Now You Do Get Fired for Buying IBM

The safest bet just became the riskiest one

“Nobody ever got fired for buying IBM.” For decades that line has been the quiet logic of the enterprise buying committee. Pick the biggest name in the room. If it works, you look smart. If it fails, nobody can blame you, because you did what everyone else did.

That logic assumed one thing: that the big vendor was the scarce, hard-to-copy option. That assumption is gone.

In 2026, the safe choice is no longer safe. If you sign the five-year platform deal, staff your account with forty people doing work a model can do in an afternoon, and let someone else hold your data and your workflows, you are not protecting yourself. Instead, you are making the decision your successor will be hired to undo.

Today, you do get fired for buying IBM. Or worse, you don’t get fired. Your company just quietly falls behind the one that didn’t.

Software is no longer scarce

The moat around enterprise software was always the cost of building it. That cost is collapsing.

A marketing ops lead with a good AI coding tool can now stand up a working internal app in a weekend. A campaign QA checker. A taxonomy validator. A dashboard that reads straight from the warehouse. Not every one of these is ready for everyday business use, but enough of them are that the economics have flipped. When software can be produced in near-infinite supply, the premium shifts away from the code and toward two things that can’t be vibe-coded: your data and your judgment about how work should flow.

Wall Street has noticed. Software valuations have been under real pressure this year as investors ask a blunt question: if a customer can build 70% of what they need in-house, what exactly is the seat license for? Meanwhile, the big platforms are responding the way incumbents always do. Bundle more. Lock in longer. Bolt “AI” onto every SKU and raise the price.

That is the moment to be most careful about signing. A vendor under pressure wants your multi-year commitment more than ever. That should tell you something about who the deal really protects.

The HoldCo math doesn’t work anymore

The agency holding company model runs on a simple formula: bodies on the account, billed by the hour or the FTE. For decades that was rational. Trafficking, QA, reporting, tagging, pacing checks, deck assembly. Somebody had to do it, and it took a lot of somebodies.

Teams can now automate much of that work. Not someday. Now. Yet look at the staffing plan on most big-agency renewals and the headcount hasn’t moved, because headcount is the business model.

Asking a holdco to automate your account is asking it to shrink its own revenue. Some will. Most will move slowly, and you will pay for every month of the delay.

There is a second cost that rarely shows up in the SOW. When the agency does the work, the agency owns the knowledge. Their spreadsheets hold your taxonomy. Their stack holds your attribution logic. Your best practitioners are their employees. Try to switch partners and you discover how much of your marketing operation you never actually owned.

The question to ask on your next review isn’t “how many people are on our account?” Instead, ask “which of these tasks should a person still be doing, and who owns the system when they’re done?”

The new career risk is renting

AI itself is not the moat. Everyone has the same models. Your competitor can license the same tools you can, by Friday.

What they can’t license is your first-party data, cleanly structured and governed. Your content, tagged so a machine can actually use it. Your workflows, mapped well enough that you know which steps to automate and which need a human. That is the raw material AI runs on, and it’s the only part of the stack that’s truly yours.

This makes the old IBM logic dangerous in a new way. The leader who rents everything (data in the vendor’s cloud, logic in the agency’s head, workflows in someone else’s platform) has nothing to point AI at. By contrast, the leader who owns those assets can plug in whatever model wins next year and move.

Boards are starting to ask the question directly: what is our AI advantage? “We bought the market leader” is no longer an answer. Instead, it’s an admission that you don’t have one.

What the smart buyers are doing instead

None of this means rip out your platforms or fire your agency. It means changing what you’re buying and what you insist on keeping.

• Own the foundation. Your data model, taxonomy and metadata standards should be documented, governed and yours, independent of any vendor contract.
• Map the work before you automate it. Know which tasks are judgment and which are just repetition. Most teams have never written this down.
• Buy outcomes, not headcount. Ask every partner to show the automation roadmap for your account, and tie fees to results rather than FTEs.
• Keep contracts short and portable. Data ownership, audit rights and clean exit terms matter more now than any feature list.
• Build where it’s cheap, buy where it’s hard. If a tool can be built in a week, don’t sign three years for it. Instead, save the big commitments for what genuinely can’t be replicated.
• Hire partners who make you less dependent on them. That’s the real test. A good partner leaves you more capable. A bad one leaves you needing them more.

Why we built Transparent Partners this way

Full disclosure: this is the bet we made at Transparent Partners. We’re independent. We don’t sell media and we don’t resell software. That’s not a slogan. It’s the only way we can give you an honest answer to “own or rent?”

Our work focuses on the foundation that makes AI useful: data strategy, taxonomy and metadata, measurement, and the operating model that ties them together. We design it, build it with your team, and hand it over. When we’re done, the system is yours and your people know how to run it.

It’s also why we move fast. Small senior teams, no layers. As Kellanova’s Chief Growth Officer put it: “You made five years of progress in six months.”

So here’s the provocation. Pull up your biggest vendor and agency contracts this week. For each one, ask three questions. What do we own when this ends? How much of this work should a machine be doing? And would I defend this choice to a board that’s asking what our AI advantage is?

If the honest answer is “we went with the safe name,” that’s no longer the safe answer.