Your 2027 Plan Has No Stop List. That Is the Real Risk.

This week more than 7,000 CIOs gathered in Orlando for Gartner's IT Symposium, and the opening keynote was blunt. AI risk is now compounding faster than AI value. Most of those leaders will fly home and approve a longer list.

Gartner projects average IT budgets will grow 3.7 percent in 2027, while AI investment grows much faster. That gap does not close by itself. Something is going to stop next year. In most companies, nobody is deciding what.

Here is the uncomfortable part. Most annual plans are not strategies. They are ceasefires. Every leader got something, so nobody had to fight, and the board got a document everyone could sign. A plan that offends no one has not made a single choice.

This is the part most planning misses. Work always stops. When leadership does not choose what stops, the organization chooses for it.

The best engineer resigns, and her project stops. A key customer escalates, and the roadmap stops. Burnout spreads, and quality stops. The stop list exists in every company. It is simply written by accident, after the plan is approved, by whoever has the least ability to push back.

In a large enterprise that waste hides inside a big budget. In a portfolio company with five million dollars of EBITDA, it has nowhere to hide. There may be three people who really carry the product. If those three are spread across eleven priorities, the company does not have eleven priorities. It has three exhausted people, and one real priority, which is keeping everyone in the leadership meeting happy.

Busy is not a strategy. It is a place to hide.

Work always stops. The only question is who decides what.

The fastest diagnostic I know takes an afternoon. Do not start with the plan. Start with where the engineering and product hours actually went last quarter.

A roadmap is a press release. A timesheet is a confession. The roadmap tells you what leadership believes. The hours tell you what the company does. The gap between the two is the first real finding of any engagement worth doing.

That gap is never random. It is custom work a salesperson promised to close a deal. Maintenance on a platform everyone agreed to retire two planning cycles ago. A pilot that was never cancelled, so it never ended. Support for a feature three customers use.

None of that shows up in the board deck. All of it shows up in the hours. Once a leadership team sees how much of its most expensive talent is pointed at nothing on the plan, the priorities conversation changes on its own. It stops being a debate about opinions. It becomes a decision about evidence.

So ask yourself the harder question. If those hours were pulled tomorrow, would you put them in front of your board? Would you put them in front of a buyer? If the answer is no, the plan you are about to approve is fiction, and somebody on your team already knows it.

A roadmap is a press release. A timesheet is a confession.

Most advice on prioritization assumes the problem is judgment. Pick better. In practice, leaders usually know what should stop. The work keeps running because something is still paying for it. There are three culprits, and the second one almost nobody checks.

The sponsor. Leaders defend what they launched. The platform they selected, the program they sponsored, the committee they created. Stopping it feels like admitting the original call was wrong. The fix is one sentence the whole leadership team learns to say out loud. This was the right decision then. It is not the best use of our capacity now. That sentence lets a sponsor retire their own work without losing face. Without it, sunk cost wins every time, and sunk cost is not a strategy. It is a memorial.

The pay plan. I have spent much of my career around acquisitions, and I have watched good products stall after a deal because the sales compensation plan stopped rewarding them. The same mechanism runs in reverse. Bad work survives because the pay plan still rewards it. If a salesperson is still paid on a product line, a custom feature, or a service tier, that work has not been stopped. It has only been defunded. Engineering gets pulled back to it the first time a deal depends on it. Show me the comp plan and I will show you the real roadmap.

The promise. In healthcare and industrial technology, roadmaps get sold. A health system or a plant manager was told a capability was coming, and sometimes the contract says so. Leadership cancels the work internally, nobody tells the customer, and the work returns six months later as an escalation at a far higher cost. Sales did not sell a feature. They sold next quarter's engineering capacity, and nobody in product was asked.

Show me the comp plan and I will show you the real roadmap.

A stop that does not change the sponsor's scorecard, the pay plan, and the customer commitment is not a stop. It is a pause with better branding.

AI was supposed to free up capacity. In a lot of companies it added a new category of work that nobody wanted to be the person to question.

Domino Data Lab's annual enterprise AI survey, released in July, found that 57 percent of enterprises still see AI returns that fail to outpace what they spend. That share has not moved since 2025, even though more than nine in ten respondents reported better production capability. Better tools. Same returns. Infosys found that close to half of large companies have no central way to track AI value at all. Gartner's name for what accumulates is "zombie AI," projects that keep consuming people and money without returning anything.

In a smaller company the damage is more specific. Every AI pilot draws on the same handful of people who carry the core product. An agent built because the board asked about agents is not free. It is paid for with the roadmap.

Most AI plans in the middle market are not strategy. They are anxiety with a budget. If nobody can name what a pilot is supposed to move, the company is not experimenting. It is decorating.

AI is a method, not a value driver. So is modernization, and so is a new ERP. The driver is the commercial or operating outcome the method is supposed to move. If a plan cannot name that outcome, it cannot tell you when to stop.

None of this requires new software. It requires four rules that leadership actually enforces, and a CEO willing to be the person who says no in public. Saying yes is free in a planning meeting. It is expensive in every meeting after.

One in, one out. Every new initiative names the work it replaces, in hours rather than slogans. If nobody can name the stop, the start waits.

Kill criteria at birth. Every initiative states, on the day it is funded, the result that would end it. Written before anyone is attached, that is a business rule. Written later, it is a political fight.

Stops get priced. Stopping work has a cost. Customer migrations, contract commitments, a pay plan change, sometimes a valued person who needs a new role. Budget for the exit of the work the same way you budget for its launch. A stop without a budget is a pause.

Stops get published. If a stop is never announced, the work does not stop. It goes underground and keeps eating capacity under a different name.

There is one more truth worth saying plainly. The stop list is almost impossible to write from the inside. Every person in the room sponsored something on it, hired someone for it, or promised it to a customer. That is not a character flaw. It is a conflict of interest built into every leadership team.

That is the real case for an outside operator. Not because they are smarter than your team. Because they did not launch anything.

Your team cannot write the stop list. Every one of them launched something on it.

The test also forces better language. Watch what happens to familiar line items when they have to name the outcome.

For PE-backed companies this is not an operating nicety. Bain's latest global report counts roughly 32,000 unsold portfolio companies worth about $3.8 trillion, with average holding periods near seven years. More assets are heading to market in front of more selective buyers. Simon-Kucher's read from this year's operating partner gatherings is that value creation is moving from a collection of initiatives to an operating system.

Technology diligence reads the roadmap, the backlog, and the org chart. Forty open initiatives and six overlapping AI tools tell a buyer that management cannot choose, and the buyer prices that risk.

A dated record of what was stopped, why, and what the freed capacity produced tells the opposite story. It is evidence of the one thing a buyer cannot model from the data room. Whether this team can make hard calls without the sponsor in the room.

Diligence teams have seen the ceasefire plan many times. They recognize it the moment they open the backlog.

Buyers do not pay for everything you launched. They pay for proof that management can choose.

1. Pull last quarter's hours. Lay engineering and product time against the 2027 plan and look at the gap before anyone debates priorities.
2. Make every start name its stop. No new initiative gets funded without the work it replaces and the result that would end it.
3. Read the pay plan last. If compensation still rewards work you intend to stop, fix the plan before you approve the roadmap.

The leadership teams that win the next two years will not be the busiest ones. They will be the ones that decided, in writing, what they were not going to do.

Someone is already writing your 2027 stop list. It should be you.