That figure is not a saving. It is unmanaged capacity: time your organization has already purchased and is not directing. It becomes a return only when leadership decides what the freed hours are now for — which is the same decision as retrain or replace.
AI WORKFORCE DECISIONS · EXECUTIVE ADVISORY
Retrain or replace?
Every executive is being asked this question by their own numbers. Almost none of them can defend their answer.
The question on the table
In MIT CISR research, executives estimated that 38% of their workforce will need fundamental retraining or replacement within three years. McKinsey found 62% of executives expect to retrain or replace more than a quarter of their people. The question is already in the room. What is missing is an answer leadership can put in writing, defend to a board, and survive if someone contests it.
What this is not
- Not a tool selection exercise
- Not generic change management
- Not training delivered and forgotten
- Not a policy nobody owns
What this is
- A role-by-role workforce decision, recorded
- Named owners for every decision taken
- Evidence behind each decision, kept
- A position the board can read and defend
The silent invoice
The hours came back. The direction did not.
BCG's 2026 AI at Work study found that 42% of employees using AI save roughly eight hours a week — and that 66% received no instruction at all on what to do with that time. Put your own numbers in and see what you are already paying for.
The 70% nobody budgeted for
Only 10% of the value is in the algorithm.
BCG's analysis of AI transformations splits the value three ways: 10% algorithms, 20% technology, 70% people, processes and the operating model. The budget almost always does the opposite.
That is why MIT's 2025 review found 95% of enterprise AI programs returning nothing measurable, and why McKinsey's 2026 State of AI found only 6% of companies reaching a 5% EBIT impact. The models work. The organizations around them were never redesigned, and nobody was made accountable for the people side of the decision.
Meanwhile leadership is deciding on bad information: 77% of executives believe their managers are ready to lead through AI adoption; only 34% of managers and 9% of employees agree. The gap between those numbers is where retrain-or-replace decisions are actually being made today — informally, undocumented, by whoever is closest to the work.
AI workforce planning
Workforce planning used to be a forecast. Now it is a decision.
Headcount models assumed the work stayed the same and only the volume changed. AI changes the work itself, task by task, inside roles that still carry the same title.
That is why AI workforce planning cannot be delegated to a spreadsheet or to a reskilling vendor. Deciding who is retrained, who is reassigned and who is replaced is an executive decision with legal, accreditation and board consequences, and it has to be made on evidence from the actual work — then written down in a form that survives being questioned a year later.
Training for employees is part of the answer, never the whole answer: training without a decision about the role is a receipt, not a plan.
The mechanism
The Retrain-or-Replace Decision Map
Produced during the 70% Audit. Tasks first, people second, evidence always.
01 · Task inventory
What the work actually is
Each role is broken into its tasks and each task is marked: performed end to end by AI, assisted by AI, or still requiring human judgment and accountability. This is done on evidence from the work, not on job descriptions.
Output
A task-level map of where the work has already moved.
02 · The decision
Retrain, reassign, or replace
Role by role, with the reasoning written down: who is retrained into judgment-heavy work, who is reassigned to where the capacity is needed, and where the task itself no longer exists. Legal exposure, documentation duties and accreditation obligations are weighed in the same conversation, not after it.
Output
A decision per role, with a named owner.
03 · The record
What survives the challenge
Every decision is recorded with the evidence that supports it, the alternatives considered and the authority that approved it — the file you need if an employee, a regulator, an accreditor or a board member asks why this role and not that one.
Output
An evidence record and a 90-day sequence of execution.
Why this desk
Four chairs, one decision.
Technology vendors cannot answer this question: they have no authority over people. Change-management consultants cannot answer it: they do not read legal exposure. Employment lawyers cannot answer it: they have not run an AI operation. HR cannot answer it alone: the decision sits above them.
I spent twenty years deciding who stays, who moves and who goes — in hospitals, in a school of medicine, in federal court, as an attorney, as an HR director and as an associate dean. Now organizations ask me the same question about AI. The difference is that these decisions are now made faster, at larger scale, and with a machine in the chain of reasoning — which is exactly when the record matters most.
How organizations engage
Start with one decision, not a program.
Four ways in. Most begin with the 90-minute session and decide from there.
Executive AI Governance Session — $1,500. Ninety minutes with leadership, a written decision memo in two business days. If the session does not produce a decision in writing, the fee is credited to the next engagement or returned.
The 70% Audit — $3,500 added to the AI Readiness & Governance Assessment. The full Decision Map, role by role, with the evidence record.
Manager Readiness Program — $25,000 to $45,000. For 20 to 40 managers: closing the gap between what the CEO believes and what the manager can actually carry.
Fractional Chief AI & People Officer — $10,000 to $25,000 per month, six to twelve months. Someone accountable in the room for AI decisions about people. Two retainer seats per quarter, because each one requires presence in real committees.
Not ready to book? Take the AI governance policy template — free, and it shows you which decisions are missing.
Start the decision
Bring me the roles you cannot decide.
Tell me where the question is stuck — a department, a role, a board that keeps deferring. You get a written reply with how the decision would be framed, whether or not you engage.
Include your email, organization and reason for contacting us. You may answer in English or Spanish. Wait for ClickUp’s confirmation after submitting.
Questions leaders ask
What is AI workforce planning?
AI workforce planning is deciding, role by role, how work is redistributed once AI performs part of it: which roles are retrained into judgment-heavy work, which are reassigned where capacity is needed, and which no longer exist in their current form. Done properly it produces a documented decision with a named owner for each role, not a headcount forecast.
What does 'retrain or replace' mean for an executive team?
It is the workforce decision AI forces on leadership: for each role, whether the person is retrained into higher-value work, reassigned, or replaced because the task itself no longer exists. It is a people decision with legal, documentation and board consequences, not an IT decision.
How do we decide which employees to retrain instead of replace?
Start from tasks, not people. Break each role into its tasks, mark which tasks AI now performs end to end, which it assists, and which still require human judgment or accountability. Roles where most remaining tasks are judgment-heavy are retraining candidates; roles whose tasks have been fully absorbed are redeployment or replacement decisions. Every one of those decisions must be recorded with the evidence behind it.
Why do most AI implementations fail to produce returns?
Because the work around the tool never changed. BCG puts only 10% of the value of an AI transformation in the algorithms and 20% in technology; 70% comes from people, processes and the operating model. MIT's 2025 review found 95% of enterprise AI programs returned nothing measurable, and McKinsey's 2026 State of AI found only 6% of companies achieved at least a 5% EBIT impact.
What is The Retrain-or-Replace Decision Map?
A role-by-role record produced during the 70% Audit: which tasks are automated, which are elevated, who is retrained, who is reassigned, who is replaced, who owns each decision, and the evidence that supports it if the decision is later contested.
What does this cost and how do we start?
Most organizations start with the $1,500 Executive AI Governance Session: 90 minutes with leadership and a written decision memo two business days later. The 70% Audit is a $3,500 module on the AI Readiness & Governance Assessment. Ongoing accountability is available as a fractional engagement, limited to two retainer seats per quarter.
Who makes these decisions defensible?
The person signing them. Workforce decisions driven by AI touch discrimination exposure, documentation, accreditation and board duty. SAGAZ is led by an attorney with twenty years of practice who has also served as HR director and associate dean — the same decision, from all four chairs.
The decision is being made either way.
The only question is whether it is made deliberately, by the people with the authority, on a record you can defend.
Talk about your workforce decisionEmail about this