Bazaar AI marketing for Indian businesses, minus the slop

AI changed marketing. Most teams changed the tools, not the work.

Bazaar means market. This is Marketing Desk: marketing and sales after AI. Source: Britannica, bazaar

Bazaar is a working desk for founders and marketing teams. Your team bought the AI subscriptions and still does the same tasks the same way. This page has the tools and playbooks to change that: what to automate first, the funnel and budget maths, a copy desk, a brief builder, SEO vs AEO vs GEO, and how marketing works differently in India.

I'm a marketer first. I've co-handled a ₹54 Cr budget at JAIN Trust, formerly USDC Global, and influenced a ₹40 Cr+ pipeline at Gnani.ai, and I use AI as leverage, not as a slogan.

Start with my record and the rules it taught me. Then use the tools with your own numbers. The reports at the back are my read of what changed and what to do about it.

Sapta as the Marketer, drawn as a sticker: black and white, in a suit and tie, arms folded The marketer in me
Want the exact stacks?Eight engines, the tools in each and where a human checks. Applied AI →

First, what I have actually run.

Fair question. Have I run any of this?

My marketing record: the numbers and the roles.

Before you take marketing advice from anyone, check what they have actually run. Here is mine, from Gnani.ai to JAIN Trust, formerly USDC Global, with the fine print most slides leave off. Hover a row to read it closely.

InstrumentCloseWhereFine print
Budget co-handled₹54 CrJAIN Trust, formerly USDC Global: JAIN Online, EdNxt.ai and about 20 collegesA year for marketing and sales, co-handled with the VP of Marketing.
Pipeline influenced₹40 Cr+Gnani.ai, enterprise AIInfluenced, not closed alone. BFSI, telecom and government buyers, with sales doing the closing.
Organic leads vs paid▲ ~120%JAIN OnlineOne organic push against what Meta and Google usually delivered. One push, not a permanent rate.
Funding round$4MGnani.ai: the Info Edge roundI built the fundraising collateral. Investors fund the business, not the deck.
Budget growth₹5–6 CrA budget I co-managed from about ₹50 lakh to ₹5–6 Cr a year, at Gnani.aiBudgets grow when the last rupee can show what it earned. That's the only argument a CFO hears.
Production cycle▼ 60%AI workflows I builtTime from brief to finished asset, not the quality bar. A person still signs off every piece.
Events run8+Indian and international, in about 18 monthsCounted as events shipped, not leads. An event is a sales tool with a venue attached.
Summit showcase221 m²India AI Impact Summit, Feb 2026A floor area, not an outcome. Square metres don't sell anything; what people do inside them does.
Campaign reach~20M viewsJAIN Online awareness, JAIN Trust (formerly USDC Global), on about ₹10 lakhViews, not enrolments. Reach is the top of the funnel; I count what it does further down.
What the AI workflows took out, in the processes I automated
Brief-to-asset time, cut60%
Final approval by a person100%
Apr 2026 to nowMarketing Manager, JAIN Trust, formerly USDC Global

JAIN Online and EdNxt.ai: a scaled online-degree business and a 0-to-1 brand.

Case notes
2022 to Apr 2026Gnani.ai

Marketing Strategist, then Marketing Manager with the Founder's Office. B2B voice and conversational AI.

Case notes
2016 to 2022Freelance writing, Apeiron Mobility, Criss Cross Labs

Copywriting and blogs, then junior mechanical engineer building battery packs and small scooter units, academic writing, and solution strategist at a marketing company.

Case notes
Since about 2013Freelance

80+ projects across marketing, brand design and strategy. Trained as a mechanical engineer.

Case notes

Numbers like these come with lessons. Here are the ones I would hand any team.

What that record taught me: seven rules I work by.

  1. Buyer psychology comes before the campaign. What do they fear, compare and need to justify to someone else? A campaign built without that is a guess with a media plan.
  2. Shipping isn't winning. Forty posts a month is proof of production, not marketing. Prove value in pipeline, revenue and cost per customer.
  3. Design is evidence. Buyers judge the company by the quality of what it puts in front of them, long before they read a word. Source: Consumer Reports and the Stanford Persuasive Technology Lab, How Do People Evaluate a Web Site's Credibility?
  4. Brand and reputation compound. Every honest case study adds to the balance. One cheap trick can withdraw years of it.
  5. AI should kill repetitive production, not judgment. With AI doing the production at Gnani.ai (video, podcast, email and visuals), production cycles got 60% shorter. The calls about where money goes stayed human.
  6. Small, sharp teams with strong systems now beat big teams with loose ones.
  7. Human-directed, AI-built. A person sets the angle and the standard, AI does the labour, a person signs off every asset.

