DIY or Done-With-You: Who Should Set Up Your AI Marketing?

Set up AI marketing yourself if you use one or two channels, can give it 3 to 5 hours a week for the first month, and mistakes would be cheap. Get done-with-you help, where an expert builds it alongside you and hands it over, when the setup links three or more systems, spends ad money, emails a large list, or you've already stalled twice.

The real difference between the options is who owns the system afterwards. Doing it yourself means you understand every piece but may never finish. Done-for-you gets it finished but often leaves you unable to change a word without paying again. Done-with-you sits between: someone experienced builds it with you, in your accounts, and you leave able to run and adjust it. Which fits depends less on your budget than on how much can go wrong and how often the setup will need changing.

Follow me on Instagram@sagnikteaches

Three ways to get it set up, and who owns it afterwards

The setup jobs, ranked by how hard they are to do alone

"AI marketing" is a bundle of separate jobs, and they vary a lot in difficulty and in what a mistake costs. The hours below are illustrative first-time estimates for a non-technical owner.

Connect on LinkedInSagnik Bhattacharya

A reasonable rule: do the easy rows yourself regardless, and consider help for the hard rows that apply to you. Most small businesses don't need all eight. For ads in particular, setting up conversion tracking before letting AI spend your budget is the job most often done wrong and most expensive to get wrong.

Subscribe on YouTube@codingliquids

Score your situation in five minutes

Give yourself the points in the column that fits best, then add them up.

• 0-5: DIY. Use the tutorials and vendor help pages; you'll learn more than you'd pay for.
• 6-10: DIY the easy jobs, get done-with-you help for one or two hard ones.
• 11-16: Done-with-you for most of it. Done-for-you only if nobody in the business will ever run it, and then insist on owning every account.

A filled-in example: an independent toy shop scores 1 (three channels), 1 (shop and email), 0 (no ads), 1 (2,100 subscribers), 1 (about 3 hours a week), 1 (can follow instructions), 1 (one stalled attempt at email flows), 1 (changes with the seasons). Total 7: DIY the voice guide, calendar and scheduler; get help with the email flows that stalled last time.

A charity shop that should do it itself

A charity shop run by a manager and a rota of volunteers posts donations on Instagram and Facebook, sends a monthly newsletter to about 400 supporters, and spends nothing on ads. It scores 3. The work is simple, mistakes are cheap, and the constraint is continuity, not skill: volunteers come and go.

A DIY plan that works in four weeks:

1. Week 1: the manager writes a one-page voice guide from the ten best past posts, and saves it in a shared assistant project. If the charity qualifies for nonprofit pricing, OpenAI lists ChatGPT Business Standard at $8 a user a month on annual billing and Claude for Nonprofits offers Team at $8 a user a month (two-seat minimum); check eligibility first.
2. Week 2: set up free scheduling in Meta Business Suite and a posting rota.
3. Week 3: write three saved prompts: "new donations post", "event notice", "newsletter intro".
4. Week 4: write a half-page "how we post" note so the next volunteer can pick it up in ten minutes.

One of the week-3 saved prompts, and an illustrative first output:

Write an Instagram post about new donations using our voice guide.
Items: {what came in, with prices}. Keep it under 70 words, lead with
the most unusual item, mention that all proceeds fund {cause}, and
end with our opening hours. Don't describe condition beyond what
I've written.

Illustrative output: "A 1960s typewriter just arrived, and it still types (we checked). Also new on the rails this morning: a stack of wool jumpers from $6 and a box of jigsaws, all pieces counted by a very patient volunteer. Every sale funds our community meals. Open Tuesday to Saturday, 10 till 4."

What the manager fixed: the typewriter line was fine because a volunteer had tested it, but the draft once added "in perfect condition" to a chair nobody had inspected. The prompt's last line, "don't describe condition beyond what I've written", was added after that.

Paying anyone to set this up would buy very little, because the hard part is the rota and the photos, which no expert can do for them.

A subscription box company that shouldn't

The opposite case, sketched with invented but realistic details: a subscription box company with about 4,200 active subscribers, a shop platform with a subscriptions app, an email platform, paid social ads, and one founder handling marketing between everything else. It wants AI to help with three things: renewal and win-back emails personalised to what each subscriber received, ad creative variations, and a monthly report that explains churn.

It scores 13. Three systems must share data about each subscriber; a mistake in a win-back email reaches thousands of people; ad spend is meaningful; and the founder has already started the email flows twice and stopped when the subscription data wouldn't map to email fields.

What a done-with-you engagement might look like here:

• Session 1 (2 hours): map the subscriber journey, agree which three flows matter, check what data each system actually holds. Homework: gather past win-back emails and churn reasons from cancellation surveys.
• Session 2 (3 hours): connect subscription data to the email platform, build the renewal-reminder flow together, write the AI prompt that personalises the opening line from last month's box contents. Homework: review 20 test emails.
• Session 3 (2 hours): build the win-back flow with a holdout group (a slice of lapsed subscribers who don't get it, so the effect can be measured). Set up an ad creative routine using the voice guide.
• Session 4 (2 hours): build the monthly churn summary, go through the runbook, and have the founder make one change unaided while the expert watches.

The founder spends about 9 hours in sessions and 6 to 8 on homework over six weeks, and comes out able to edit any of the flows. Compare that with the two stalled DIY attempts, which cost more hours and produced nothing live. If you'd rather map this kind of setup with someone, that's what my AI implementation consultation covers.

A quick sum: what DIY really costs

DIY looks free because the hours don't arrive as an invoice. Put a number on them. In the subscription box example, the founder's two stalled attempts took an illustrative 25 hours between them. At a notional $60 an hour of founder time, that is $1,500 spent with nothing live, plus the months in which lapsed subscribers got no win-back email at all. If the working flow wins back even an illustrative 20 subscribers a month at $35 each, every month of delay costs $700 in revenue.

