, regional language rules, and conditions that require human review.
• Smaller brands can start with a clean catalogue, a claim library and a basic approval process.

For brands planning a campaign

This is the second half of the cleanup. Pair catalogue work with a claim library and a simple approval process, giving generated creative something to check against.

Khalid describes declining to state an offer on an early call, because they first want to know whether the client's back end can handle scale, whether the supply chain is set up for the orders, and whether operations are ready.

• Most agencies would have quoted a retainer by then, they note.
• Their argument is that running ads is the easy part; what determines whether a business makes money is everything surrounding the ad.
• They give a specific example: Meta's algorithm will serve a brand's existing customers back to it at the same acquisition cost as a first-time buyer, and the CAC looks identical on the dashboard even though one conversion is profitable and the other is a retention failure the brand paid to repeat.
• Profit, they argue, comes from the second, third and fourth order, through WhatsApp and email sequences that bring warm audiences back without putting them into cold acquisition campaigns.
• They name the ventures Forever Muslim from 2018 and Sirat Perfumes after that as the source of this conclusion.

For brands planning a campaign

Use this as the diagnostic frame before discussing spend: readiness questions about fulfilment and retention come first, media buying second.

Khalid's point is that if the only event a business returns to Meta is "form submitted," the platform learns how to generate more form submissions, and a serious prospect looks identical to a casual click.

• They describe working on a Zoho CRM and Meta integration where the useful half was not moving the lead into the CRM but sending the outcome back: contacted, qualified, converted or rejected.
• Since April, newly created campaigns cannot use Meta's qualified-leads performance goal without a Conversions API integration, they note.
• Meta reports that advertisers using CRM feedback saw roughly a 15% reduction in cost per quality lead and a 44% improvement in the rate at which leads became qualified.
• The integration alone will not fix poor qualification: sales still needs a consistent definition of "qualified," CRM stages have to be updated properly, and rejected leads need meaningful reasons rather than sitting in an open stage.

For brands planning a campaign

Scope the CRM-to-platform feedback loop as a deliverable, and agree on qualification definitions before building the integration.

The client's complaint, in the post's own words, was that Meta leads were arriving completely useless: the data showed auto-rickshaw drivers filling out forms about investing in Thailand property.

• Khalid's response was that this was an optimisation problem, not a Meta problem, and the agency offered to take the account on a promise of no qualified leads, no invoice.
• They built a fresh campaign in under 24 hours.
• The first event was in Bangalore, and for the second they chose Ahmedabad, a smaller and less obvious market on paper; the results were strong enough that the client held a second event there the following week.
• What changed, Khalid writes, was everything before someone clicked submit: high-friction forms that filter out the wrong person before they reach the CRM, creatives built around investor psychology rather than mass reach, and audiences optimised for quality rather than volume.
• The reported outcome was 344% more reach at 91% lower CPM than the previous agency.

For brands planning a campaign

A performance-fee structure tied to qualified leads, paired with form friction and audience-quality targeting, is the model this experience supports.

Khalid compares two stores: Adil Qadri, a well-known D2C fragrance brand on Shopify Plus, with a mobile PageSpeed score of 34, and Sui Kaari, which they say they built on WordPress and shared hosting, with a score of 82.

• They acknowledge Adil Qadri’s product, brand recall and scale, then identify a common mistake: founders treat website performance as an infrastructure problem and upgrade platforms, moving to Shopify Plus or a premium theme, when the platform is not usually the problem.
• They point to unoptimised hero images that take 16 seconds to load, JavaScript bundles that block rendering before a product appears, and Total Blocking Time of 970ms against a benchmark of under 200.
• Sui Kaari’s figures are LCP of 3.4 seconds and Total Blocking Time of 0ms.
• On mobile, on slow 4G in India, they argue, those differences separate a bounce from a purchase.

For brands planning a campaign

Include page-speed fundamentals in the audit. They affect the same agent-facing surface as catalogue data and do not require a platform migration.

An agent comparing sellers needs to know whether the correct variant is in stock, whether delivery will land on time, whether specifications match the request, what the return conditions are and whether the merchant can keep the promise.

• Much of that lives outside the website: catalogue, inventory system, shipping rules, marketplace feeds, CRM, support documents and, in smaller businesses, spreadsheets and WhatsApp threads.
• The recurring failure is disagreement between systems: the page says one thing, the feed another, inventory is updated by hand, or a delivery promise covers only a few pincodes.
• A human buyer can ring the company or trust the brand; an agent has no reason to fill those gaps.
• The suggested fix is unglamorous: clean product attributes, dependable inventory, explicit policies, realistic delivery estimates and systems that agree with each other.
• A companion post makes the same point about discovery shifting from searching to asking, where thin product data, stale reviews and slow pages remove a brand from the answer.

