I saw versions of the same slide over and over during my years in investment banking IT. It had one number on it, and the room loved it every time.
A new or newly promoted CIO would announce a plan to cut technology costs by moving development and support to lower-cost locations, often India, sometimes a cheaper onshore city like Houston. The operating expense line dropped sharply and the leader was credited with the savings. More than once, the move because the launchpad for a bigger global role.
18 to 24 months later, delivery had slowed and quality had slipped. Firms hired project managers and business analysts just to translate between the business and teams thousands of miles away, and costs crept back up. By then the original decision had already been scored as a success, and nobody connected the two.
It happened often enough, under enough different leaders, that I came to see it as something the system was producing. Today I’m watching organizations pick up the same playbook with AI written on the cover.
Why does the pattern keep repeating?
The leaders running this play were responding rationally to how the system scored the decision. Savings were counted in the year they were announced. The cost of separating the work from the people who understood it arrived two years later, spread across budget lines, and was never traced back. When an organization rewards the number at announcement and never audits the outcome, swapping the person in the seat leaves the incentive exactly where it was.
“It happened often enough, under enough different leaders, that I came to see it as something the system was producing.”
What does the AI version look like?
A leadership team engages a large consulting firm to supply an AI playbook, then reduces headcount in anticipation of the efficiency AI agents are expected to deliver. The problem is that the cuts are made on a promise that’s running well ahead of the evidence. Nobody, including the firms selling the playbook, has a long track record of running AI agents at enterprise scale inside organizations like yours.
The data is starting to catch up. In Orgvue’s 2025 survey of more than 1,000 senior leaders, 39% had made redundancies because of AI, and 55% of those now say it was the wrong decision. Forrester’s Predictions 2026 report forecasts that half of AI-attributed layoffs will be quietly reversed, with the work coming back offshore or at lower salaries. The AI story loops straight back into the offshoring one.
“39% had made redundancies because of AI, and 55% of those now say it was the wrong decision.” - Orgvue, 2025 survey of 1,000+ senior leaders
What leaves when the people leave?
In the offshore era, the most valuable thing a development team had was proximity. Developers who sat near the trading desk knew which exceptions mattered and who to call when a number looked wrong, and nine of it was written down.
AI agents carry the same dependency. An agent can only work with the context it’s given, and the person who knows why month-end still has a manual step is that context. Remove the person before the agent has learned the work, and you’ll pay someone to reconstruct what they knew. That’s the business analyst bridge all over again.
Why doesn’t anyone in the room say so?
Every organization running this play has people who can see where it’s heading, and most of them stay quiet for two reasons that reinforce each other.
The first is biological. Questioning a high-profile initiative risks being placed outside the group, and the labels arrive quickly: not on board, or worse, change resistant. Being pushed tot he edge of a group registers in the brain in a way that overlaps with physical pain. In a 2003 study published in Science, Naomi Eisenberger and colleagues found that being left out of a simple online ball-tossing increased activity in the dorsal anterior cingulate cortex, a region also involved in the distress of physical pain. Researchers call this social pain. For a sponsor facing a confident CIO and a polished deck full of jargon, asking “how do we know this will work?” feels like risking a seat at the table.
“Being pushed to the edge of a group registers in the brain a way that overlaps with physical pain.”
The second is trust. Many people have already concluded their opinion isn’t wanted, because they’ve seen the last concern raised in a steering meeting thanked politely and then ignored. Once people stop believing the room will listen, silence becomes the sensible choice, and from the top that quiet looks a lot like alignment. Often the person staying quiet is the one who understands the work best.
What should leaders do differently?
None of this is an argument against AI. It’s an argument against removing the knowledge AI depends on before you know what it needs.
Define the outcomes before you count the savings. Agree on what the business needs to be true in two years: delivery speed and quality, tracked through cycle time and rework, along with whether sponsors are customers are getting what they need. If those slip whole costs fall, the savings have been borrowed from the future.
Sequence the reduction behind the evidence. Run the agent alongside the people doing the work today, and reduce capacity only once it has handled real volume, exceptions included.
Score the decision at month twenty-four. Build a look-back against those outcomes into the business case, with a named owner who stays accountable wherever their career goes next.
Make the hard question safe to ask. Separate “how do we know this will work?” from “are you n board?” and ask any outside advisor where they’ve delivered this at your scale. When a concern changed the plan, say so openly and credit the person who raised it. That is how trust gets rebuilt.
If you lead at the executive level: Look at the last major cost decision your organization celebrated. Who is tracking whether it delivered, two years on? If the answer is nobody, your AI business case is about to be scored the same way.
Strategic Change Execution focuses on executing change where complexity and pressure collide, partnering with leaders to deliver outcomes without weakening the organization in the process.
If you’re weighing a workforce decision tied to AI, Newgrange starts with a situational assessment of where the knowledge and capacity in your organization really sit, then builds a customized capability development plan so AI strengthens what you have instead of replacing what you’ll need. Let’s talk before the cost-cutting slide gets made.