Rule seven, drawn: the human-directed, AI-built loop

Crimson steps are people, cobalt steps are machines. The loop closes on commercial numbers, which become the next brief.

Human-directed, AI-builtPositioning and the brief stay humanPerson sets the anglewho it's for, the standard, banned wordsAI does first drafts and research summariesAI researches, draftsfrom real queries and callsProduction runs as a systemAI produces variantsresize, caption, dub, clipNothing ships without a person approving itPerson checks, signs offfacts and taste, every assetMeasured in commercial outcomes, not volumeMarket answerspipeline, CAC, trust
Swipe the diagram sideways to follow the loop.
The same loop runs every engine on the Applied AI page. Where I've run it, at Gnani.ai, AI-assisted production cut the production cycle by 60%.

The full case notes on my work page →   The exact AI stacks →

Now the tools. Start with the question every team asks first.

Enough of my numbers. Let's run yours.

Should you automate this task? A three-question test.

How often does it happen?
How much judgment does it need?
What does a mistake cost?
Answer 3 questions

Pick one option in each row and the verdict prints here.

The test as a decision tree

This is exactly how the verdict above is decided: one veto question first, then points.

YesNoLots of judgment ANDa costly mistake?Keep it humanAI preps the research;a person owns the callAdd up the points2 each: daily · little judgment · cheap mistake1 each: weekly · some taste · some embarrassment5 or 6 pointsAutomate it3 or 4 pointsDraft with AI0 to 2Not yet
Swipe the diagram sideways to see every branch.
Most marketing tasks land on "Draft with AI": a machine does the first 80%, a person makes the final call.

Passed the test? Then there is an order to do it in.

What to automate first: the order I would go in.

The order, drawn: eight jobs, one to eight

Start here: easy to check, a mistake is cheap1Reportingthe week's numbers2Resizing andreformatting3Captions andsubtitles4Researchand briefs5First draftsof repeat formats6Follow-upsand reminders7Regionallanguage versions8Schedulingand publishingNever on the list: these stay with a personPositioning · the final approval · pricing · anything you'd have to apologise for in public
Swipe the diagram sideways to follow the order.
The queue from this section as one path. It starts where the output is easy to check and ends with publishing, which runs once everything before it is approved.

Start where the output is easy to check and a mistake is cheap. Here's the order I'd go in on a typical marketing team. Tick off what you've already put on a system.

  1. ReportingPulling the week's numbers into one summary. Pure pattern, easy to verify against the source.
  2. Resizing and reformattingOne creative into every size and platform. Nobody went to design school for this.
  3. Captions and subtitlesFast to generate, fast to proofread. Most people watch video with the sound off in public.Source: Streaming Media, on the Verizon Media and Publicis Media caption study
  4. Research and briefsCompetitor scans and source summaries, grounded in documents you choose.
  5. First drafts of repeat formatsNewsletters, product posts, event invites. The format is fixed; a person sets the angle.
  6. Follow-ups and remindersEvent and lead reminders on email and WhatsApp, with opt-in, with a person writing the template.
  7. Regional language versionsDubbing and translation into Hindi, Kannada and more, with a native speaker checking before it goes out.
  8. Scheduling and publishingThe calendar runs itself once everything above is approved.

0 of 8 on a system.

Never on the list: positioning, the final approval, pricing, and anything you'd have to apologise for in public. Those stay with a person. Run a task through the test ↑

The same work, packaged by outcome: the eight engines.

The eight AI marketing engines: pick the outcome you want.

Eight engines, one person in charge

ContentBlogs, LinkedIn,newsletterVideo and reelsReels, Shorts,carouselsSocial media managerCalendar, posts,replies, reportsSEO and AEOAudit, rewrite,schema, rankSales collateralDecks, case studies,proposalsOutbound leadsLists, emails,sequencesWebinars and eventsPages, invites,reminders, replaysResearch and reportingMarket scans, briefs,dashboardsA person stays in charge of each one
Swipe the board sideways to see all eight.
The eight listings below on one board, with what each one puts out. Every line runs back to the same place: a person in charge.

Pick the outcome you want. Each listing opens the exact stack, the workflow and where a person stays in charge.