Frames the change as overdue operational plumbing rather than futuristic AI, and names the specific records an agent consumes.

For brands planning a campaign

Useful for a brief that starts before any media spend: audit catalogue, feed and inventory alignment, and write delivery and returns promises as explicit rules. Treat the work as operations, not an AI feature to bolt onto the storefront.

Examples: Source 1 · Source 2

Nauman Khalid sets out what to check in the Sales channels area: which channels can read the catalogue, whether direct checkout is available and switched on, whether new channels are added automatically, and what product and order data are involved.

• The post records Shopify's position that a direct-checkout channel receives an order's name, email, phone number and delivery address but not the full order history or customer database, and that channel-level performance reporting sits in the same place.
• Ask how the channel will perform: catalogue access alone does not show that a product will rank, appear in an AI answer or produce a sale.

Confirm eligibility and direct-checkout availability for each channel and market; settings vary.

For brands planning a campaign

A brand can ask for a governance pass: decide who owns the AI-channel switch, agree with the person responsible for customer data on what gets shared, then measure by channel rather than assuming catalogue access equals sales.

Examples: Source 1

The first covers a client whose Meta leads included auto-rickshaw drivers filling property-investment forms for a Thailand development.

• The diagnosed cause was not the platform but what happened before the submit button: high-friction forms that screen the wrong person out, creative built around investor psychology, and audiences tuned for quality rather than volume.
• The account was taken on under a guarantee of no qualified leads, no invoice, and the rebuilt campaign launched in under a day, reporting 344% more reach at 91% lower CPM than the previous agency.
• The second post explains the mechanism: if the only event sent back to Meta is form submitted, the platform learns to produce more form submissions, and a serious prospect looks identical to a casual click.
• Returning CRM outcomes—contacted, qualified, converted, rejected—teaches the campaign what the business values.
• Meta now requires a Conversions API integration for the qualified-leads goal on newly created campaigns, and Nauman Khalid cites Meta's average figures of a 15% reduction in cost per quality lead and a 44% improvement in the rate leads became quality leads.
• The integration alone does not fix qualification: sales needs a consistent definition, CRM stages need updating and rejected leads need real reasons rather than being left in an open stage.

Uses a performance guarantee to reframe responsibility: if an agency blames the platform, it is covering for itself.

For brands planning a campaign

Strong fit for a lead-generation brief where volume looks acceptable and margin does not. The deliverable is a closed loop: qualification criteria agreed with sales, CRM stages wired to the ad platform, and creative and forms designed to filter rather than maximise submissions.

Examples: Source 1 · Source 2

Nauman Khalid's reply is that consumers pay for perceived value, brand trust and category benchmarks, not input cost, and that even Nike sells at roughly 10 to 12 times manufacturing cost.

• On marketplaces, the working rule given is cost of goods times four equals survival, which leaves only 7 to 8 percent net after ads, returns, fees, shipping and operations—not 30 or 50 percent.
• A portfolio post shows the same discipline at launch.
• For Kardhini, the first conversation was what to launch rather than a logo or theme: the founders had a limited budget and wanted a luxury position, so the discussion covered whether a price premium could cover product, packaging, delivery, operations and the cost of acquiring a customer.
• The result was a chosen direction, one logo mark with two lockups, brand guidelines, wooden-box packaging, a pricing ladder and a live storefront.

Anchors pricing advice in an explicit formula and margin figure, then applies it to an actual launch decision.

For brands planning a campaign

Useful when a brand is setting launch price or deciding whether a premium position is viable. Expect the conversation to include unit economics, packaging, service cost and acquisition cost before creative work.

Examples: Source 1 · Source 2

The enterprise store scored 34 on mobile PageSpeed; the small one scored 82.

• The pattern named is founders treating performance as infrastructure and upgrading the platform, hosting or theme, when the causes are unoptimised hero images loading in 16 seconds, JavaScript bundles blocking render and Total Blocking Time of 970ms against a benchmark under 200ms.
• The smaller store's figures—3.4 second Largest Contentful Paint and 0ms blocking time—are not exceptional globally but matter on mobile, on slow 4G, in India.
• The closing point is that the most expensive platform and theme mean nothing if the fundamentals are not handled.

The post states the comparison is about fundamentals rather than a verdict on either brand's product or reputation.

For brands planning a campaign

A brand planning a replatform or a premium theme can ask for a performance audit first, checking image weight, render-blocking scripts and blocking-time figures before committing any migration budget.

Examples: Source 1

Nauman Khalid points to reports from advertisers of the operational problem underneath: garments changing structure, products acquiring features they do not have, brand messages altered, and teams finding AI enhancements inside live campaigns they did not expect.

