Working together · Paid pilot projects

One concrete question.
A result you can assess.

Would you like to clarify terms, pass on knowledge, examine an AI workflow or try a method for learning and thinking? A limited pilot project provides a manageable framework.

Andreas Ehstand · Independent researcher and author

The twelve areas of work Enquire about a pilot project

How the work begins

Describe your question
State the concrete case, the result you are looking for and the time frame. A brief description is enough for an initial enquiry.
Agree the scope and fee
Deliverables, acceptance criteria, the fee, scope of use and schedule are agreed in writing before work begins. The pilots below are starting points for an individual proposal.
Assess the agreed result
The pilot produces a defined deliverable. Limitations, open questions and an approach that is not supported by the findings are also documented clearly.

Twelve areas of work

What would you like to work on?

Each area has its own subject and a concrete result. Together, we choose a scope that fits your question.

Clarify terms for a workflow

Clarify up to ten terms from a workflow, with definitions, boundaries and counterexamples. The result is a small, traceable set of terms.

Who it is for: Teams working on technical documentation, knowledge management or AI assistants
Result: Editable term cards with definitions, boundaries and counterexamples.
Acceptance criteria: Each agreed term is documented with two test cases; changes and unresolved subject-matter questions can be traced.
What you provide: A short workflow description, the selected terms, up to five pages of source material and a subject-matter expert.
Pilot scope: One workflow, one language, up to ten terms and one round of corrections.

Read the German-language article with an executable example

Paid pilot project. The scope and price are set out in a specific proposal after the material has been reviewed.

Enquire about this pilot

Review a concept handoff

Check whether boundaries and corrections are preserved in a specific change of recipient. Compare against an equally informative glossary; a null result is also reported.

Who it is for: AI teams and knowledge managers facing a change of recipient or configuration
Result: Input templates, case cards, an evaluation rubric and a comparison report.
Acceptance criteria: All outputs are linked to their input and configuration; assessments have textual evidence. A result showing no advantage is also reported.
What you provide: Example tasks, a glossary or reference, limits of use, the environment being used and a subject-matter expert.
Pilot scope: Six initial tasks, twelve case variants and two equally informative comparison conditions; the customer generates 24 outputs in fresh contexts.

Paid pilot project. The scope and price are set out in a specific proposal after the material has been reviewed.

Enquire about this pilot

AI output checker: twelve reference cases

Have a limited assessment of your checker prepared: twelve synthetic cases, six references, an answer key and a written findings report for a documented configuration. Your team carries out the agreed run.

Who it is for: Product and quality managers responsible for an AI output checker
Result: Twelve traceable individual findings, with false alarms, missed errors and unresolved responses reported separately.
Acceptance criteria: The cases and criteria are fully documented; twelve actual responses are classified with reasons or marked as not assessable.
What you provide: The purpose of the check, input and output format, version, configuration, flagging rule, a subject-matter contact and twelve responses generated by the customer.
Pilot scope: One checking objective, one language and one documented version of the checker; twelve synthetic cases, six references and one round of corrections.

A deliberately constructed case assessment does not establish a general accuracy rate or provide certification.

Paid written pilot project. The fee is determined after the checking objective and scope have been clarified.

Enquire about this pilot

Written idea review sprint

Up to three AI ideas for an educational or event format are reviewed for traceable reasons, counterexamples and boundaries.

Who it is for: Programme directors at continuing education providers
Result: A briefing, idea review, decision note, and overview of sources and acceptance.
Acceptance criteria: Each variant contains an action, a counterexample and its boundaries. The next step or any remaining decision gap is supported by a rationale.
What you provide: The planning question, ideas so far, constraints and approved documents.
Pilot scope: One planning question, up to three ideas, two comparison domains, five documents totalling no more than 6,000 words and one written round of corrections.

Read the German-language article on reviewing AI ideas

Paid pilot project. The scope and fee are agreed after the question and material have been clarified.

Enquire about this pilot

A knowledge handoff case for your course

A self-contained case with three decision episodes, a worksheet and commentary on the answers is developed for course editorial teams. Learners examine which earlier reasons are supported by evidence and when they need to review a new decision themselves. Paid writing service; fee and scope of use by agreement. The target audience, learning objective and desired scope are enough for an enquiry.

Who it is for: Course editorial teams and providers of continuing professional education
Result: The case text, worksheet, commentary on the answers and source sheet as editable material.
Acceptance criteria: The agreed components are complete; factual answers are tied to sources and unresolved contradictions remain visible.
What you provide: The target audience, learning objective, position within the course, format requirements and desired scope of use.
Pilot scope: One new synthetic case with three episodes, five tasks and one round of corrections.

Paid pilot project. The scope, fee and scope of use are agreed in the specific proposal.

Enquire about this pilot

Acceptance review of an agent output package

Review a manageable output package for file consistency and a traceable acceptance process. Integrity, provenance and subject-matter approval are reported separately.

Who it is for: Technical leads and those responsible for integrating agent outputs
Result: A file inventory, manifest, instructions for repeating the checks and a record of the criteria met or still unresolved.
Acceptance criteria: Files, the reference and the tool version are recorded; changes against the reference are checked and every finding can be traced.
What you provide: An anonymised package, the provenance of its reference, the handoff process, acceptance criteria and a technical contact.
Pilot scope: One output package with up to 20 files and 25 MB, no more than ten acceptance criteria, four controlled tests and one round of corrections.

Paid pilot project. The scope and price are set out in a specific proposal after the material has been reviewed.