A content engineBlogs, LinkedIn, newsletter

Articles, posts and a newsletter every week in your voice, built to rank and to be quoted.

A video and reels engineReels, Shorts, carousels

Cut from one long video, one webinar or one idea, on a schedule.

A social media managerCalendar, posts, replies, reports

Posts made and scheduled, replies drafted, a weekly report, you approving.

SEO and AEO, fixedAudit, rewrite, schema, rank

Found on Google and quoted by ChatGPT, Gemini and Perplexity.

More from Bazaar

An engine is a few tools wired together. Here is which tool does which job.

AI tools for marketing, by job: the classifieds.

The desk, by job: who fills which seat

In my everyday stackPicked per jobThink and writeClaudeChatGPTGeminiResearchNotebookLMPerplexitySearch ConsoleVisual and videoGoogle VeoElevenLabsDaVinci ResolveCanvaAutomateClaude CodeClickUpDistribute and sellWhatsApp BusinessInstantlyVacancy, humans only: Head of Taste
Swipe the chart sideways to see every tool.
Every classified below on one chart: fourteen tools across five jobs, ten of them in my everyday stack. The one seat no tool fills is Head of Taste.

Which AI tool for which marketing job, written as classified ads. Tools marked daily are in my everyday stack; the rest I pick per job. Filter by the job you need done.

Vacancy: Head of TasteMust decide what is good and what to cut. Must say no. Cannot be automated, outsourced or prompted. Humans only. Apply within.
DailyPosition filled: strategist and writerStrategy, writing in your brand voice, research synthesis. Claude
DailyPosition filled: builderBuilds the automations: edits, sites, scripts, pipelines. This paper's own page included. Claude Code
DailyPosition filled: ideas deskIdeation, scripts, image prompts. ChatGPT
DailyPosition filled: long-document readerLong documents and Google Workspace work. Gemini
DailyPosition filled: grounded researcherReads only the sources you give it and answers from them. NotebookLM
Position filled: fact finderCited web research, links included. Perplexity
Position filled: search listenerThe real queries people type, and whether your pages are indexed. Search Console
DailyPosition filled: motion artistImage to motion, video generation. Google Veo
Position filled: voice departmentVoice-overs and dubbing. ElevenLabs
DailyPosition filled: editor and colouristPro editing and colour, scriptable. Cuts my own travel series. DaVinci Resolve
DailyPosition filled: in-house designerCarousels, creatives and decks on brand templates. Canva
DailyPosition filled: managing editorContent calendar, tasks and approvals. ClickUp
Position filled: follow-up deskReminders, follow-ups and broadcasts, with opt-in. WhatsApp Business
DailyPosition filled: outbound clerkCold email at scale, with warm-up. Instantly

The full toolbox, by job →

Automation saves hours. The next tools are about money.

Funnel calculator: from cost per lead to cost per customer.

Your agency reports cost per lead. Your CFO asks for cost per customer. Here's the bridge between them. Enter your own numbers.

The formula: CAC = CPL ÷ (qualify rate × buy rate). That's the whole trick. The rest is arithmetic.

Cost per lead
Cost per qualified lead
Customers a month
Customer acquisition cost

Where the money leaks: your funnel, drawn

The same numbers as the calculator, drawn as a funnel. Move the sliders and watch where people drop out.

Leads1,250 Qualified375 Customers37.5 875 drop out70% don't qualify 338 don't buy90% of qualified leads CAC = ₹400 ÷ (30% × 10%) = ₹13,333 per customer
Swipe the diagram sideways to see the drop-offs.
Lowering CPL helps less than people think: doubling the buy rate halves the cost per customer, the same as halving CPL.

Read the full story →

Once you know what a customer costs, decide where the next rupee goes.

Budget split: where should the money go?

The benchmark, drawn: 60% brand, 40% activation

The Binet and Field benchmark, for an average consumer brand60%Brand40%ActivationYour split below goes into four pots that always add up to 100%Brand buildingPerformanceContent and organicExperiments
Swipe the chart sideways.
The bar is the fixed benchmark from Les Binet and Peter Field's work for the IPA: a starting point, not a law. The four pots carry no numbers here because those are yours to set with the sliders. Source: IPA, The next chapter for The Long and the Short of It

Set a budget and move the sliders. The split always adds up to 100%, and the desk tells you what your mix says about your plan.