• The model produced something plausible because it did not know what the company considered true.
• The proposed layer is a structured record of immutable product attributes, approved and prohibited claims, visual details that must survive generation, current pricing and availability, regional language rules and conditions that force human review.
• Large businesses will build sophisticated versions; a smaller brand can start with a clean catalogue, a claim library and a basic approval process.

Turns a platform capability story into an infrastructure requirement, naming the fields a brand would need to maintain.

For brands planning a campaign

Relevant to a brand planning to hand creative production to a platform. Ask what the brand truth record would contain before scaling generated variants, and who signs off on claims and product details.

Examples: Source 1

Where more than a quarter of an engineer's wider peer group had already used Copilot CLI, the likelihood of trying it rose 216%.

• Continued use followed a different pattern: engineers stayed when the tools fitted the coding they already did.
• Adopters were estimated to merge around 24% more pull requests during the study; merged pull requests measure output, not value.
• The argument is that most corporate AI programmes start with procurement, training and an adoption dashboard, then return employees to a workflow designed before the tool existed.
• A better rollout starts with a few credible practitioners, a small number of recurring jobs, public use, and honest reporting of what was completed, corrected, reviewed and failed.
• The post extends the reasoning to sales, operations and marketing teams.

Notes that the study measures output, not value.

For brands planning a campaign

Useful when a brand wants internal adoption of new AI tooling rather than a licence purchase. The practical need is a small pilot with named practitioners, visible results and a published account of where the tool struggled.

Examples: Source 1

The self-serve Ads Manager is live, the US pilot crossed $100 million in annualised revenue within six weeks, and more than 600 advertisers had joined by late March.

• Yet advertisers testing it still ask for targeting, experimentation and reporting that Meta and Google made ordinary years ago.
• Nauman Khalid reads that gap as a possible opening, not just a weakness: inside a chat, the user states the problem in a sentence—payroll software for a 40-person company, a CRM for a small real-estate team, an agency to rebuild a website—rather than being interrupted between photos or reels.
• The proposed test is narrow: can a brand appear during a high-intent conversation, make a credible promise, and carry that intent into a landing page written for the question asked?

Nauman Khalid cautions against new-platform launch narratives: early reach is rarely the same as repeatable acquisition.

For brands planning a campaign

A brand testing a new intent-led channel can be asked to run a short, bounded experiment with a purpose-built landing page and a defined success measure, rather than be judged against an established channel's cost per lead.

Examples: Source 1

The reasoning is that any competent media buyer can spend budget and generate traffic, while the factors that decide whether the business makes money surround the ad.

• Meta will serve existing customers back at the same acquisition cost as a new buyer, so the same conversion can be a profitable first order or a retention failure paid for twice.
• Profit in D2C is described as coming from the second, third and fourth orders, and from WhatsApp and email sequences that return warm audiences while keeping them out of cold acquisition campaigns.
• Khalid cites running Forever Muslim from 2018 and building Sirat Perfumes, and learning with the founder's own money that ROAS on a dashboard and profitability in the business are different numbers.

Makes a performance guarantee and a refusal to quote before diagnosis part of the sales approach.

For brands planning a campaign

Sets expectations for how an engagement is scoped: retention, CRM and operations sit inside the remit, and the first deliverable may be a readiness review rather than a media plan. The stated fit is a brand whose ROAS looks fine but whose margins do not.

Examples: Source 1

Nauman Khalid names Mecnex, scaled to Rs 8 crore; HoneyVeda from Shark Tank India; and digital systems built for the United Nations, alongside a description of a small team where interns run campaigns, write copy and track results from the first week. The Kardhini post follows the same shape: a portfolio piece that walks through the launch decision, the direction chosen, the assets produced and the live storefront.

For brands planning a campaign

Ask for case studies in this format—the decision made, the assets produced and the outcome—to judge whether the studio's range matches the campaign's requirements.

Examples: Source 1 · Source 2

The post argues that official productivity figures understate physical output because they measure value added, brand, IP, pricing power and tariffs. Its conclusion is that China's advantage is a systems advantage—process discipline, supply-chain density, faster iteration, infrastructure and engineering throughput—not low wages.

Included as range evidence: this is macro commentary, not advice a brand can apply to a marketing brief.

For brands planning a campaign

Signals a wider business and operations perspective that can be useful in sourcing or supply-chain conversations.

Examples: Source 1

View example

Khalid describes funding it themselves while learning how retention, WhatsApp and email sequences affect profitability.

View example

The post says Khalid was helping three brands with launch decisions at the time.

View example

A mobile PageSpeed comparison against Adil Qadri (82 versus 34) argues that platform upgrades do not fix unoptimised images and blocking scripts.

View example

The post does not specify their role or stake in the studio; ask them to clarify their relationship. It also names Mecnex (scaled to ₹8 Cr), Shark Tank India's HoneyVeda, and the United Nations as work the team has done, without describing each engagement.

