Doing your own digital marketing: when it works, when it costs you, what agents change.

The dishonest framings first, so we can skip them: "anyone can do this in ten minutes a day" (they can't; the tools are easy, the judgement isn't), its newer cousin "the AI will do it for you" (it will draft it; it won't know which draft is true), and "leave it to the professionals" (for half of this list, you are closer to the customer than any professional will ever be). The truth splits down the middle.

The genuinely DIY-able list

Everything on this list rewards proximity (being the person closest to the customers and the work) more than expertise. Your Google Business Profile: categories, services, hours, photos of real work; nobody knows the truth of the business better. Reviews: the ask at the end of the work is inherently yours, since no agency was there when you delivered it (the review engine is fully DIY-able). Social capture: the thirty-seconds-a-time habit that feeds the feed (pass the inspection); again, only you were there. Ads hygiene: reading your own search-terms report monthly and adding negatives is teachable in an afternoon and catches the worst waste. First drafts: AI tools can turn your notes into posts, ad variants and page copy in moments, which moves production onto this list, as long as you check every claim, because that part was always yours. None of this requires a marketing brain. It requires the owner's eyes and a habit.

The specialist list

The other half punishes learning-by-error, because the errors are silent, expensive, or both. Technical SEO (crawling, indexing, site migrations, the structured data machines read), where a wrong move can cost rankings you spent years earning and the symptom shows up months later. Tracking plumbing (conversion events, call attribution, server-side APIs, consent settings), where a subtle misconfiguration means every downstream decision is made on wrong numbers that look right. Campaign architecture at real budgets, where the difference between a fenced account and a leaky one compounds monthly, and automated campaign settings widen the leak when nobody sets the fences. The pattern: DIY-able work fails loudly and cheaply; specialist work fails quietly and dearly. Sort any task by how it fails and the list writes itself.

The real cost of DIY

The DIY budget looks free because nobody invoices you. It's paid in two other currencies. Hours: the evenings that were going somewhere else, indefinitely, because marketing is never finished. And tuition-by-error: the months a campaign runs unfenced, the tracking that counted wrong while decisions leaned on it, the ranking asset dented by a well-meant change. AI tools cut the first currency more than the second: they save hours on drafting, and they can copy a mistake across every page at once. Sometimes that tuition is cheap and worth paying; that's how owners genuinely learn. The mistake is not counting it. If an hour of your time is worth real money, an honest DIY costing multiplies the hours by it and adds the errors, and sometimes DIY still wins that sum. Then it's the right call, made properly.

The middle path nobody sells you

Between "hand it all over" and "figure it out alone" sits the option the industry under-advertises because it ends: mentored DIY. You run the accounts, own every login, do the work, while someone who does this daily reviews it, catches the silent failures early, and teaches the judgement the tools don't come with. The economics favour you: expertise arrives by the hour, capability stays in the business, and the arrangement is designed to make itself unnecessary. It's what we'd recommend to a friend with more time than budget. Strategy & mentorship is that lane here, and team training is its cousin for when the "you" is staff rather than the owner.

The agent-native lens: your facts checked by software, your hours shared with it

When the buyer's assistant checks your homework. The proximity work on the DIY list is much of what assistants read when they research for a customer: your business profile, your reviews and replies, the facts on your site. That puts the premium on accuracy over polish. An assistant comparing options is looking for hours, coverage, prices or specifications that agree wherever it looks, and for recent evidence behind what you claim. Keeping those facts current and consistent is squarely DIY work, because nobody knows them better than you. Expressing them as schema.org JSON-LD so software can read them reliably sits nearer the specialist list: when markup is wrong, nothing visibly breaks.

When the DIY team is you plus agents. Agents shift the first fork more than the others. They can draft posts from your notes, propose ad negatives, check pages for broken links and markup, and summarise last month's numbers, so "do the hours exist?" gets easier to answer yes. They don't move the stakes fork: an agent can make a silent failure faster, and at a scale no person would. Run one the way you'd run a capable junior. Read access first, write access per task, every account in your name, its working shown, and your sign-off on anything that spends money, publishes as you or touches tracking. The mentored lane still fits; the expert eyes simply review the agent's output alongside yours.

Decide by stakes, not pride

Both wrong answers come from ego. Owners too proud to pay for help run high-stakes plumbing themselves and pay tuition they never counted; owners too intimidated to touch anything delegate the ask-for-a-review (work only they can do) to a retainer. Use the tree: hours first, failure-cost second, appetite third. Most growing businesses land on a blend: DIY the proximity work, delegate the silent-failure work, mentor the gap between. Agents are likely to keep moving tasks across those lines; sorting by how a task fails is the test that stays put. And if the answer comes up "delegate", the next skill is picking who: how to choose a digital agency is the companion piece, red flags included.