Brand (benchmark)60%
Activation (benchmark)40%

The best-known benchmark comes from Les Binet and Peter Field's work for the IPA: roughly 60% brand and 40% activation for an average consumer brand. It's a starting point, not a law. Categories and stages differ. Source: IPA, The next chapter for The Long and the Short of It

Behind every split is one argument. Here it is, both sides.

Brand vs performance: which one should get the money?

Point: performance pays the bills

Performance marketing is honest. You spend a rupee, you see what came back. Leads, sales, a cost per acquisition you can defend in a review.

When a business needs revenue this quarter, a beautiful brand film is a luxury. Founders and CFOs trust what they can count, and they're not wrong to.

And AI made performance faster: more ad concepts, more variants, quicker reads on what wins.

Counterpoint: brand lowers the bill

Performance harvests demand. It doesn't create much of it. Someone has to make people know you before they're ready to buy, or every click is a cold one.

When everyone can make a hundred ads a day with AI, the ads stop being the advantage. The thing a cheaper competitor can't copy is a name people already trust.

Brand is the reason the same ad converts better for one company than another.

Brand vs performance over time: why both lines matter

Performance bursts lift sales quickly and fade when the spend stops. Brand builds slowly and raises the floor every burst starts from.

Launch6 months1 year2 years3 years Effect on sales Performance: quick spikes that fade Brand: slow, and it compounds
Swipe the chart sideways.
An illustration of the pattern Les Binet and Peter Field describe in The Long and the Short of It (IPA, 2013), not plotted data. Source: IPA, The next chapter for The Long and the Short of It
The editor's ruling. Both, in a ratio you choose on purpose rather than one you drift into. Most teams I meet are all performance because it's easier to report. Measure brand by what it does to your performance numbers over time, and stop treating it as a separate religion. Set your own ratio in the allocator ↑

Read the full story →

Budget set. Now the words: test your copy before it ships.

Copy desk: test your draft for AI-speak.

What the red pen does: an example markup

EXAMPLEBrands must leverage cutting-edge AI to superchargeseamless growth. This game-changer will revolutioniseyour funnel — and take your brand to new heights!7flagged phrasesstruck out1em dashcountedSentence lengthwords per sentence
Swipe the example sideways.
A fixed example, using the sample draft that sits in the desk when the page opens. Paste your own copy below and the live count takes over.

Paste a draft and my sub-editor marks it up. It strikes the words that make copy sound machine-made, counts the dashes and checks how long your sentences run. It's a blunt tool, the way a red pen is. Source for the word list, in part: Wikipedia, Signs of AI writing

Spiked

Better copy starts with a better brief.

Prompt brief builder: write the brief, not the wish.

Six answers make one brief

1What you're making2Who it's for3What they should do after reading4Voice5Must include6AvoidThe briefwritten for a fast, literal juniorwho has never met your customerwho it is for · what good looks likewhat to avoid
Swipe the diagram sideways.
The six fields of the builder below, drawn as the parts of one brief.

A prompt is a brief to a fast, literal junior who has never met your customer. Tell it who it's for, what good looks like and what to avoid, and the output changes completely.

The wishWrite a LinkedIn post about our new course.
The briefWrite a LinkedIn post for working engineers in Bengaluru thinking about an MBA. Goal: get them to the webinar. Voice: plain, a little dry, no buzzwords. Include the date and one real outcome. Give me three openings and tell me which you'd pick.

Build your brief


      
More Applied AI notes

One good brief is a draft. A system of them is an engine.

Still curious? Here's my read of what changed.

AI content without slop: build an engine, not a tap.

Two production lines, drawn side by side

The slop machineA topic anda word count›No source butthe internet›Published becauseit exists›Measuredby volumeThe content engineAn angle aperson chose›Real queries,calls and data›A voice guide withbanned words›Checked beforeit ships›Measured byleads and trustSame model, different system.
Swipe the diagram sideways to follow both lines.
The two lists from this section, laid out as production lines. The engine has one more station than the tap: a check before anything ships.

AI writes average because average is what the brief asked for. A content engine starts from real inputs (the questions people search, what buyers say on sales calls, your own numbers), has a voice guide with banned words, and keeps a person on the angle before drafting and on the facts before shipping. Same model, different system.

AI content without slop: build an engine, not a tapThe full column: why the brief makes slop, not the model, and how the engine is wired.Read the story →

See the content engine stack →

The slop machine
  • A topic and a word count
  • No source but the internet
  • Published because it exists
  • Measured by volume
The content engine
  • An angle a person chose
  • Real queries, calls and data
  • A voice guide with banned words
  • Checked before it ships
  • Measured by leads and trust

An engine is only useful if people find what it makes.

