Skip to content
Why 95% of GenAI Pilots Produce No P&L Impact (MIT 2025)

Why 95% of GenAI Pilots Produce No P&L Impact (MIT 2025)

Yaroslav Maxymovych· with AI assistance11/9/202676 views6 min read

TL;DR

  • •MIT 2025 research found 95% of GenAI pilots produce no measurable P&L impact.
  • •Most pilots fail because they start with a tool, not a workflow decision.
  • •P&L impact requires visibility into current operations before automation begins.

After watching dozens of founders launch AI pilots that never touched the bottom line, I keep seeing the same pattern: the technology works in a sandbox, but the business never changes.

Definition: GenAI pilot

A time-boxed experiment that tests whether generative AI can perform a specific business task, usually with a small group and limited scope.

Definition: P&L impact

A measurable change in profit and loss—revenue increase, cost reduction, or margin improvement—directly attributable to the AI intervention.

Definition: Pilot purgatory

The state where AI experiments run indefinitely without ever being scaled into production or killed.

Why MIT Says 95% of GenAI Pilots Miss P&L

MIT's 2025 research on enterprise GenAI adoption found that the vast majority of pilots never translate into financial results. The issue is rarely the model. It is the context around the model.

Most companies launch pilots to "see what AI can do." That is a technology question, not a business question. Without a pre-defined financial target, a pilot can look successful—employees use it, managers are impressed—and still deliver zero dollars to the P&L.

The pilots that do move numbers share one trait: they start with visibility into what the team actually does today.

What to Look for Before You Approve a Pilot

1. Does the pilot target a specific cost or revenue line?

A pilot must be tied to a line item on the P&L before it starts. If you cannot name which cost center or revenue stream the AI is meant to affect, you do not have a pilot. You have a hobby.

2. Do you have a baseline for the current process?

You cannot claim savings if you never measured the status quo. How many hours does the team spend on this task weekly? What is the error rate? What is the cost of delay? Without these numbers, post-pilot "improvements" are unverifiable.

3. Who owns the outcome after the pilot ends?

Pilots die when no one is accountable for integrating the result into daily operations. The owner must name a single person responsible for moving the pilot to production—or killing it.

The Baseline Problem: You Can't Optimize What You Can't See

Before any AI tool is selected, the founder needs to see the current state of operations. This means mapping the workflow, measuring time spent, and identifying where value is actually created versus where it leaks.

This is why our first step with companies is a visibility diagnostic: we look at what the team does day to day, not what AI vendors promise they could do. Only then can a pilot be scoped against a real baseline.

Tool tip (AiAdvisoryBoard.me): Before commissioning a GenAI pilot, run a Plan → Fact → Gap review of the target workflow. Compare what leadership believes the team does (Plan) with what actually happens (Fact). The Gap is where your first AI candidate lives—or where you realize automation is premature. See how the 7-day diagnostic works.

From Pilot to P&L: The 90-Day Filter

Most companies get stuck in pilot purgatory because they lack decision points. We use three gates:

Day 30: Scope decision. Is the workflow clearly defined? Is the baseline measured? If not, kill or narrow.

Day 60: Controlled pilot. Does the AI output meet quality standards with human review? Is time or cost actually decreasing? If not, kill.

Day 90: Scale or kill. Can this be handed to the team to run without the vendor or consultant? If not, kill.

Pilots that survive all three gates are rare. That is the point. The filter exists to protect the P&L from theater.

Manager Scan: What a Pilot Review Looks Like

Here is how a founder should review an AI pilot after 30 days, using plain operational language:

  • Pilot name and owner: who is accountable?
  • Target workflow: which process is being changed?
  • Baseline: hours per week, cost per transaction, error rate before AI.
  • Fact: what actually happened in the last 30 days?
  • Gap: difference between expected and actual results.
  • Decision: scale, adjust, or kill.

If the founder cannot fill in the baseline and fact rows, the pilot is not ready for a second month.

Micro-Case: What Changes After 7–14 Days

A 120-person professional services firm launched a GenAI pilot for client reporting. After two weeks, the team was using the tool, but the partner could not tell whether report production had sped up or slowed down. There was no baseline.

The founder ran a visibility diagnostic first. They mapped the current reporting workflow, measured time spent by role, and identified the actual bottleneck: data gathering, not writing. The AI pilot was redirected to the gathering step. Within 14 days, the team had a measurable baseline and a clear go/no-go criterion for month two.

Note on this case: This example is illustrative — based on typical patterns we observe with companies of 30–500 employees, not a single named client. Specific numbers are rounded approximations of common ranges, not guarantees.

FAQ

Does MIT really say 95% of GenAI pilots fail? The MIT 2025 research cited here indicates that 95% of GenAI pilots produce no measurable P&L impact. This does not mean the technology failed; it means the pilots were not connected to financial outcomes.

How is this different from a normal software pilot? GenAI pilots often lack deterministic outputs. You cannot test them like a calculator. They require behavioral change from employees, which most pilots do not plan for.

Should we kill all pilots and wait? No. Kill pilots that lack baselines, owners, or P&L targets. Keep pilots that start with a specific workflow and a defined financial hypothesis.

Who should own the pilot inside a 100-person company? One operational leader—usually a COO, head of operations, or department head—with the founder setting the financial target and reviewing progress weekly.

What if we already have five pilots running? Audit them against the three gates above. Keep the ones with clear baselines and owners. Pause or kill the rest before they consume more budget.

Conclusion

MIT's 2025 finding is a warning, not a sentence. Most GenAI pilots produce no P&L impact because they start with the tool instead of the workflow, and because no one measures the baseline before automating.

Before approving your next AI experiment, map what the team actually does today. That visibility is the cheapest pre-AI investment you will make.

If you want a system that surfaces the Plan → Fact → Gap automatically—every day, across the company—see how the 7-day diagnostic works. https://aiadvisoryboard.me/?lang=en

Frequently Asked Questions

Read with AI

Open this article in your assistant — it will summarize it and help apply it to your company.

Show the prompt

Read the article https://aiadvisoryboard.me/blog/why-genai-pilots-no-pl-impact-mit-2025.md and summarize the key points. Then ask me about my company (industry, team size, what takes the most time) and explain which ideas from the article apply to us and where to start.

More on this topic
90-Day AI Implementation Roadmap: The SMB Guide to Adoption →

The pillar guide for "AI Implementation Roadmap" linking every article in this cluster.

Yaroslav Maxymovych
Author
Yaroslav Maxymovych
Founder & CEO, AI Advisory Board

Implements AI agents in companies and teaches founders and their teams to work with them — through courses and corporate programs.

This article was prepared with AI assistance, based on Yaroslav Maxymovych's methodology and materials. Spotted an inaccuracy — let us know via the form below.

For companies

Your company's first 3 AI automations — in 2 weeks

A corporate AI-transition program: 4 live sessions with your team plus a video course for every employee. Up to 20 people for one fixed price. If it doesn't work — money back.

Working automations in 2 weeks
Up to 20 employees, one price
Money-back guarantee
See the program & priceIt's the program page, not a checkout — a 2-minute read
Newsletter

AI adoption resources — one email

Leave your email to receive a welcome email with blog resources. There is no regular newsletter at present

No spam. Unsubscribe anytime.