Evidence
Case Studies
What is published here
One documented engagement, written to a fourteen-field template that records baseline, intervention, human oversight, measurement method, attribution, and limitations. Every numerical claim states whether it is measured, estimated, client-reported, or modeled. The list is short because the standard is strict and earlier work was not documented this way at the time.
51-word direct answer
Key takeaways
- One case study. The count reflects the documentation standard, not the volume of work done.
- Every figure is labeled measured, estimated, client-reported, or modeled. Unlabeled numbers are not published.
- The template includes limitations and attribution, the fields that make a case study checkable rather than promotional.
- Where a field was not documented at the time, it says so rather than being quietly omitted.
Published case studies
Nonprofit · Content production
AI-assisted production for the TLE Foundation
Training videos, workbooks, and audio materials produced under a compressed deadline using AI-assisted workflows. Documented with a full account of what was and was not measured.
- Outcome is client-reported, not a measured benchmark
- No baseline was captured, stated plainly
- Client permission for attributed use: pending
Why there is only one
Earlier engagements were not documented at the time to a standard that survives scrutiny: no captured baselines, no recorded measurement methods, no written permissions. Reconstructing those retrospectively is not possible, and inventing them is not an option.
Publishing aggregate client counts and efficiency percentages without the evidence behind them is exactly what this rebuild removed. Those claims are listed as unverified on the credentials page.
Engagements from this point are documented as they run. One honest case study is worth more than a page of unfalsifiable numbers.
The fourteen-field template
Every case study here answers all fourteen. Where a field cannot be answered, the case study says so; an unanswerable field is information too.
Client and permission status
Named where permitted, with the permission position stated explicitly.
Context
The organization, its scale, and the situation the work sat in.
Baseline
What was measured before anything changed. If nothing was, that is recorded.
Problem
The specific operational problem, not a general aspiration.
Constraints
Deadline, budget, capacity, data, and policy limits that shaped the approach.
Intervention
What was actually built or changed.
Human oversight
Who reviewed what, and at which point output could take effect.
Systems and tools
What was used, at the level of detail the client permits.
Measured results
Outcomes, each labeled measured, estimated, client-reported, or modeled.
Measurement method
How the result was arrived at. Without this, a result is an assertion.
Attribution and limitations
What can and cannot be credited to the intervention, and what else was going on.
Client quotation
In the client's own words, used only with permission.
Date and current status
When it ran, and whether the workflow is still in use.
Related expertise and service
Where to read more, and where the commercial equivalent lives.
Principle
A case study without a measurement method is an advertisement. The method is what allows a reader to disagree with the conclusion, which is what makes agreeing with it worth anything.
On attribution
The strongest temptation in a case study is to let a sequence of events imply a cause. A client used AI-assisted workflows, then secured funding; writing both facts next to each other invites the reader to connect them.
That connection is not evidence. The specification governing this rebuild is explicit on the point: do not imply that AI caused funding approval when the evidence only shows that AI-assisted production helped the client complete materials. Funding decisions turn on the merit of a proposal, the relationship, the funder's priorities, and timing.
So each case study separates what was done, what was observed, and what can be credited, and states the third conservatively.
Client permission
No client is named, quoted, or identified by inference without permission. Where a client prefers not to be named, the engagement can still be documented with the organization described by sector and scale.
Where permission is pending rather than granted, the case study says so at the top. That is the current position for the TLE Foundation study: the underlying facts are supportable and the wording is the approved form, but written consent for attributed use has not yet been recorded.
Logos are not displayed without written permission for that specific use. Generic logo bars are a credibility risk, and unapproved client logos are not published here.
Frequently asked questions
Why not include more examples, even briefly?
Because a brief example with no baseline, method, or permission is the thing this site is built to avoid. A list of client names and impressive-sounding outcomes reads well and tells a reader nothing they can check.
The list will grow as engagements are documented properly from the start.
Can I contact the client to verify?
Where permission has been granted for a reference conversation, yes. Ask via contact and it can be arranged with the client's agreement.
Contacting a named client directly without that arrangement is not something this site invites, out of respect for their time.
What if my organization does not want to be named?
That is entirely normal and easily handled. An engagement can be documented as, for example, “a Hawaiʻi-based nonprofit of roughly thirty staff” with the substance intact.
Naming is never a condition of the work, and no engagement is documented publicly without agreement.