A case study without a before number is a testimonial

Most AI case studies are testimonials with a logo on them.

They say some version of "saved time," "the team loves it," or "changed how the team works." A buyer reads that and learns nothing they can repeat to their boss. There is no starting point, so there is no change, so there is no proof.

No before number, no case study. Just a testimonial with a logo on it.

Why do numbers with a before and after work?

Because two numbers and a direction are hard to argue with.

The line I keep coming back to on LinkedIn is the Monday.com click-through rate dropping from 2.94% to 0.84%. Nobody argues with it. It has a named company, a before, an after, and a direction. That is the whole structure of a claim that sticks.

Your case study needs the same shape. Before: how long the job took, how much it cost, how many people did it. After: the same measure, same units, a set amount of time later. The gap is the story.

Why is this harder for AI products?

Because buyers have already been burned by AI claims.

Enterprise buyers have watched every vendor claim state of the art. That is why private, task-based evals are getting funded. I wrote about it in private AI benchmarks as proof, not slogans. The buyer question moved from "who won the chart" to "does it survive my workflow."

A case study is your version of that eval, run by a real customer. If it names the job, the before, and the after, it answers the question the buyer is actually asking. If it says "10x productivity," it is one more slogan they already tuned out.

What should an AI case study include?

Five things. Most skip the first two.

1. The job. One specific task. "Researching accounts before outbound." Not "sales productivity."
2. The before number. Hours per week, cost per unit, error rate, or volume. Measured before the product went in, or reconstructed honestly from their records.
3. The after number. Same unit, same measure, with the time window stated. Thirty days is fine. Say it.
4. What the humans still do. Who reviews the output, where the product stops. Buyers trust a case study more when it admits limits.
5. A customer who signs off. A named person who approved the numbers. If they will not put their name on it, the number is not ready.

That list also turns the case study into a proof asset before the demo. Short, specific, and mapped to a job the buyer recognizes.

Why is this a Grow problem?

Because proof is what makes paid and word of mouth compound.

Company-level numbers are investor proof. TechCrunch reported Ema's Series B with more than 50 enterprise deals, over 1 million active users, and about 180% net dollar retention. Those numbers help raise a round. A buyer running a five-person team cannot map themselves onto 1 million users. They need one team like theirs, with one job, before and after.

Every ad, outbound note, and sales call gets easier when you have that. The same first line lands harder when it can name a team like the buyer's, the job, and both numbers, and all of it is true. Without the case study, every channel is asking for trust on credit.

G2 reviews help too. They are public and written by people you do not control. But they are testimonials by design. Use them for volume of voices. Use case studies for the numbers.

What if I only have one good customer?

Write the one, then go get the next three.

One strong case study beats a logo wall. But one story that your whole narrative depends on is a risk. I covered that in don't build your GTM story on one mega-customer. Use the first case study to sell, and use it as the template to capture before numbers from every new customer on day one.

That is the move most founders miss. The before number disappears the moment the product goes live. If you do not record it at kickoff, you will never have it.

What should I do this week?

Add one question to your onboarding call: "How long does this job take today, and how do you know?" Write the answer down with the date. Thirty days later, ask again.

Then publish the ones where the customer will sign their name to both numbers.

Record the before. Or keep publishing testimonials.

Start Signals, Convert, Grow

FAQ

What is the difference between a case study and a testimonial?

A testimonial is an opinion. A case study shows one job, a before number, an after number, a time window, and a named customer who approved the numbers.

What metrics should an AI case study use?

Pick one measure tied to the job: hours per week, cost per unit, error rate, or volume. Use the same unit before and after, and state the time window.

What if the customer never measured the before?

Reconstruct it honestly from their records, or skip the number and say so. For new customers, ask for the before number at kickoff.

Are G2 reviews case studies?

No. G2 reviews are useful public testimonials from people you do not control. They add volume of voices. Case studies add the numbers.

How many case studies do I need?

Start with one strong one, then capture before numbers from every new customer so you are not dependent on a single story.