Somewhere in your ad accounts right now, a machine is writing your headlines. Google’s Performance Max pulls extra headlines from your landing page by default, and it keeps doing that until someone turns it off. In February, OpenAI started testing ads in ChatGPT, so there’s also a brand new channel asking for a piece of your budget.
Both of those are decisions your agency should be making on purpose and telling you why. And there’s no slack in the budget to cover a wrong call. Gartner’s 2025 survey put marketing budgets at 7.7% of company revenue, flat, and 59% of CMOs said they don’t have enough to execute their strategy. At the same time, 80% of the 435 marketing leaders in a Supermetrics survey said they feel moderate or significant pressure to adopt AI within 12 months. Only 6% had fully embedded it. So you’re being told to spend on AI with a budget that didn’t grow, and nobody handed you a scorecard.
Ken Griffin put it bluntly to Bloomberg: “Corporate America has been very uneven in its ability to extract value from AI models.”
I hear the same thing from sales leaders, in different words. They can’t see what marketing produced, and they don’t trust the leads. When a lead takes nine months to become revenue, the only fair way to judge an AI investment is by the quality of those leads and the deals they close. Cheap clicks don’t count.
So here’s how I think about our job. At Level, we’re your AI fiduciary. We test the AI tools and models, pick the ones that earn a place on your account, manage what they spend, and answer for the results. Most of that testing you’ll never see. Working through a hundred options to find the one that fits your account is part of what you’re paying for. Budget goes behind what works and comes off what doesn’t.
Where the word comes from
The term comes from financial planning. A good planner learns your income, your timeline, and how much risk you can stomach, then builds a portfolio to match. They bring you new options when they fit and talk you out of the ones that don’t, even when the pitch sounds great.
That’s the posture to expect from an agency managing your AI. Pick the tools around your business goals. Answer for how the budget gets used. And when the agency’s favorite platform and your goals point in different directions, they change course to stay aligned with your interests.
Here are four things any agency managing your AI owes you. We’ll go first.
1. Your agency teaches the AI which leads matter to your business
Ad platforms optimize toward whatever you tell them to value. Google’s Smart Bidding optimizes for the conversion types in the account’s main Conversions column. If that column counts form fills, the AI gets very good at buying cheap form fills, and with a long sales cycle you won’t see the bill for months.
Fixing that means knowing your funnel better than the platform does. Your agency should be sending what happens after the form fill (a qualified lead, an opportunity, a closed deal) from your CRM back to the platforms, so bidding learns which leads turn into revenue. Google accepts offline conversions for up to 90 days after the click. If your sales cycle is nine months, that window won’t reach a closed deal. At Level, we find the earliest funnel stage that reliably predicts one and agree on it with your sales and finance teams before we let it steer the bidding.
2. Your agency tests new advertising tools and tells you when to pass
ChatGPT ads will get plenty of attention this year. Attention isn’t a reason to spend your money there.
The reason, if there is one, starts with your buyers. In a Responsive survey of 350 B2B buyers, a quarter used generative AI more than traditional search for vendor research, and nearly two-thirds used it at least as much as search. If that’s how your buyers find vendors, a test may be worth your time. If it isn’t, you should hear that from your agency in writing, with the reasoning.
When a test does make sense, your agency sets the rules before a dollar moves: a capped budget, a target cost per qualified lead or opportunity, and a date to decide. At Level, we judge the test against what your account already produces. If it falls short, we stop and put the rest of the money back into the channels that were working, instead of extending the test because we’d already paid for it.
3. A named person answers for anything AI produces
AI may already be writing some of your ads. Google calls it text customization, and it’s on by default in Performance Max. In a regulated category, that’s a compliance question. FINRA has told member firms its content standards apply whether a person or a technology tool generated the communication. AI-written copy gets the same review as copy a person wrote. No exceptions for the robot.
Inside Level, when a tool or workflow earns its place, we write down how to use it, train the team on it, and audit accounts so every client it fits gets it. Nothing goes to a client until a person has reviewed and approved it. Every automated workflow at Level has an owner, and that owner is accountable for what it produces.
Expect the same standard for the automation inside your ad platforms. Someone on your agency team should be able to tell you which automated settings are on in your accounts and why, without looking it up.
4. You see what AI costs and where it spends your budget
There are two costs here. The smaller one is tools, and your agency should be able to tell you which tools it runs and what they cost. At Level, we fund our own AI innovation and test new models against the ones we’re already using, keeping whichever does the job better or cheaper. We use the most cost-efficient model for each task and don’t tie ourselves to any single one.
The bigger cost is your media budget, which the platforms’ AI spends on your behalf every day. Google now offers channel performance reporting for Performance Max, so you can see how a campaign delivered across its channels instead of getting one blended number. Your agency should be using that reporting to show you where the money went and what it produced. If an agency can’t show you both the tool costs and the media spend, you don’t have transparency. You have a summary.
Six questions to ask your agency
Ask us these. Ask any agency these. The answers will tell you whether its AI decisions connect to business outcomes, how it tests new tools, who’s responsible when something goes wrong, and how it reports on costs and results. Ask for examples from your own account.
- Which conversion is our bidding optimizing for, and how far into our funnel does that data reach?
- How many AI tools did you evaluate last quarter, which did you adopt or drop, and why?
- If we tried ChatGPT ads, what would the test cost, how would you judge it, and when would you decide whether to continue?
- Which person reviews AI-written copy on our account, and does our compliance team see it?
- Which automated settings are on in our ad accounts, and who approved them?
- Where did the automated campaigns spend our money last month, and what did it produce?
A good financial planner manages your money in your interest. They bring you what fits, steer you away from what doesn’t, and show you the results. An AI fiduciary does the same thing with the AI on your account and the budget it spends. That’s the job we’ve signed up for at Level. Hold us to it.