SEO vs AEO vs GEO: what each one means and what to do.

Where you show up in each one

SEOSearch engine optimisationAEOAnswer engine optimisationGEOGenerative engine optimisationyouThe direct answeryouSource: youYou rank in the listand people click throughthe classic list of resultsYou are pickedas the direct answersnippets, voice, AI summariesYou are cited insidean AI-written answerChatGPT, Gemini, Perplexity
Swipe the diagram sideways to compare all three.
Three sketches of the three definitions below: a link in the list, the answer box, and a citation inside a written answer. In practice AEO and GEO ask for the same habits.

People still search. More of them now get the answer from ChatGPT, Gemini, Perplexity or an AI summary on Google, without clicking anything. Answer engine optimisation (AEO) is how you become the source that answer is built from. Source: Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results

SEOSearch engine optimisation

Getting a page to rank in the classic list of results, so people click through to you.

AEOAnswer engine optimisation

Getting your content picked as the direct answer: featured snippets, voice answers, AI summaries.

GEOGenerative engine optimisation

Getting cited inside answers written by generative AI such as ChatGPT, Gemini and Perplexity. The term comes from a November 2023 research paper, "GEO: Generative Engine Optimization", later published at KDD 2024. Source: Aggarwal et al., GEO: Generative Engine Optimization (arXiv)

In practice AEO and GEO ask for the same habits, so the table treats them together.

 
SEO
AEO and GEO
The goal
Rank a link in the results
Be the source an AI answer quotes
The unit
A page and a keyword
A question and a clear fact about you
A win looks like
A click to your site
A mention or citation, sometimes with no click at all
How you write
Cover the topic well
Answer first in plain words, then go deep
Structure
Titles, internal links, speed
Clear headings, FAQs, schema markup, the same facts about you everywhere
You measure
Rankings and clicks in Search Console
Whether AI answers name you, and branded search

What I do on every page now

  • Put the direct answer in the first two sentences under each heading.
  • Write headings as the questions people actually ask, taken from real search queries.
  • Add FAQ and article schema so machines can read what the page is.
  • Keep facts about the brand identical across the site, profiles and listings.
  • Cut the filler. An answer engine skips padding, and so do readers.
My takeMost of AEO is good SEO done honestly. The new part is writing so a machine can lift one clean paragraph and still be right.

The SEO and AEO stack →

Read the full story →

All of this runs in India, which has rules of its own.

Marketing in India: what works differently here.

Five things that work differently, on one board

22languages in theEighth ScheduleOne languageis a slice, not a marketRegional versions from one pipeline;a native speaker checks every lineThe conversationlives on WhatsAppreminders and answers,with clear opt-inThe phone isthe storefrontsmall screen and a patchyconnection firstProof beatspromisesreal people, real numbers;label what's paidThe calendaris regionalDiwali is national; Onam, Pongaland Durga Puja by state
Swipe the board sideways to see all five.
The five dispatches below as one board. The 22 squares are the scheduled languages; the single lit one is what a one-language campaign covers by count.

Most marketing playbooks online were written for a different country. A few things work differently here, and AI changes the cost of getting them right.

One language is a slice, not a market

The Constitution's Eighth Schedule lists 22 languages, and many buyers would rather read and hear their own. An English-only campaign reaches a slice of the country and calls it India. Sources: Britannica, Indian languages, Wikipedia, Languages with official recognition in India, The Week, on the KPMG and Google report Indian Languages: Defining India's Internet

Regional versions used to mean a new shoot or a new agency. Now dubbing and translation can come out of the same pipeline: I've run Hindi and Kannada versions from one production line. The rule that doesn't change is a native speaker checking every line before it ships, because a wrong word in someone's mother tongue is worse than no version at all.

The dubbing and video stack

Now the reports: what changed, what died and the conditions today.

Report: what AI moved in marketing, and what it didn't.

The six verdicts, sorted

Now runs as a systemStill human workMovedFirst draftsfrom a draft, not a blank pageProductionresize, caption, dub, clipShiftingDistributioncheaper to feed, with a systemTrustreal proof is worth moreDidn'tPositioningwho you are for, and whyTastewhat is good, what to cut
Swipe the chart sideways to see all three columns.
The six cards below sorted by verdict. Two moved, two are still shifting, and two did not move at all: positioning and taste.
First draftsCopy, scripts, decks and briefs now start from a draft, not a blank page.
Moved
ProductionResizing, captioning, dubbing and clipping can run as systems.
Moved
PositioningKnowing who you are for and why they should care is still human work.
Didn't
TasteChoosing what is good, and what to cut, still decides whether anything works.
Didn't
DistributionOrganic channels got cheaper to feed, if you have a system to feed them.
Shifting
TrustGeneric AI content made real proof and real people worth more, not less.
Shifting

Some tactics did not survive the move.

