The Architect’s Dilemma

Byline: Patricia Collins ran growth strategy across IBM's $30B Cloud portfolio and held the CMO seat at one of the first IoT startups. She now advises operators living the gap she's describing.

For the VP of Revenue Operations who built the revenue engine — and can’t stop the AI now running inside it.

You built the revenue system. The data model, the routing, the lead scoring, the tools, the integrations — the whole machine the company runs revenue on. So when the forecast misses, the question comes to you. When the AI inside the scoring quietly drops a segment and a quarter of pipeline disappears, you’re the one explaining it. You know the revenue system better than anyone. And you can’t stop it.

The condition, in plain terms

This is the Architect’s Dilemma. You’re accountable for a system you’re not allowed to govern.

That’s not “get a seat at the table.” It’s not “show your impact better.” Those are coaching answers to a structural problem, and they waste your time. The real problem is simple. You were given the accountability for the AI. You were never given authority over it. You can name every system you’re responsible for. You can’t name one you’re allowed to shut off.

And here's the question under it — the one that decides whether you do anything. Are you authorized for the scope you carry? Or are you holding the risk of the whole revenue system with the decision rights of a function lead? Those are two different jobs. The org chart treats them as one.

What it actually looks like

Monday morning. Forecast review.

The CRO points at the board slide.

"Pipeline conversion dropped three points — is the AI scoring model still good?"

Everyone turns toward you.

You know the scoring model shifted after a 6sense intent update. You know SDRs started rewriting outbound with ChatGPT. You know marketing layered another tool into routing.
You can explain every moving part. You can authorize changes to almost none of it.
Then the second question lands: "So who can actually fix it?"
The room waits for you to answer that one too.

Take the VP of Revenue Operations at a $120M company. She built the lead routing. She built the scoring that decides which accounts reps see first. Eighteen months ago, the company added an AI model on top to “optimize” it. She didn’t pick the vendor — the CRO and procurement did. She didn’t sign off on the training data — that meeting happened without her. But the scoring model runs inside her system. When it works, it’s “the RevOps engine.” When it breaks — when it quietly buries a profitable segment because the data under-weighted it — the CRO asks her why that pipeline dried up.

She can explain exactly what went wrong. She can’t fix it. She’s not allowed to override the model, pause it, or change its inputs. Those calls belong to the CRO, who doesn’t understand the system, and the vendor, who doesn’t work for her. So she owns the outcome of a system she built but can’t control.

She spends three weeks building workarounds — manual overrides, side spreadsheets, a shadow scoring layer. Now she’s more essential and less visible at the same time. The better she patches it, the more the gap disappears from view. Upstairs, everything looks fine. Until the next time it breaks. Then it runs again.

Why it gets worse the longer it runs

Here’s the trap that keeps good operators stuck for years. Every workaround you build to cover for missing authority makes you more essential and harder to read. The shadow layer, the manual fixes, the side process that keeps the AI from doing damage — each one quietly takes the risk back onto you. And each one hides the proof that the structure is broken.

From above, the revenue stack looks healthy. It looks healthy because you’re holding it up. Your skill is covering the defect. That’s how a structural problem hides as a personal strength. It’s why the people best placed to expose the gap are the least likely to. To expose it, you’d have to let the thing you built wobble in public.

Eventually the company calls this a retention problem. Operators carrying scope without the authority to govern it hit a wall. They check out, or they leave, and the knowledge of the system walks out with them. The company calls it culture. It wasn't. It was an authority problem nobody fixed until the only option left was quitting. The Conference Board and ADP tracked this pattern in January 2026: roles expanding faster than the authority and title assigned to them.

The wrong fixes

The company has a name for what she’s living, and every name is wrong.

It calls it a tools problem. So it buys more governance software. Now there are governance dashboards no one is allowed to act on.

It calls it a communication problem. So someone tells her to “make her impact more visible.” Now there’s more reporting. Nothing changed about who can stop the scoring model.

It calls it a teamwork problem. So it books a cross-functional meeting. Everyone agrees the AI matters. No one leaves with the authority to halt it.

The most expensive wrong answer is the one that sounds smartest: influence without authority. “Lead through influence” really means “keep absorbing risk you have no power to control.” It works until it doesn’t. And when it doesn’t, no one asks how much influence you had. They ask who owned the decision. In 2026, the American Management Association found 69% of leaders already spend at least half their time influencing people they have no authority over. That’s not a skill to build. It’s a broken structure, running everywhere, that companies have written into the job.

What this costs you

Here’s the number that should change how you see your situation. In Grant Thornton’s 2026 AI Impact Survey of about 950 senior U.S. executives, three in four boards had approved major AI investments — but only 52% had set clear AI governance expectations. The money went in. The authority to govern it didn’t. Unassigned authority doesn’t vanish; it settles on the lowest person who can still be blamed — and that’s always the one who understands the machine. That’s you.

Play that forward, because this is where it stops being abstract. The scoring model you can’t override keeps making calls — which accounts get worked, which segments get buried, which leads a rep ever sees. Run it long enough and one of those calls is wrong in a way that leaves the building: a profitable segment starved for two quarters, a compliance-sensitive lead routed somewhere it shouldn’t go, a forecast the board acted on that was built on a model no one was authorized to question. Now it isn’t a RevOps inconvenience. It’s a revenue miss the company can’t explain, or a decision a regulator wants accounted for — and the first question is always the same: who owned it?

This isn’t hypothetical anymore. In Moffatt v. Air Canada (2024), a tribunal held the company liable for what its AI told a customer and rejected the argument that the system was a separate entity answering for itself. The organization owned the output of a system it didn’t directly control. Read that as the architect: you can be held answerable for an automated decision you had no authority to make. That principle is now in the open.

And it compounds. Every quarter the gap runs, more decisions move into models, tools, and automations — and the authority to govern them doesn’t move with them. The pile of un-owned automated risk grows, and all of it routes, by default, to the person closest to the machine. When it finally surfaces, “no one owned it” isn’t a defense. It’s the finding. And the name on the system becomes the answer to “who owned this” — not because you were given the authority, but because you were never given a way to say no to the accountability.

What fixed looks like

The fix isn’t more influence. It’s structural. The authority around the system gets redrawn so it finally matches what you’re already accountable for — owned by the right people, not resting on you by default.

What that looks like in the end is simple. Your title and your pay catch up to the system you actually run. The risks you were never meant to hold come off your name. The ones you control stay.

How you get there depends on your specific setup — which is what the Authority Checklist is for. It shows you where the gap is and what to deal with first.

The move

One thing you can do this week.

Take the AI decisions you’re accountable for but can’t override.

Bring them to your skip-level.

Not as a complaint — as a question: who’s allowed to stop each one?

Then watch. The silence isn’t them not knowing.

It’s the risk you’ve been carrying, finally out in the open.

Run the BluShift Authority Checklist.  The Authority Checklist locates your authority gap — where you’re accountable for outcomes you don’t have the authority to govern — and what to redraw first.

Patricia Collins • Founder, Blumaverick • Author

Sources:

Grant Thornton 2026 (52% boards w/ clear AI governance), AMA 2026 (69%), Moffatt v. Air Canada 2024 (tribunal)