View example

Khalid's posts walk through OpenAI and Stripe's Agentic Commerce Protocol and Google's Universal Commerce Protocol, covering product feeds, checkout endpoints and the fact that the retailer stays merchant of record.

Shivers works from the protocol side, explaining how a retailer plugs into a published standard and the implementation steps involved. Khalid starts from their client work, tracing failures to mismatches among catalogue, inventory, CRM and delivery promises, with less attention to the protocols themselves.

Read original · Original 2

Posts on Walmart's title changes and OpenAI's agent ambitions share Khalid's subject matter.

Sinclair writes fast platform news, with breaking-format hooks, growth figures and calls for comments to get further breakdowns. Khalid writes slower, first-person analysis that ends in an operational recommendation; they are more cautious about early platform claims.

Read original · Original 2

Argues that product data now functions as ad creative, so thin descriptions cause the agent to sell against the brand.

Stupar writes tactical, urgency-driven playbooks for UK sellers, including TikTok Shop launches and creator exclusivity windows. Khalid is an India-based D2C operator who emphasizes backend readiness, margin and cautious testing of new channels over first-mover speed.

Read original · Original 2

Tom's posts cover Google's Universal Commerce Protocol and OpenAI + Stripe's Agentic Commerce Protocol: how AI agents consume product feeds, compare items, build carts and complete checkout through delegated payments while the retailer stays merchant of record.

• He lists concrete retailer steps: provide a structured product feed to OpenAI, implement ACP checkout endpoints, support delegated payments, keep pricing, inventory and fulfilment logic clean and consistent, and track whether AI systems cite the brand.
• That is useful when the campaign needs to decide which protocol to support, how ChatGPT or Gemini shopping surfaces fit the roadmap, and what a machine-readable feed and checkout integration must contain.

The tradeoff is practical depth in store operations.

• Nauman's posts stay at the standards, feed and integration level; they do not walk through what to do when the Shopify catalogue, marketplace feed, warehouse count and delivery promise disagree, or how to assign ownership of an AI-channel setting.
• Their posts are stronger on operational reconciliation and storefront speed as the first diagnostic.
• Tom is also writing about US-led platforms and protocols; the comparison here does not show campaign delivery for an India-first D2C operation.
• If the brand needs a working data-cleanup programme rather than a protocol briefing, Nauman is the closer match.

Examples: Source 1 · Source 2

Max's posts give a specific Walmart title rule change - rejection over 90 characters, a strict Brand + Product Line + Key Features + Size + Count structure, and the argument that clean titles are raw material for Walmart's Sparky agent - plus a fast read on OpenClaw's growth and Peter Steinberger joining OpenAI to build personal agents.

• He also connects agentic commerce to consumer-brand recommendation logic: visibility shifts from shelf placement to algorithmic preference, and commerce infrastructure has to integrate with protocols such as OpenAI's Agentic Commerce Protocol.
• That is a better choice when the immediate task is marketplace listing compliance, keyword restructuring for an AI shopping agent, or explaining how an agent-led buying shift changes product content on a specific retail platform.

The tradeoff is breadth of operational readiness.

• The posts are marketplace- and news-led; they do not cover the full chain Nauman examines across catalogue truth, inventory, returns policy, CRM feedback and delivery promises.
• Max's format also includes comment-gated calls to action ('Like and comment ... for our full breakdown'), so the public text is a shorter briefing than a complete methodology.
• If the brand's problem is one marketplace's title format, Max's material is closer; if the problem is systems that disagree across the whole store, Nauman's is.

Examples: Source 1 · Source 2

Marko's posts cover ChatGPT Ads arriving inside Shopify and HubSpot: US merchants installing the app, international markets opening, HubSpot auto-tagging URLs, Sponsored Agents in test, and the practical warning that product data becomes the creative because sponsored agents answer from the feed and product detail.

• He also analyses Sephora launching new products on TikTok Shop before its own site, with creator retainers, AOV ceiling and exclusivity windows as the seller implications.
• That is a better choice when the campaign needs to run a bounded test on a conversational ad surface, set up attribution inside Shopify or HubSpot, or respond to a retail partner moving a category onto TikTok Shop.

The tradeoff is that Marko's emphasis is on channel entry, timing and commercial tactics, not on repairing the underlying catalogue, inventory and fulfilment data.

• He explains why messy specs or thin descriptions will make an agent sell against the brand, but the posts do not provide the audit method, owner checklist or claim-library structure that Nauman's posts set out.
• His TikTok Shop post is a US pilot analysis and the ChatGPT Ads post is time-sensitive launch coverage, so the durable operational content is thinner.
• If the brand's priority is a fast channel test, Marko's posts speak to it; if the priority is making the backend systems consistent before any channel test, Nauman is the fit.

Examples: Source 1 · Source 2