Now the charity shop: about 8 hours of the manager's time over four weeks, no ad spend at risk, and a list of 400 supporters. Its DIY cost is low, the cost of delay is small, and nothing an expert adds would change the outcome much. Same method, opposite answer, which is why the scoring table asks about stakes and systems rather than budget.

Signs to switch from DIY part-way through

Starting DIY and bringing in help later is a perfectly good plan, as long as you notice when to switch. Watch for:

• One job has made no progress for three weeks or more.
• You've found yourself copying settings from forum posts without understanding them.
• A test has sent a real email or post to real customers by mistake.
• The job involves ad spend and you can't say what the platform is optimising for.

Any one of these is a reasonable trigger. Keep what you've built, write down where it stands, and bring that to whoever helps; it makes the guided part shorter.

What a done-with-you engagement should hand over

If you pay for guided help, judge it by what you can do without the expert afterwards. At minimum:

• Every tool and automation in accounts you own, with the expert's access removed at the end.
• Saved prompts and the voice guide in a place your team can open.
• A runbook: what each automation does, what to check weekly, and what to do when it breaks.
• One change made by you, not them, before the engagement ends.
• A short list of what to review in three months.

A runbook entry doesn't need to be long. An illustrative one for the subscription box company:

AUTOMATION: Win-back email, lapsed 30 days
What it does: emails subscribers 30 days after cancelling, with an
AI-written opening line based on their last box. 10% holdout.
Where: Email platform > Flows > "Win-back 30d"
Weekly check (Mon, 5 min): flow is live; last 7 days sends vs
cancellations roughly match; no error alerts.
If it breaks: most likely cause is the subscription app renaming a
field. Check Settings > Integrations > field map. Pause the flow
rather than let it send blank personalisation.
Owner: founder. Last reviewed: {date}

How each route tends to fail

DIY fails by stalling. A sports equipment shop owner built three Zaps over two months: one to post new products, one to add enquiries to a spreadsheet, one to email a discount to newsletter sign-ups. Two stopped working after a form was changed; the third sent the discount code twice to everyone. Nobody noticed for three weeks because error alerts were off. The individual pieces were fine; what was missing was testing and a weekly check.

Done-with-you fails when the homework doesn't happen. The sessions are only half the work. If the business can't find a few hours between sessions, the expert ends up building alone and the handover becomes a done-for-you project in disguise.

Done-for-you fails at the exit. A common version: an agency sets up ads and email in accounts it created, the business moves on, and the ad account history, audiences and email templates stay with the agency. Starting again costs months of learning data. Before signing anything, check who owns every account, as covered in the questions to ask before paying anyone to set up AI marketing.

Telling a real done-with-you offer from a rebadged agency

• Does the work happen in your accounts, from day one?
• Are there defined sessions with homework, or just "we'll send it over when it's ready"?
• Is there a written handover, and will they show you an example?
• Does the offer end, or does it roll into a retainer by default?
• Will they tell you which jobs you should do yourself?

The last question is the most revealing. Someone genuinely offering done-with-you help will happily tell you that your scheduler and voice guide are DIY jobs. The broader comparison of these models, beyond marketing, is in done-for-you versus done-with-you AI implementation.

A middle path: build it yourself, then pay for a review

For businesses scoring around 6 to 10, a one-off review is often the best value. You build the setup, then someone experienced spends an hour or two checking it before you rely on it. In the sports shop example above, a review would have caught the missing error alerts, the untested form change and the duplicate discount step in under an hour. Prepare for it by listing every automation, what triggers it and what it should do, and by running each one with a test entry while the reviewer watches. Good reviewers will also tell you what to switch off, which is often as useful as anything they add. Ask for the findings in writing, ranked by risk ("could email the wrong people" above "could be tidier"), so you can fix the dangerous items first and the rest over the following weeks. Before you start building, what to prepare before setting up AI marketing covers the groundwork that makes either route faster.

Questions about getting AI marketing set up

How long does a done-with-you AI marketing setup usually take?

It depends on scope, but most small-business setups fit into a handful of working sessions spread over four to eight weeks, with tasks for you between sessions. The spacing matters: you need time to use what was built and bring back real problems. Be wary of anything promising a full setup in a single afternoon, and equally of open-ended engagements with no defined handover date.

Can I start DIY and switch to done-with-you later?

Yes, and it is often the sensible order. Doing the easy parts yourself first, such as a voice guide, a content routine and a scheduler, teaches you what you actually need and makes any later help cheaper, because the expert starts from your working pieces. Keep notes of what you built and where, including logins and any automations, so the person helping you can see it quickly.

What should I prepare before a done-with-you session?

Bring admin access to every account involved (email platform, shop, social, CRM, ads), ten examples of marketing you're proud of, your current numbers for the metrics you care about, and a short list of the jobs you most want off your plate. Most wasted setup time goes on waiting for passwords or permissions, so sort access out beforehand.

Further reads

• How to Choose an AI Consultant: 20 Questions to Ask First — Questions to ask anyone offering guided setup.
• Do You Need an AI Consultant to Set Up Your Marketing Automation? — Whether marketing automation needs a consultant at all.
• Red Flags When Buying AI Marketing Software or Services — Warning signs in marketing tools and services.
• How Long Does AI Marketing Automation Take to Set Up and Pay Off? — Realistic setup time and payback periods.
• Should You Hire an Automation Consultant or Build It Yourself? — The build-or-hire question for automations generally.
• AI Consultant vs Agency vs Freelancer: Which Should You Hire? — Which kind of outside help fits your situation.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI and Anthropic nonprofit pricing pages; Zapier plan limits (checked September 2026). Hour estimates are illustrative.