Case study · Nonprofit
AI-Assisted Production for the TLE Foundation
Summary
The TLE Foundation needed a complete training program, video lessons, workbooks, and audio materials, produced under a compressed deadline with a small staff. AI-assisted production workflows were used to draft and generate the material, with the organization reviewing and approving output. The program was delivered. No baseline was captured beforehand, so delivery is the outcome on record and nothing here is a measured benchmark.
64-word direct answer
Key takeaways
- Delivered outcome: a complete multi-format training program produced under deadline by a small team.
- No baseline was captured before the work, so no efficiency figure can be published, and none is.
- No client quotation appears on this page. The testimonial was withdrawn, so the record here is Martin's own account of his own work.
- No funding or grant outcome is claimed here, because the available evidence does not support a causal link.
At a glance
The fourteen-field template from the case studies standard. Fields that were not documented at the time say so.
| Field | Record |
|---|---|
| Client | TLE Foundation, named with approved wording. No attributed quotation is published, so no permission is outstanding |
| Context | Small nonprofit staff delivering a structured training program in multiple formats |
| Baseline | Not captured. No prior cycle time, cost, or output measurement exists for comparison |
| Problem | A complete multi-format program required under a compressed deadline, beyond conventional in-house production capacity |
| Constraints | Small staff, fixed deadline, nonprofit budget, multiple output formats required simultaneously |
| Intervention | AI-assisted production workflows for drafting and generating video lessons, training workbooks, and audio guides |
| Human oversight | Organization staff reviewed and approved material before release. Specific review points were not documented as a formal design at the time |
| Systems and tools | Not documented at the level of named tools. Not reconstructed retrospectively |
| Results | Program delivered in the required formats and on the compressed deadline. No outcome beyond delivery is claimed. The client characterization that previously appeared here has been removed |
| Measurement method | None. No instrumentation was in place; the outcome is qualitative and client-described |
| Attribution | AI-assisted workflows contributed to production capacity. No further outcome is attributed |
| Client quotation | None. Removed 5 August 2026 by Martin's instruction rather than held pending permission. What remains is his own account of his own work |
| Date and status | Date not recorded to a standard worth publishing. Current status of the materials unknown |
| Related | AI implementation · content and production systems |
Six of the fourteen fields are incomplete. That is the honest state of a project documented after the fact rather than as it ran, and it is the single best argument for the discipline described on the adoption framework page.
What the engagement involved
The TLE Foundation needed to deliver a training program consisting of video lessons, training workbooks, and audio guides. The deadline was compressed, and the staff available was small relative to what multi-format production conventionally requires: three distinct production pipelines, each with its own drafting, recording, and assembly work.
The approach was to use AI-assisted workflows for the drafting and generation stages of each format, with organization staff reviewing and approving output. That is the “draft and review” pattern described on the AI implementation page: the system produces a first version, a person edits and releases it. It is usually the lowest-risk place to start and frequently the highest immediate return, because production drafting is genuinely labor-intensive and the failure mode, a poor draft, is visible and cheap to correct.
The program was delivered in the required formats.
Takeaway
Martin helped the TLE Foundation use AI-assisted workflows to produce training videos, workbooks, and audio materials under a compressed deadline.
How to read the outcome
Everything above is Martin's own account of work he performed. The client's assessment of the result is not reproduced here, because it was withdrawn rather than held pending written permission, and describing it in other words would republish it by paraphrase.
Nobody counted the hours a conventional team would have taken, and no comparable production run was measured. There is therefore no productivity multiple to report, which is precisely why earlier material's “team of fifty” framing does not appear here as a statistic. What is documented is the scope delivered and the deadline it was delivered against, both of which Martin can attest to directly.
What was not measured, and why that matters
This is the most useful section of the case study, and the least flattering.
No baseline was captured. Nobody recorded how long comparable production had taken before, what it had cost, or how much material the team could previously produce in a given period. Without that, no efficiency claim is available: not “faster”, not a percentage, not a multiple. Any such figure would have to be constructed after the fact, which is how unfalsifiable statistics get made.
No measurement method was designed. The outcome is qualitative: the material was produced, the client was satisfied, and the client described the result in strong terms. That is real, and it is a different kind of evidence from a measured result.
Oversight was not formally designed. Staff reviewed output, which is the right behavior, but the review points were not specified in advance with a risk tier and a named reviewer. In a content-production workflow the consequences of that are mild. In a workflow touching client records or eligibility decisions they would not be.
The general lesson is the one on the adoption framework page: the baseline is captured at stage two, before anything changes, because it is the one input that cannot be recovered later. This engagement predates that framework being written down. It is part of why it was.
Attribution and limitations
What can reasonably be attributed: AI-assisted workflows contributed to the organization's production capacity within the deadline, in the client's assessment.
What cannot: any quantified productivity gain, any cost saving, and any downstream outcome of the program, including funding, enrollment, or participant results. No such outcome is claimed on this page, and if you encounter one attributed to this engagement elsewhere, it is not supported by anything on record here.
Other factors were present. The organization's existing subject expertise, its curriculum design, staff effort, and prior material all contributed to what was delivered. An AI-assisted drafting workflow accelerates production; it does not supply the expertise being taught.
Single case, no generalization. One engagement in one nonprofit on one content-production problem. It is not evidence that comparable results are available to other organizations, and it should not be read as a forecast.
Caution
A sequence of events is not a causal chain. That a client used AI-assisted production and later succeeded at something else are two facts; joining them is an inference the evidence does not support.
If your situation is similar
The pattern here transfers reasonably well to any organization producing structured material, such as training content, onboarding, documentation, or course material, with a small team and a real deadline. Drafting is the labor-intensive part, review is where the expertise lives, and separating the two is the whole idea.
Two things would be done differently now. The baseline would be captured first, even roughly: how long the last comparable piece took, and how much material exists today. And the review point would be written down, with a named reviewer, before the first draft was generated.
Neither adds much time. Both change what can honestly be said afterwards.
Frequently asked questions
Why is there no percentage or time-saving figure?
Because no baseline was captured before the work, so there is nothing to compare against. Any figure would have to be reconstructed retrospectively, which makes it unfalsifiable.
Aggregate figures of that kind are not published here. They are now listed as unverified on the credentials page and are not republished.
Did the AI work help the organization secure funding?
No such claim is made, and the evidence on record does not support one. What is supportable is that AI-assisted workflows helped the client complete their training materials under deadline.
Funding decisions turn on proposal merit, funder priorities, relationships, and timing. Attributing a funding outcome to a production workflow would be exactly the causal overreach the case study standard prohibits.
Which tools were used?
Not documented at the level of named tools, and not reconstructed after the fact.
The pattern matters more than the products in any case: AI-assisted drafting for each output format, with human review before release. The specific tools available for that change every few months.
Is the program still in use?
Unknown. Current status was not tracked, which is another gap this page records rather than glosses over.
Engagements documented from the start include a status field that is actually maintained.