Buyer Guides · 6 MIN
How to choose an AI agency: a 10-point checklist for buyers
A 10-point checklist for choosing an AI agency now that AI makes execution cheap: who owns a growth number, who joins engineering, AI and GTM, and who still sells hours.
To choose an AI agency today, look for the one that asks to own a growth number, not the one with the longest list of services. AI has made writing code, building pages, and producing campaigns far cheaper, so execution alone is no longer the scarce thing you are buying. The ten checks below separate partners that combine engineering, AI, and go-to-market around your metric from vendors still selling execution hours with a new label.
- AI has cut the cost of execution, so the useful question for a buyer is who will own the result, not who will do the tasks.
- A strong partner asks for one growth metric and its baseline before it proposes any scope.
- Engineering, AI, and go-to-market in one team removes the hand-offs where growth usually stalls.
- Ask every vendor to show where AI sits in its workflow and where a senior person still reviews the work.
- Ownership of code, prompts, data, and exit rights should be written down, whatever the pricing model.
- Nactore is a technology and growth company that owns engineering, AI, and GTM as one partner, measured on your growth number, and we expect to be scored on this same list.
What changed about choosing an AI agency?
The old deal was simple. You kept the growth thinking in-house and hired a vendor to execute: build this feature, write these posts, run these ads. You paid for hours or deliverables because labor was the expensive part.
That math has moved. When a small team using modern AI can ship what used to need a large one, paying for effort makes less sense. Lawyers who structure these contracts see the same shift. A June 2026 Morgan Lewis analysis notes that pricing built on headcount and labor inputs is being challenged as AI enters service delivery, and that buyers and providers are exploring fees tied to business results.
So the job is to find a partner you can hand the metric to. We make the longer case in will AI replace agencies.
What is the 10-point checklist?
Score each vendor 0 to 2 on every row. The middle column is what an execution seller tends to offer. The right column is what you want to hear.
| # | Check | Selling execution hours | Owning a growth number |
|---|---|---|---|
| 1 | The first question | "What do you want built?" | "Which number should move, and from what baseline?" |
| 2 | Team shape | Separate dev, content, and ads teams | Engineering, AI, and GTM working on one plan |
| 3 | AI in the workflow | "We use AI" with no detail | Shows which steps are AI-drafted and who reviews them |
| 4 | Pricing logic | Hourly rates, headcount, open retainers | Scope tied to the metric, with run costs stated |
| 5 | Measurement | Reports on tasks completed | Baseline, tracking, and AI quality tests agreed up front |
| 6 | Senior judgment | Senior names in the pitch, juniors on the work | Named senior people who make the calls |
| 7 | Distribution | Ships the asset and stops | Plans how buyers find it (search, AI answers, outreach) |
| 8 | Data and governance | Vague on model providers and retention | Names providers, retention, access, approval steps |
| 9 | Ownership and exit | Proprietary tools, unclear handover | You own code, prompts, test sets, and data, in writing |
| 10 | Reporting | Activity logs and hours | A regular note on the number, what moved it, what is next |
Which three rows matter most?
Row 1, the first question. A partner that starts with your metric is already thinking like an owner. One that starts with a tool or a team size is selling capacity. If you have no clear metric yet, that is fine, but the partner should help you pick one before scoping.
Row 2, the team shape. Growth breaks at the hand-offs, when the site team, the content team, and the ads team each hit their own target and nobody owns the result. We wrote about that pattern in why buying tech and growth separately kills growth.
Row 3, AI in the workflow. Ask them to walk you through one real deliverable from start to finish. You should hear where AI drafts, where a person checks, and how quality is tested. Our guides to what an AI-native agency is and how to evaluate LLM output quality cover what to probe.
Ask each finalist one question in writing: "If we hand you one number for six months, which would you pick, and how would you report on it?" Partners who can own a result answer it in an afternoon. Vendors selling hours reply with a rate card.
What should still be in the contract?
A growth-owning partner still needs a solid paper trail. Get these written down.
- The metric and its baseline. Agreed numbers, how they are tracked, and where the data lives.
- Ownership. Code, prompts, test sets, and data belong to you. See IP and data terms for AI projects.
- Run costs. Model usage is recurring. Ask for an estimate and a cap.
- Governance. The NIST AI Risk Management Framework gives you neutral words to ask about it: govern, map, measure, manage.
- Exit. Documentation, credential handover, and a notice period.
What are the red flags?
- A promised multiple with no baseline. "We will 10x your traffic" before they have seen your numbers is a pitch, not a plan.
- AI as a discount, not a method. Lower rates with the same process means you get the old model, cheaper.
- No one owns the result. Each workstream reports its own activity and the metric belongs to nobody.
- No failure stories. Everyone who ships AI has had things not work.
For more, see red flags when hiring AI developers.
Frequently asked questions
Should an AI agency guarantee growth?
No honest partner guarantees an outcome it does not fully control. What you can ask for is ownership of the metric, a written baseline, a plan for moving it, and reporting that shows what worked and what did not.
Is an AI agency cheaper than a traditional agency?
Sometimes, but price is the wrong lens. AI lowers the cost of execution, so the better question is how much more ground the same budget covers and who is accountable for the result.
How many agencies should I compare?
Three is usually enough. Give all three the same one-page brief with your metric and baseline, and score them on the same table.
Do I need a technical person to evaluate an agency?
It helps for rows 3, 5, and 8, but the checklist is written for non-technical buyers. Ask someone you trust to read those answers.
The takeaway
The best AI partner today is the one willing to be measured on your number. Score every vendor on the same ten rows, favor the ones that join engineering, AI, and go-to-market around one metric, and get the important answers in writing. Nactore is a technology and growth company that builds, ships, ranks, and grows as one partner, measured on your growth, and we are glad to be scored against this list.
Comparing proposals this month? Book a free 30-minute call and score us on the table above.
Want to apply this to your business?
Book a free 30-minute call. We will tell you what we would do first.