Should a marketing team build AI systems, buy tools, or hire someone?
The short answer
Buy an off-the-shelf tool when it genuinely runs the workflow end to end, typically $100 to $500 a month for what it covers. Hire when you need judgment across many problems, not throughput on one. Build a custom system when a specific workflow eats recurring hours and no tool runs it end to end: building is the only option where the capability compounds and you still own it when you stop paying.
The question behind the question
Most marketing teams added AI this past year. Most got faster drafts and busier dashboards, and the workload didn’t shrink. That’s the tell that the team bought tools when the problem was workflows.
A tool helps a person do a task faster. A system runs the workflow end to end… it pulls the data and does the analysis, then drafts the output… and a person refines the result. Shaving minutes off a task isn’t capacity. Removing the task entirely is. Which option gets you there depends on the workflow.
The comparison at a glance
| Buy a tool | Hire a person | Build a system | |
|---|---|---|---|
| Typical cost | $79 to $500/mo | $120,000+/yr loaded | Low four to five figures, fixed |
| Time to value | Days | Months (ramp) | Weeks |
| Who owns it | Rented… gone when you cancel | Walks out with notice | Yours at handoff: code, docs, runbook |
| Best for | Generic workflows thousands of teams share | Judgment spread across many problems | A specific recurring workflow no tool runs end to end |
| Fails when | The workflow is yours, not generic | You needed throughput, not judgment | Nobody on your team owns it after handoff |
When to buy a tool
Buy when the workflow is generic enough that thousands of teams share it. Email sends, basic scheduling, transcription, design templates. The vendors have seen more edge cases than you ever will, and $79 to $500 a month is usually unbeatable for what they cover.
The two costs people miss: tools are rented, so the capability walks out the day you cancel… and every tool is another silo someone manages. Ten point tools is ten logins, ten exports, and a human gluing them together. The glue is usually the actual workload.
When to hire
Hire when you need judgment spread across many problems: a marketer who decides what’s worth doing, not a faster way to do it. What hiring doesn’t solve is throughput on repetitive work. A $120,000 hire spending mornings assembling reports is the most expensive reporting system ever built, and in 2026 they’ll quietly resent it, because they know a system should be doing it.
When to build
Build when a specific, recurring workflow eats real hours and no tool runs it end to end against your data. The signature cases we see:
- Reporting. Data lives in four platforms, the deck is due weekly, and someone senior burns half a day assembling it. A built system drafts it from live data and a human refines it.
- Monitoring. Spend, pacing, feed health, disapprovals… things a human checks daily until the one day they don’t. Systems don’t skip days.
- Analysis at depth no human sustains. Reading every search term weekly. Watching every frame of every ad and correlating with performance. Humans can do this once, not every week forever.
- Structured production. Campaign buildouts, product-page generation, variant drafts… anything that’s mostly structure with a judgment layer on top.
What custom builds cost
The build option was mostly theoretical for small teams until recently, because custom marketing software historically meant engineers, five to six figures, and quarters of calendar time. Agent-era tooling changed the shape: purpose-built marketing systems now commonly land in the low four to five figures and take weeks, not quarters. Our own builds run from about $5,000 fixed with the buyer owning everything at handoff… code, docs, runbook. When you stop paying, the system stays. That ownership line is the sharpest difference between building and everything else: tools and agencies are rent, a system is an asset that compounds.
The decision rule
Ask one question per workflow: does something off the shelf run this end to end?
- Yes → buy it. Genuinely. If a cheap tool covers the workflow, a custom build is vanity.
- No, and the workflow eats recurring senior hours → build, and make sure a person on your team owns the system after handoff.
- No, and the problem is judgment rather than throughput → hire, or rent judgment fractionally.
Most teams need all three in different places. The expensive mistake isn’t picking the wrong one once… it’s defaulting to tools for everything and wondering why the team is still buried.