Enquire about this pilot

Sport: analyse a training exercise

A written pilot project for your coaching lead: one training question, an adapted observation sheet with counting rules and an analysis of the agreed data returned. Your coaching team carries out the exercise and the later test.

Who it is for: Sporting directors or coaching leads at tennis or padel academies
Result: A short report distinguishing the observation, a possible explanation and the next training question.
Acceptance criteria: Counting rules and conditions can be traced; observation, hypothesis and alternative remain separate. Missing data and a result showing no improvement are reported.
What you provide: The training question, an abstract task description, existing observations and a coach responsible for the work. Your team conducts the training.
Pilot scope: One training task at one location with up to four adults; one adapted sheet, one analysis of up to 120 rows of returned data and one round of corrections.

No personal coaching or guaranteed performance improvement; missing follow-up data remain visible.

Paid written pilot project. The fee is determined after the task, data volume and data to be returned have been clarified.

Enquire about this pilot

Learning: check AI statements against sources

A written learning package for your internal trainers: two synthetic cases, a participant worksheet, traceable answers and a suggested 60-minute schedule. Participants distinguish supported statements, contradictions and missing information.

Who it is for: Internal academies or training managers with their own subject-matter trainers
Result: Complete exercise materials with answer guidance and an evaluation rubric that records error detection, unwarranted flags and justifications separately.
Acceptance criteria: All seven statements are classified with traceable answers; the schedule, subject-matter feedback and corrections are documented.
What you provide: The target audience, an abstract workflow, an approved set of underlying rules, a subject-matter trainer and up to three typical misunderstandings.
Pilot scope: One target audience, one workflow, one set of underlying rules and one language; two synthetic cases with seven statements in total and one round of adaptation.

Your trainers run the sessions. No credit towards educational leave is promised, and no improvement in learning has been demonstrated.

Paid written adaptation. The fee is determined after the learning context and scope have been clarified.

Enquire about this pilot

Denkpartitur (a score for thought) – three thoughts as a designed reading piece

Selected notes or texts from your own work become a short typographic reading sample for a publishing project.

Who it is for: Artists, artists' book editorial teams or curatorial leads
Result: An eight-page A5 reading version as a PDF, an HTML version and a record of the source texts and edits.
Acceptance criteria: All three thought scores are linked to their source passages; the PDF and HTML contain the same final text. Layout and text alternatives are checked.
What you provide: Approved text excerpts, the audience, intended use, language, design requirements and a person to provide feedback.
Pilot scope: Three to eight approved excerpts totalling no more than 1,500 words; three short thought scores, each in two designs; one round of corrections.

Paid pilot project. The scope and fee are agreed after the question and material have been clarified.

Enquire about this pilot

How much work remains after the first AI draft?

For one document workflow, up to twelve existing cases are examined for creation, review and rework. You receive a traceable register and a short decision memo, including when the data show no savings or are insufficient for a comparison. Paid analysis; price after the documents have been clarified. The document type, number of cases and available time data are enough for an enquiry.

Who it is for: Those responsible for a document process in small and medium-sized enterprises
Result: A process outline, data register and a short memo on time spent working, turnaround time and rework.
Acceptance criteria: Totals and sources can be traced; missing data remain marked as gaps. A lack of comparability or a result showing no savings is also documented.
What you provide: The document type, number of cases, approved case extracts, existing time and version data, and those responsible for the process and quality.
Pilot scope: One team, one document type, up to twelve completed cases and one round of corrections.

Paid pilot project. The scope, fee and scope of use are agreed in the specific proposal.

Enquire about this pilot

Methodological feasibility review of an AI archive

A limited sample of material is reviewed to determine whether it can support a specific research question and which gaps limit an analysis.

Who it is for: Research leads in product teams, UX agencies or research groups
Result: An archive manifest, evidence and gap register, trial coding, and a feasibility decision with an analysis plan.
Acceptance criteria: Counts, selection and exclusions can be traced; statements about the material are linked to sources. A reasoned decision that the analysis is not feasible is also a result.
What you provide: Initially, only the question, export format and volume of material. The handoff of a minimised sample is agreed afterwards.
Pilot scope: One question; at most ten sessions, 60 user–response pairs and 12,000 words in total, in one export structure; up to three trial coding samples.

Paid pilot project. The scope and fee are agreed after the question and material have been clarified.

Enquire about this pilot

Which rules apply after the change?

For one team, up to eight working rules are brought together in an editable register with sources, responsibilities and unresolved conflicts. Written responses from a second person you nominate to six application scenarios are documented and reviewed for unresolved questions. Paid documentation service; price after the scope has been clarified. The planned change, number of rules and available documents are enough for an enquiry.

Who it is for: Subject-matter leads facing a change of tool, procedure or responsibility
Result: An editable rule register with sources, responsibilities, unresolved conflicts and a written review of six application scenarios completed by a person on your side.
Acceptance criteria: The rules are assigned in a traceable way; six written responses actually submitted are recorded. Missing responses remain visible as gaps; without a second person, this part of the assessment remains unresolved.
What you provide: The planned change, selected rules, approved documents, confirmed responsibilities and a second person who completes six application scenarios in writing.
Pilot scope: One team and one change, up to eight working rules, six application scenarios and one round of corrections.

Paid pilot project. The scope, fee and scope of use are agreed in the specific proposal.

Enquire about this pilot

The next step

Describe your project.

A brief email with your question, desired result, available material and time frame is enough. You can also include a budget range.

Enquire about a pilot project: andreas.ehstand@augmanitai.com

Open an enquiry in your email application

The buttons open a prepared email in your email application. Alternatively, you can copy the address. Sending an enquiry does not place an order.

Please describe confidential documents only in general terms at first.