Beth Mach: Stop Automating the Mess: AI Doesn’t Fix a Bad Operating Model, It Makes It Move Faster
The fastest way to destroy value with artificial intelligence (AI) is to point it at a process nobody ever bothered to fix. Companies across media and advertising are doing exactly that, layering automation over manual workarounds, inconsistent data entry, and decisions no one can explain after the fact, then congratulating themselves when the reports arrive sooner. Beth Mach, who leads operations at Spacely Media, argues the speed is the problem, not the payoff. A broken process that runs at machine pace does not become efficient. It becomes efficiently wrong, at scale, with fewer people left in the loop to notice. And as buyer-side and seller-side agents begin transacting media directly through open protocols, the tolerance for a weak operating model is about to collapse.
Speed Is Not The Same As Structure
Mach is blunt about what most efficiency gains amount to. “Things work faster, reports are generated faster, you don’t need as many people to do the work,” she says, “but if your data isn’t consistent, if you don’t have consistent decision-making and you’ve got a lot of manual tasks that you’re still kind of doing and maybe even working around stuff, that’s when it starts to break down because you’re just building technology on basically bad data.” The verdict follows directly: “It looks faster, but truly it’s worse, because ultimately that means that the process isn’t structured and repeatable.”
That reframing has changed how Mach talks about her own market. Rather than selling speed, she describes the work as building infrastructure where none existed, taking inventory, rates, and specs that live in scattered manual-entry places across agency and brand systems, and rendering them as a consistent view. Her test for any workflow is worth borrowing: What can be structured? What can be transparent? What can be transferable? Anything that fails all three is not a candidate for automation. It is a candidate for redesign. Mach draws the parallel to the early big data years, when organizations celebrated volume before anyone asked what to do with it. “We have so many things that we can do, but what do you do with it, and does it matter?”
The Tell That Separates Slow From Broken
Leaders routinely misdiagnose their own operations, treating a broken process as a slow one and automating it into permanence. Mach’s diagnostic is the reliability of the outcome, and the traceability behind it.
A slow process may take too long, but it still produces consistent, explainable results. A broken process creates repeated work, conflicting answers, or decisions that no one can later explain. “If the person responsible cannot walk you through how or why a decision was made, the problem is not speed,” she says. “It is a lack of clarity, ownership, or a reliable record.” Those are the fingerprints of a process that needs rebuilding before any software touches it.
The corollary is uncomfortable for anyone thinning out experienced headcount. Mach’s concern is whether the remaining workforce can still think critically enough to challenge a plausible-looking output. She describes a veteran media professional reviewing a one-sheet and spotting a basic error in the market calculation: the age segment had been prorated incorrectly, which inflated the estimated audience size. Left uncorrected, those inflated numbers get guaranteed to an advertiser, the campaign underdelivers, and the seller pays it back in refunds or free inventory while its reputation absorbs the rest. “If you don’t have some of that explicit experience to challenge questionable assumptions,” Mach says, “then we’re just going to be using bad data everywhere.” Judgment, in other words, is not a soft skill. It is the last control on a system moving faster than its inputs deserve.
What 12 Months Of Repair Work Looks Like
Agentic buyers already transact digital programmatic inventory, and Mach is clear about why that channel was ready first: clear definitions, structured data, current availability, and tight transactional rules. Channels without a digital backbone – print, out of home, in venue – have none of that, which means open protocols expose them rather than serve them. What agents cannot do is extract what sits inside relationships: the seller who understands what a buyer meant, or the operations person who catches a mistake before it becomes an order. Agents will not surface conflict, unclear pricing, or duplicate opportunities the way a human does. “If you don’t have some of the nuances accounted for, you’re not ready for it.”
Mach’s prescription for the next year is narrow and practical. Standardize what data should look like and document it. Define what technology can do today versus what it is expected to do next. Establish a clear source of truth: where the data comes from, how often it updates, and whether you can consistently trust it. Then audit the human steps, using a distinction Mach considers the heart of the matter. Ask where judgment creates value, and separate that from “where human involvement exists only because the process has never been redesigned.” Some tasks are held onto because they make people feel valuable, not because they require a person. “Just because you’ve always done it that way doesn’t mean that’s the way it should always be done.” Her read on who wins is equally direct.
Companies that make their supply and rules clear and executable, with decision-making traceable, will be ready for autonomous buying. Those clinging to inventory and keeping it illegible are making themselves hard to transact with, usually out of fear that visibility costs them their jobs. Mach sees the opposite risk. “If they don’t do those things, they will not be ready, and that will cost them business. Period.” The window for that repair work is narrower than most leaders assume, because a market that moves to structured, agent-readable transactions does not wait for the unstructured to catch up.
Follow Beth Mach on LinkedIn for more insights on operating model design, data standardization, and preparing media businesses for autonomous buying.
