Frameworks

Costs climb faster than your prices. Define what AI has to win before you pilot it.

Chris McCarthyNovember 1, 20267 min read
ai for manufacturing marginsai roi manufacturinghow to measure ai roiai pilot success metricsmid-market manufacturing ai

Costs climb faster than your prices. Define what AI has to win before you pilot it.

By Chris McCarthy, Founder & Chief AI Officer, OB.1

When input costs rise faster than you can raise prices, every new spend has to show where the margin comes back. AI is no exception. The first question to settle is plain: what counts as success, and who owns it? Answer that in dollars, with a name next to it, before any pilot starts.

This is part one of our Pilot to Production series. Each part takes one problem we hear from mid-market manufacturing leaders and pairs it with the question that moves AI forward. This one is for owners, presidents and CEOs.

The squeeze is real, and it lands on you

The numbers back up what you feel. In the National Association of Manufacturers' second-quarter 2026 outlook survey, manufacturers expected raw material and other input costs to rise 5.8% over the next 12 months. They expected prices on their own products to rise 4.2%. For small and medium-sized firms, the expected figures were 6.3% for raw materials and 4.8% for prices, a gap of about a point and a half by our own math. Raw material costs were a top challenge for 83.1% of respondents. Closer to home, the Cleveland Fed's October 2026 Beige Book reported that nonlabor cost pressures remained robust for an eleventh straight period in the Fourth District.

You feel this in the monthly numbers before anyone names it. Material costs go up. Freight goes up. Your largest customers push back on every price letter. The gap between what you pay and what you can charge comes straight out of EBITDA.

That pressure changes how AI gets judged in your building. A pilot that saves "some time" does not hold up in a margin conversation. A pilot that recovers a named dollar amount, owned by a named leader, does.

Why AI pilots stall without a success definition

Most AI pilots start with a tool demo. Someone sees a capability, the team gets excited, and a test begins. Three months later, the question in the leadership meeting is whether it worked. Nobody wrote down what "worked" meant, so the answer is a debate.

That debate is the real cost. The pilot may have helped. Without a baseline and a target, nobody can prove it, and it never earns the budget to scale.

The firms that get AI right tend to grow. The National Center for the Middle Market reported in April 2026 that middle-market firms using AI saw average year-over-year growth of 12.9%, more than double the 5.8% reported by firms without AI. In its mid-year 2026 indicator, nearly two-thirds (66%) of firms reported positive returns on AI investments. The gap shows up where measurement is weak. In RSM's 2026 middle market AI survey, unclear ROI was a reason cited by 30% of manufacturers whose pilots saw only moderate or limited success.

The question to ask: what counts as success, and who owns it?

This question has two halves, and both matter.

What counts as success means a number tied to margin. Pick one or two. Good candidates in a mid-market plant:

  • Hours spent building quotes, per quote, before and after
  • Scrap rate on one line or one product family
  • Order entry errors that lead to credits or rework
  • Days to close the month
  • Expedited freight spend caused by planning misses

Write down today's value. Write down the target. Write down the dollar value of closing that gap over twelve months.

Who owns it means one leader whose name sits next to the number. That person decides when the pilot is working, when it needs to change, and when it should stop. Ownership by committee is the same as no owner.

What a good answer looks like

Here is the test we use. A strong success definition fits on one page and answers five things:

ItemExample
The processQuoting for custom fabricated parts
The baseline6 hours of estimator time per quote today
The target3 hours per quote within 90 days
The dollar valueEstimator hours freed, priced at loaded labor cost, plus win rate on faster quotes
The ownerVP of Sales, with the lead estimator as day-to-day contact

The numbers above are an illustration. Yours come from your own data. If you cannot fill in the baseline, that is your first finding: the process needs measuring before it needs AI.

How OB.1 fixes this

We start every engagement with an Operational Blueprint. It is a two-week look at how your work runs today, done with the people who run it. For a margin problem, the Blueprint does three things.

First, we map the process where the margin leaks. We talk to the people closest to the work and trace each handoff. That shows where hours, scrap or errors pile up.

Second, we set the baseline and the target with you. Each opportunity gets a dollar estimate, an owner, and a ranking. You see which fixes can recover margin first.

Third, we hand you a roadmap your leadership team can act on. Vetted build partners can then build to that spec in the Architect phase. OB.1 stays neutral on tools because we do not sell build hours.

The point is simple. AI earns its place when it moves a number you already care about.

Read the rest of the series

  1. Costs climb faster than your prices (this article)
  2. Your ERP data sits in silos nobody trusts
  3. The plant you just bought runs a different system
  4. Staff paste pricing into public AI tools
  5. Your best tech retires, and the fixes leave too
  6. Automation could ship the wrong order
  7. Teams lean on the system and lose the judgment to step in

Book a Blueprint Session

Sources

Frequently Asked Questions

How do I measure AI ROI in a manufacturing company?

Pick one process and one or two numbers tied to margin, such as quote hours, scrap rate or order errors. Record today's baseline, set a target, and price the gap over twelve months. Then compare results to that baseline.

Who should own an AI pilot in a mid-sized manufacturer?

One senior leader who owns the process being changed, such as the VP of Operations for a scrap problem or the VP of Sales for quoting. That person decides whether the pilot is working.

What is a good first AI pilot for margin improvement?

A process with a clear baseline, a named owner and a direct link to cost or revenue. Quoting, order entry and scrap tracking are common starting points because the numbers already exist.

What does an Operational Blueprint include?

A two-week review of how one area of your business runs today. It delivers a baseline, input from the people doing the work, and a ranked roadmap with owners and dollar estimates.

Get Started

Book a Blueprint Session.

That is the first conversation.

chris@ob1ai.co | (234) 602-0500

Talk to OB.1

Tell us where AI fits in your business. We start with the Operational Blueprint and Diagnostic AI Consulting, and we reply within one business day.