BeforeScaling — Enable your workforce before you scale AI

of enterprise GenAI pilots deliver no measurable P&L impact

MIT · 2025

0%+

of agentic AI projects will be canceled by end of 2027

Gartner · 2025

of employees already bring their own AI to work — unmanaged

Microsoft & LinkedIn · 31,000 workers

of AI transformation success is people and process — not technology

BCG

The losing playbook

What companies get wrong about AI

Every failed rollout runs on at least one of these. If any sound like your plan, you're about to buy shelf-ware.

“We'll replace people with agents.”

Agents don't know your business. The people who do are the only ones who can apply AI to it. Firing the knowledge and keeping the tool is backwards — and it's why 40% of agent projects get canceled.

“We bought Copilot. We have a strategy.”

A license is not a capability. Most employees with AI access stall at “faster Google” — and the investment shows up nowhere.

“AI is an IT project.”

IT can deploy a model. IT cannot make your controller or project manager work differently. That's a management problem, not a technical one.

“We banned it until it's safe.”

78% of your people already use AI on their own. Banning it doesn't stop usage — it stops governance of usage. The worst of both worlds.

“Automate the process as-is.”

Automating a bad process gives you the same bad output, faster. Someone must redesign the work first — and that someone has to understand it.

“We tried it in January. It wasn't good.”

The tools improve monthly. A verdict from six months ago describes a product that no longer exists.

Exhibit 1 — Where AI success actually comes from

Algorithms & models Technology & data People & processes

Source: BCG's 10-20-70 rule for AI transformation.

Most budgets invert this — and fail in proportion.

Companies spend 70% of attention on models and licenses, and treat people and process as a footnote. BCG's research says the ratio runs the other way: the spend that pays is training, workflow redesign, and adoption. That 70% is our entire business.

The asset was never the model

You can't fire your way to AI.

Models are rented, identical for you and your competitor, and replaced every quarter. What compounds is your people's ability to apply them to your business — judgment, domain knowledge, and customer history no vendor ships. In the largest field experiment to date, the bottom half of performers gained the most. AI lifts the middle of your organization — precisely the people the fire-and-replace plan discards.

The talent math

You can't hire this quadrant. You build it.

With today's technology, enabling your experienced people is cheaper and more impactful than hiring or replacing them — and the agents everyone wants to buy can't be built right without the knowledge that lives in those people's heads. Enabling them isn't one option among several. It's the only move left.

Exhibit 2 — Where the leverage lives

High Knows your business Low

Your veterans

Deep business knowledge, no AI. The know-how lives in binders and habits — and retires with them.

Your enabled workforce

Business knowledge × AI capability. The only quadrant where AI compounds — and it's built, not hired.

New to both

Long ramp, high churn risk. Transformation doesn't start here.

AI-native hires

Fluent with the tools, blind to your business. Output without judgment — polished and wrong.

Low Can work with AI High

The research closes every other door. Replacing is expensive: it costs one-half to two times the employee's annual salary (Gallup), and external hires cost 18–20% more while underperforming for two years (Wharton). Buying AI talent is more expensive every year: the AI-skills wage premium hit 56%, doubling in twelve months (PwC, 2025). And tools without your people's knowledge fail: 95% of AI pilots deliver no P&L impact precisely because they never absorb how the work actually gets done (MIT, 2025). That's why 85% of employers now prioritize upskilling their existing workforce (World Economic Forum, 2025).

The other side

The winners are measured, not imagined

About 6% of companies — McKinsey calls them AI high performers — already attribute 5% or more of their EBIT to AI. What separates them isn't the model they bought. It's how far up their people operate.

Exhibit 3 — The higher your people operate, the more likely the payoff

Deploy the basics — data, tech, talent (operational)20%

Build solutions the business actually uses (managerial)39%

Learn with AI — humans and AI iterating together (strategic)73%

Probability of significant financial benefit from AI. Source: MIT Sloan Management Review × BCG, global survey of 3,000+ companies.

Learning with AI nearly quadruples the odds.

Deploying tools gives you a one-in-five chance of real financial impact. Organizations where humans and AI learn together reach 73% — and those with multiple ways for humans and AI to learn from each other are 6× more likely to amplify their success. The climb isn't optional. It's where the money is.

The framework

AI is not the same thing for every employee

That climb has a map. There are three levels of working with AI — and every role peaks at a different one. One expectation for everyone is how rollouts die; part of our job is telling you which is which, honestly.

