What marketing tasks should a small business hand to AI first?

Creative resizing and variation, first drafts of routine documents, transcription of calls and meetings, and reading campaign reports for changes. These are high-volume, low-judgement tasks where the tool is reliable and the time saved is real. Start there, not with strategy or customer-facing copy.

Can AI write ads in Hindi or regional languages?

It can produce them; it cannot make them sound local. For anything a real audience in a district town will hear, treat the AI draft as a starting point that a native speaker will substantially rewrite. Voice, idiom and rhythm are the whole point in regional creative, and they are exactly what a translation loses.

Is it safe to put client data into AI tools?

Only if you know which tool, what its terms say about training on your inputs, and whether the client has agreed. Campaign numbers, customer lists and unreleased creative should not go into consumer-grade tools by default. Ask your agency the same question.

Should an agency charge less because it uses AI?

Where the tool has genuinely reduced hours on mechanical work, something in the pricing should reflect it. Where the agency's value was never the mechanical work — the idea, the market knowledge, the accountability — the price is for that and should not move. A good agency can tell you which is which.

Will AI replace marketing agencies?

It is replacing the mechanical parts of what agencies did — resizing, drafting, media mechanics — and the platforms themselves are absorbing targeting and bidding. It is not replacing the judgement, the idea, the local knowledge or the accountability. Agencies built on the first set of things are in trouble; agencies built on the second are getting faster.

Automation

What AI does in a marketing team, honestly

Two years of demos have produced two camps: people who think AI has replaced the agency, and people who think it is a toy. It is neither, and both camps are wasting money.

4 September 20267 min read

The honest ledger

AI is now in every ad platform, every design tool and every inbox. The question is no longer whether to use it. It is which tasks it genuinely does well enough to hand over, and which ones look handed over until a client notices.

Here is the ledger as it actually stands in day-to-day agency and in-house work, with the failures included, because the failures are where the money goes.

Where it saves real time

Creative variants. Ad platforms now treat creative as the targeting, which means a campaign needs many genuinely different executions rather than one. AI is very good at the volume part: resizing, reformatting, generating a dozen openings from one idea, producing the twenty text variations an ad set wants. The idea still has to be good. The multiplication used to take a week and now takes an afternoon — and that is most of what stops creative fatigue from being a monthly emergency.

First drafts of anything routine. Briefs, run-of-show documents, post-event summaries, the first version of a proposal. Not the final version — the first one, the one that used to be a blank page for two hours.

Transcription and summary. Client calls, dealer feedback sessions, recordings from an activation. Turning an hour of audio into a page of notes is now instant, and it was never a good use of a person.

Reading reports. Pointing a model at three months of campaign data and asking what changed is faster than building the pivot table, and often catches the thing nobody was looking for — the placement that quietly took the budget, the creative that died in week three.

Research and reconnaissance. What a competitor is running, what a category's ads look like in another state, what questions people are asking about a product. The kind of reading that used to be a junior's whole afternoon.

Where it fails, quietly

Language that has to sound local. A model will produce Hindi, Marathi, Bundeli or Malvi. It will sound like a translation, because it is one. For anything a real person in a district town will hear or read, the voice has to come from somebody who speaks that way, and the AI's draft is at best a starting point they will mostly rewrite. This is the same failure as subtitling a metro ad for a village, automated.

Facts about the client. A model does not know what the dealer margin is, which products are in stock in Ratlam, or that the launch moved to November. It will fill the gap with something plausible. Every number, name and date in AI-drafted client material has to be checked by a person who knows, every time.

Judgement about people. Which dealer needs a phone call rather than a message, whether a complaint is serious, whether a client is unhappy. Anything that requires reading a person is not delegable, and the tools that claim to do it are guessing.

Original ideas. It is excellent at more-of-the-same and poor at the thing nobody has done. Ask it for a campaign idea and you will get the average of every campaign it has read. The average is not what a client is paying for.

Anything it produced that nobody read. The most expensive AI failure in agencies is not a wrong answer. It is a plausible one that went out because the person who should have checked it was told it was already done.

The one rule

A person's name goes on it, and that person read it.

Everything else follows. AI can draft, resize, summarise, translate a first pass, and read the numbers. A human who is accountable for the client decides what goes out, and has actually looked. The moment output travels from the tool to the client without passing through somebody who would be embarrassed if it were wrong, the tool is running the account, and it will eventually run it into a wall.

What clients should ask their agency

Not "do you use AI" — everybody does. The useful questions:

The answers tell you whether the agency is using AI to do better work faster, or to do the same work with fewer people looking at it.

Which parts of my work does AI produce, and which does a person?
Who checks the facts, and how?
Is my data going into a tool, and which one? Client lists, campaign numbers and creative should not be quietly training somebody else's model.
Does the price reflect the time saved? If the drafting takes a fifth of the time, something in the retainer should have moved.

Where it is heading

The direction is clear and it is not replacement. Ad platforms are absorbing the mechanical parts of media buying — targeting, placement, bidding — and answer engines are absorbing the mechanical parts of search. What is left for people is the part that was always the point: knowing the market, knowing the client, having the idea, and being the one who stands there when it goes wrong.

That is the work worth paying for, and the tools make it more visible, not less. The automation that pays is the kind that gives a person back the hours to do it.

Questions people ask

What marketing tasks should a small business hand to AI first?: Creative resizing and variation, first drafts of routine documents, transcription of calls and meetings, and reading campaign reports for changes. These are high-volume, low-judgement tasks where the tool is reliable and the time saved is real. Start there, not with strategy or customer-facing copy.
Can AI write ads in Hindi or regional languages?: It can produce them; it cannot make them sound local. For anything a real audience in a district town will hear, treat the AI draft as a starting point that a native speaker will substantially rewrite. Voice, idiom and rhythm are the whole point in regional creative, and they are exactly what a translation loses.
Is it safe to put client data into AI tools?: Only if you know which tool, what its terms say about training on your inputs, and whether the client has agreed. Campaign numbers, customer lists and unreleased creative should not go into consumer-grade tools by default. Ask your agency the same question.
Should an agency charge less because it uses AI?: Where the tool has genuinely reduced hours on mechanical work, something in the pricing should reflect it. Where the agency's value was never the mechanical work — the idea, the market knowledge, the accountability — the price is for that and should not move. A good agency can tell you which is which.
Will AI replace marketing agencies?: It is replacing the mechanical parts of what agencies did — resizing, drafting, media mechanics — and the platforms themselves are absorbing targeting and bidding. It is not replacing the judgement, the idea, the local knowledge or the accountability. Agencies built on the first set of things are in trouble; agencies built on the second are getting faster.