Report: marketing tactics that died, and what replaced them.

The four, by how dead they are

KeywordstuffingOld enough to rememberSurvived by peoplewho write for humansThe blankpagePassed quietlyDied when every briefstarted with a draftThe one-languagecampaignGravely illSurvived by vernacular cutsand a native speakerLast-clickcreditStill walking aroundNot dead.Should be.
Swipe the row sideways to see all four.
The four obituaries in a row, from long gone on the left to still walking around on the right. The dashed stone is the one that has not been filled yet.

The jobs moved on, and so did the tricks that came with them. Tactics that passed away, or should have. Mourned by some.

Keyword stuffingOld enough to remember Hover or tap for the obituary
Repeated itself until search engines stopped listening. Survived by people who write for humans and still rank. Source: Google Search Central, Spam policies for Google web search
The blank pagePassed quietly Hover or tap for the obituary
Died when every brief started with a draft. Nobody misses it, though a few writers say the coffee tasted better back then.
The one-language campaignGravely ill Hover or tap for the obituary
Struggled once regional versions became cheap to make. Survived by vernacular cuts and a native speaker with a red pen.
Last-click creditStill walking around Hover or tap for the obituary
Not dead. Should be. Gave the final ad all the credit for a journey it joined at the end. Source: Google Ads Help, About attribution models

What is left is today's weather.

Report: marketing conditions now, and what to do in each.

The forecast at a glance

REGION OF YOUR PLANCONDITIONWHAT TO DOAI content volumeFlood warningPublish less and prove moreSearch clicksOvercastWrite to be quotedTrustHigh pressure, clearPut people in frontMeasurementFogAgree a baseline before you startVernacularWarm frontOne more language, done properlyWhatsApp commerceSteady breezeAutomate the reminders, not the relationship
Swipe the board sideways to read the advice.
All six cards on one board, with the advice each card keeps behind a tap. My read of conditions, not a measurement.

Today's forecast, by region of your plan. This is my read of conditions, not a measurement. Tap a card for the advice.

Source for Search clicks: Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results. The other five cards are my read from running campaigns.

Before you fold the paper, the questions I get asked most.

Founder FAQ: what founders and marketing heads ask me.

The questions founders and marketing heads ask me first, paraphrased, with names left out.

A black-and-white portrait of Sapta, smiling and looking up to one side

"Sir, is AI content going to look like AI slop?"

A founder

Slop happens when nobody directs the work. I set the concept and the standard, AI does the production, and a person approves every asset. If it looks generic, it doesn't ship.

"Will this replace my team?"

A marketing head

It replaces repetitive, low-judgment work. Your people move to strategy, relationships and decisions. Whether you then hire fewer people is your call, not the software's.

"How will we know it worked?"

A founder

We agree the baseline first: hours spent, cost per asset, cost per lead or enrolment, pipeline. Then we track those, not likes. Try the funnel math with your own numbers.

"Which tools should we buy?"

An operations lead

Fewer than you think. Start from the outcome, then pick the few tools that get you there and wire them together. The tools by job list what I use.

Corrections and clarifications. Many marketing decks in recent years said AI would replace marketers. It replaced tasks. We regret the error, and so do the decks. Write to the editor →

The longer stories live in the columns.

Stories and columns: my marketing and AI notes.

Sapta as the Marketer, drawn as a sticker: black and white, in a suit and tie, arms folded
The columnist. Me, drawn as the Marketer: the same sticker that opens this page. The stories and columns below are my notes on marketing and AI.
From freelancer to the Founder's Office: what each marketing job taught meThirteen years, six desks and the lesson each one left behind.Read the story → Shipping isn't winning: why content volume is not marketingThe editorial, in full: buyer psychology, design as evidence and proving value in outcomes.Read the story →
Loading…

The other six worlds are my field research. People buy with their gut on Yantra, plan for weeks on Yatri, ask why on Kyun, argue with themselves on Manthan, judge by looks on Kala and by taste on Rasa.

Every Bazaar note →   Every Applied AI note →   Work with me →