Level 1 · OperationalAutomationAI moves the data: gathering, recording, integrating. Manual labor gets faster. Real — but bounded.
Linear
Level 2 · ManagerialAugmentationAI works on the content: drafting, summarizing, transforming, synthesizing. Knowledge work gets faster — but the thinking hasn't changed hands.
Linear
Level 3 · StrategicCollaborationAI becomes a thinking partner — bent by what's unique about the person using it: taste, judgment, domain expertise. Only reachable with levels 1 and 2 under your feet.
Exponential

Exhibit 4 — AI pays inside its frontier, and bites outside it

Speed — tasks inside AI's frontier+25%

Quality — tasks inside AI's frontier+40%

Correct answers — tasks outside the frontier, AI users vs. no AI−19 pts

Field experiment, 758 working consultants. Harvard Business School × BCG, Navigating the Jagged Technological Frontier.

Knowing where the frontier runs is a skill.

On tasks AI can't do well, consultants who used it were 19 percentage points less likely to get the right answer than colleagues working without it — because the wrong answer comes out looking polished. Frontier Recognition is one of the six skills every participant trains. And the bottom half of performers gained the most: AI lifts the middle of your organization, not just the stars.

The thesis

Capability is built in one order. There is no shortcut.

You can't scale AI over people who've never gotten results from it. Built in this order, company systems automate workflows that already work. Built in any other order, they join the 95%.

Enable

Each individual builds real results in their own work — new practices, habits, and automations — until the capability reaches critical mass and they keep going by themselves.

Individual · to critical mass

Redesign

Teams redesign their shared practices and methodologies deliberately. This is the step research ties most directly to bottom-line impact — and the one 80% of companies skip.

Team · the step that pays

Scale

Your IT builds enterprise-level tools on top of workflows that already work — with governance fed by real data, not guesses. This step is yours; our job is getting you ready for it.

Company · with IT, not instead of it

Exhibit 5 — The step everyone skips is the one that pays

Layer AI on top of existing processes~80%

Have redesigned even one workflow21%

AI high performers who fundamentally redesign55%

Source: McKinsey, The State of AI, 2025.

Redesign is the #1 bottom-line driver.

McKinsey finds fundamental workflow redesign is the organizational change most correlated with EBIT impact from AI — and four in five companies never do it. It's step two of the sequence, and it only works on top of step one.

The offer

The Enablement Program

Six weeks. One cohort of up to ten people from your company. Three rules: real workflows — never exercises; one thing at a time; and proof you can see.

Working session: mapping a real workflow on the board

Week 1Leader kickoff

Expectations per role, and the one commitment that makes it work: protected hours.

Weekly90-min 1:1 session

Each person rebuilds a real workflow from their own job on AI — learning by doing, not a demo.

WeeklyLeader report

Who's progressing, what's built, where the blockers are. Every week.

BetweenProtected practice

Capability is built between sessions. A few protected hours per person, contractual.

Week 6Demo day

Every participant runs their workflow live. No slides about potential. Working things.

Real workflowsThe weekly report, the proposal, the client deck — the actual work, running with AI.

One thing at a timeNobody advances until the current stage runs. Skipped stages are how capable people end up with nothing working.

Owned context, rented brainsEverything is built in files your company owns — portable across vendors. No tool lock-in, by design.

An honest promiseEvery participant ships at least one real workflow — worth ≥10% of their time back, compounding. We don't promise 10×. The 95% who failed were promised 10×.

What you get

An X-ray of your organization no audit produces

This is not a report on your workforce. It's your organization with its processes documented — knowledge that today exists only in your people's heads. That's the indispensable raw material for agent workflows at the enterprise level: prioritize the right projects, surface the real areas of improvement, and de-risk the day an experienced person walks out the door.

How work actually gets done

Real documentation of the workflows your people follow to hit their objectives — closing the gap between your SOPs and reality.

What tools are really in use

Which tools, by whom, and how — including the AI your people already brought to work without telling anyone.

Your future-project pipeline

The needs, mapped per team — the concrete input for governance, observability, and enterprise-level implementation.

Who's who with AI

Which of your people operate at automation, augmentation, or collaboration — so you know exactly what you can build on.

And it doesn't stop when the program ends. The six weeks give every employee a built-in incentive to keep documenting their processes — that documentation is the context their own agents run on — and a new habit: how the work is approached gets decided before the work even starts.

This is for you if

A fit

You run a company of roughly 30 to 250 people.
You already know AI matters — you don't need convincing, you need a way to start that works.
You're willing to give your people a few protected hours a week to build the capability.

Not a fit

You want someone to build agents while your team keeps working the old way. That's the expensive mistake we exist to prevent.
You're looking for a tool recommendation. The tool was never the problem.

Who's behind this

Built the hard way — running a real company on it

Christian Trujillo built and ran his own company on exactly this approach — AI workflows running finance, billing, operations, and reporting, built person by person, before anyone called it enablement. BeforeScaling exists because the method that worked there works anywhere knowledge workers do repeatable work: meet each person where they are, build with their real workflows, and prove it with working artifacts — not slideware.

He's the consultant who will tell you “not yet” when your team isn't ready to scale — which is exactly why “now” means something when he says it.

Before you scale AI into your company,
can your people actually use it?