The first in a series drawn from the Women in Marketing Procurement Summer Series, featuring The Brandtech Group’s Karen Bennett (Jellyfish), Matt Franceschi (Pencil), and Tamara Carvalho (Oliver).
Fifty percent savings. Thirty percent efficiencies. 10x content for the same cost. Etc. Etc.
This reflects the conversation, with some variation, that you will have with every AI content platform pitching your organization this year.
It is also the number that should make every marketing procurement professional sit up and ask the question that we put to the panel: how real are the dollars?
We talk about hard dollars and soft dollars in this business for a reason. “If my CMO reinvests the savings, I need to know exactly how much there is to reinvest”, said a webinar attendee. “If we are having a tough year and finance wants it on the bottom line, I need to see the money” she continued. So, which is it?
The answer, it turns out, is both. And neither. And that is precisely why procurement needs to be in this conversation from day one.
The demo is impressive. The P&L question is harder.
Let us be clear about what we saw. Matt Franceschi walked us through Pencil, The Brandtech Group’s AI operating system, and the capability is genuinely striking: end-to-end workflows from strategy through production to performance, dozens of AI models aggregated under a single agreement, predictive scoring of creative before a dollar of media is spent. Karen Bennett was candid about where the market is: “Demos are over, it is time to prove the scale”. The technology has moved past the proof-of-concept phase. Brandtech points to a global consumer brand running content at 10x volume and a global liquor brand pairing a reported ROAS lift with a 33% cost reduction.
But headline numbers are where the procurement work begins, not where it ends. So, we pushed on the savings question, and the answers were more instructive than the slides.
Where the money actually shows up
Franceschi’s most tangible example was studio avoidance. “There’s an amount of money Brands just don’t have to spend”,he explained, describing how a brand like L’Oreal can generate environments, backgrounds, and product swaps rather than mounting a shoot, traveling to location, or, the classic budget-killer, reshooting because something was missed the first time.
That is hard dollars. It was in the production budget, and now it is not. You can find it, baseline it, and bank it.
But then the answer shifted. Improved ad performance, a lower cost per click, a higher return on ad spend: “maybe that’s a slightly different line item in the business”, Franceschi acknowledged. He is right, and procurement should pay attention to what that means. The efficiency shows up in the media budget, not the production budget. It is real value, but it belongs to a different owner, a different baseline, and a different conversation with finance.
Tamara Carvalho added a third category that will resonate with anyone who has audited agency fees: revision cycles. Brands burn enormous amounts of time and money getting to approval. Build your brand guidelines, your compliance requirements, and even your “we never say this” rules into the platform, and assets are delivered closer to “perfect” the first time. “That’s a huge benefit right there”, she said. True. But is it a hard saving, or is it reclaimed capacity, hours your team and your agency no longer waste, which only becomes money if you reduce fees, reduce scope, or redeploy the time to work you would otherwise have paid for?
Three categories of savings, then. Avoided cost: hard, findable, bankable. Media efficiency: real, but it lands in someone else’s budget line. Reclaimed capacity: soft until you contract it into something hard.
Agencies and Tech partners will quote you the sum of all three as one big number. But you know that your CFO will not accept it as one big number. And probably, neither should you.
The industry data says: verify before you bank
This has nothing to do with anyone doubting the technology. It is building discipline into the claim, and market data supports the caution re recommend.
Gartner reported last year that generative AI for procurement had entered what it calls the trench of disillusionment, with many organizations experiencing uneven ROI against expectations, often because of fragmented, low-quality data. Deloitte Digital’s survey of marketing leaders found that among brands that had implemented generative AI in marketing operations, 41% said it had already reduced content production costs. That is a meaningful result. HOWEVER, it is also a long way from “everyone saves 50%”, we hear so often.
In addition to the savings math, there is also a cost side that rarely makes the pitch deck. Digiday reported on the hidden costs of AI-assisted production, noting that Coca-Cola’s AI Christmas ad involved some 70,000 individual prompts, and that un-disciplined use of premium models can quietly (or loudly) erode the efficiency case. Add licensing, integration, training, governance, and the change management that Karen Bennett rightly flagged as the “people-tech challenge”, and your net savings number moves a fair amount in the wrong direction.
None of this means the 50% is wrong. It means the 50% is unverified until you do the work.
What procurement should do before signing anything
This is where our function earns its seat. Four practical moves:
Baseline before you deploy. You cannot claim savings against a number you never established. What does a piece of content cost you today, by type, by market, fully loaded? If you cannot answer that, the vendor’s savings number cannot be validated, and those unverified numbers have a way of often appearing in someone’s Board deck.
Agree the savings taxonomy with finance up front. Define, in writing, what counts as hard (avoided external cost), what counts as soft (reclaimed hours, cycle time), and where media efficiency gets booked. Do this before the pilot, not after, when everyone has an incentive to grade their own homework.
Run the analysis. One genuinely useful offer from the session: Franceschi described taking a representative cross-section of 60 to 80 briefs and sorting them into three production lanes, fully AI-suitable, AI-plus-human-craft, and mostly human. That analysis, done honestly, is your savings model. It tells you what proportion of your content can generate savings for your Brand. Do this work with your agency, understand the math, then keep a savings score card on hand for every production.
Contract for the measurement, not just the tool. If the platform tracks time, cost, and generations from brief to approved asset, as Pencil does, then that audit trail should be yours, specified in the agreement, with reporting you define. The savings claim should be a contractual commitment with a measurement methodology.
The reinvestment conversation is the real prize
Franceschi made a point near the end of the session that deserves more attention: the goal is not just to spend less, it is to reinvest what you free up, in more strategic work, in growth, in the things your teams never have time for. He is right, and it is a better conversation than pure cost-cutting to have with stakeholders.
But here is the commercial reality: you cannot reinvest what you cannot count. The reinvestment case and the bottom-line case both depend on the same discipline, knowing which dollars are hard, which are soft, and which belong to a different budget entirely.
AI content platforms are going to transform production economics. We strongly believe that after what we saw. But transformation without measurement is just a story, and procurement’s job, now more than ever, is to make sure the story reconciles.
Ask the question. How real are the dollars? Then do not accept an answer you cannot take to your CFO.