The marketing workflow, mapped to where AI helps

The stages of a marketing month

A working marketing function moves through eight stages, sometimes in sequence, often overlapping. Strategy sets what we are trying to do. Brief writing translates strategy into instructions for production. Copy and creative production is the bulk of the visible work. Channel adaptation cuts the work for each platform. Campaign launch is the moment of shipping. A/B testing runs the experiments that tell us what actually worked. Analytics and reporting close the loop. Next-cycle planning decides what we keep, what we kill, and what we change for the following month.

Of these eight, AI moves time inside three or four and barely touches the rest. Walking the map honestly is the difference between using AI well and pretending it has rewritten your job.

Stage 1 — Strategy. AI leverage: low.

The first week of the cycle is the senior marketer and the founder or country head deciding what this campaign is for. Are we acquiring new wallet users, defending against a competitor, raising average transaction value, getting dormant users back? What does Dashain mean for the brand this year — gratitude, family, festive spending, financial discipline? Which audience segments matter — the salaried urban user, the diaspora son in Qatar sending tika money home, the small merchant accepting QR payments? What does the country look like in Ashwin 2083 that did not look the same last year?

These are judgements. They depend on data the model does not have — your acquisition cost trends, what your CEO will say yes to, what the regulator quietly signalled last quarter, what a competitor just launched. AI is not useless here. A senior marketer can use it to summarise last year’s campaign report, to draft alternative positioning statements to react to, or to compress a competitive scan. But the decision is human. The strategy meeting takes the same number of hours it took last year.

Stage 2 — Brief writing. AI leverage: medium.

Once the strategy is set, briefs go to the production team. A creative brief, a content brief, a media brief, a PR brief. These documents are formulaic enough — objective, audience, key message, tone, channel, deliverables, deadline, success metric — that AI handles the first draft well. A senior marketer who used to spend an hour writing a brief writes it in twenty minutes by giving the model the strategy summary and editing the output. The judgement is still hers; the typing compresses.

The risk at this stage is that a thin brief generated quickly will produce thin work for the rest of the month. The leverage is real, but the discipline of writing a specific brief — naming the audience precisely, defining what success looks like, marking the one thing the campaign must do — is what makes downstream AI usage productive. A vague brief plus AI gives you a lot of vague copy, fast.

Stage 3 — Copy and creative production. AI leverage: high.

This is where the month visibly compresses. Headlines, captions, ad copy, email sequences, landing-page copy, blog drafts, video scripts, image prompts, the long tail of product-detail pages. A team that used to spend two weeks producing the festival campaign’s full asset set produces a complete first-draft set in three days. Junior copywriters become editors and quality controllers; the senior writer spends less time generating and more time selecting and sharpening.

The work does not disappear. It changes shape. A copywriter no longer types the third version of a banner headline from a blank screen; she reads forty AI-generated options, kills thirty-five, sharpens five, and decides which two go into testing. The skill that matters is taste — which most marketing teams under-developed because the old workflow rewarded volume of production over volume of selection.

Stage 4 — Channel adaptation. AI leverage: high, with caveats.

The Dashain campaign exists across Facebook, Instagram, TikTok, Viber broadcast, YouTube pre-roll, email, the in-app banner, the merchant-side WhatsApp groups, the printed POS material at partner stores, and the diaspora-targeted version that runs in the Gulf and Australia. Re-cutting a single creative concept across eleven channels used to take a junior writer most of a week. AI does it in an afternoon.

The caveats matter. The model needs to be told the specific channel norms — TikTok openings, Viber broadcast length, the polite register Nepali email subscribers expect, what works for a diaspora audience that has been away for years. Without that briefing the cuts read generic. With it, the work compresses dramatically and the marketer’s time goes to the harder question: which cuts to actually run.

Stage 5 — Campaign launch. AI leverage: low.

Launch day is logistics, approvals, and nerve. The creative director signs off the final assets. The performance lead loads the ad sets. The community manager warms up the social accounts. The legal review clears the regulatory copy. Someone — usually the most senior person on Slack at 6 a.m. — pushes the button. None of this changes because of AI. The model can help draft the internal launch announcement and the press release, which are minor wins. The actual launch is a human act of coordination, judgement, and accountability.

Stage 6 — A/B testing. AI leverage: medium.

Variant production for tests is free, as the previous section noted. The model generates ten subject-line variants in the time it takes to read this paragraph. The interesting work — what to test, what hypothesis the test is checking, how to read a result that is statistically borderline — does not change. The marketer who knows that the meaningful test for the Tihar push is “does invoking family obligation beat invoking aspirational spending” still has to design that test. The model writes the variants once she has decided.

Stage 7 — Analytics and reporting. AI leverage: medium-to-high.

Once data is in, AI is excellent at producing the narrative around it. Paste the Meta Ads breakdown, the GA4 conversion path, the email performance summary into the model and ask for a one-page client report or an internal week-in-review. The arithmetic is yours; the prose is the model’s. The Monday standup deck that used to take a half-day takes forty-five minutes. The monthly client report that used to take two days takes half a day.

The risk is letting the narrative outrun the data. The model will produce a confident story from numbers that are noisier than the story suggests. The marketer’s job at this stage is to be the sceptic the model is not — flagging “this is one week of data, not a trend,” refusing to confuse correlation with causation, naming the things we still do not know.

Stage 8 — Next-cycle planning. AI leverage: low.

The end-of-month meeting decides what to keep, what to kill, what to change. This is strategy again, with one cycle of fresh data attached. The model can help summarise the month’s results and flag patterns. The decisions — kill this audience segment, double the budget on this creative direction, scrap the diaspora cut and rebuild it from scratch — are human. The strategy meeting that opens the next month is the same act of judgement that opened this one.

The shape of the new month

Look at the eight stages together and a pattern emerges. AI compresses time heavily inside production, channel adaptation, and reporting — and barely at all inside strategy, launch, and planning. A marketing month that used to be 30% strategy, 50% production, 20% reporting becomes 35% strategy, 30% production, 35% reporting. The total hours often do not fall. What changes is where they go. The team that was drowning in execution can finally think.

This is the optimistic reading. The pessimistic reading is that teams who simply ship more in the saved hours, without using them for sharper strategy and better selection, end up doing more bad marketing per month than they used to do mediocre marketing. The compression is real either way; whether it improves the work is a managerial choice, not a technical one.

Check your understanding

Quick check

—

Which of the following pairings of marketing stage to AI leverage is most accurate for a Nepali fintech running a Dashain campaign?

1. AStrategy: high leverage. Production: low leverage.
2. BStrategy: low leverage. Copy and creative production: high leverage.
3. CCampaign launch: high leverage. Analytics and reporting: low leverage.
4. DNext-cycle planning: high leverage. Brief writing: low leverage.

Check answer→

Quick check

—

A founder says: let us have AI write the entire Dashain campaign in one hour and ship it. What is the most defensible pushback?

1. AAgree — the model is fluent enough that an hour of generation plus a quick read is fine for festival work.
2. BRefuse outright — AI should not touch festival campaigns because cultural calibration is unsafe.
3. CAccept the compression but insist that strategy, brief, brand voice check, cultural review, and final selection stay human — AI shortens production, not judgement.
4. DCounter-propose using AI only for the diaspora audience and writing the domestic campaign manually.

Check answer→

What comes next

The map shows that one stage runs through every other stage — the brand voice that production must stay inside, the brief must encode, the cultural review must check against, the launch must protect. It is the single thing AI must not eat. The next section is about exactly that: what brand voice is, what models can and cannot hold, and how a Nepali marketing team keeps the voice human while letting AI accelerate everything around it.