Fixed scope · $2,500
Human–AI Practice Diagnostic
A focused investigation for a team whose use of AI is inconsistent, shallow, confusing, or less valuable than expected.
A recurring problem that ordinary explanations have not resolved.
- Different people obtain radically different value from the same system.
- Advanced models produce generic work, or repeatedly miss the real task.
- Users distrust the AI, compensate for it, or avoid it without being able to name why.
- The team’s mental model of AI appears to limit the exchange.
- Training has not produced stable, independent capability.
- A useful workflow feels psychologically, linguistically, or conceptually malformed.
Observe, interpret, diagnose, name, report.
- Observe
- We watch two or three people use AI in real work and ask how they understand what they are doing.
- Interpret
- We read the enacted process through its language, assumptions, subject positions, policies, complexes, and relations.
- Diagnose
- We identify the unused capacity, misconception, or governing structure producing the pattern.
- Name
- We formulate the underlying principles in language the team can remember and use without us.
- Report and debrief
- We present evidence, recommendations, and a durable written account; then test the findings together.
One bounded engagement.
The people who observe the work also write the report and lead the debrief.
- Intake questionnaire and team-lead interview
- Two or three observation sessions
- Review of selected prompts, instructions, outputs, policies, and relevant materials
- Identification of unused capacities, recurrent pathologies, and governing misconceptions
- An 8–15 page diagnostic report with evidence
- Three to five named corrective principles
- Prioritized recommendations and a live debrief
- One brief follow-up after the team has tried the recommendations
Diagnosis is not an open-ended transformation program.
The engagement does not include mass training, model fine-tuning, organizational restructuring, ongoing management, indefinite support, or validation of a predetermined AI strategy.
It is also not open-ended implementation contracting. If a useful implementation artifact emerges, it can be proposed separately only when it is specific and bounded.
We do not accept work intended to improve coercive surveillance, military targeting, deceptive persuasion, predatory finance, labor domination, or comparable systems of harm.
What this work is—and is not.
- Is this prompt engineering?
- Prompt design may be one intervention, but the diagnostic examines the whole mind-process: context, expectations, subject-position, policy, habit, power, and interpretation.
- Are you claiming AI is conscious?
- No metaphysical verdict is required. We use the interpretation when it produces better understanding and intervention.
- Why a written report?
- It makes the observations durable, specific, and usable beyond the debrief, and gives the team a shared vocabulary for recognizing the patterns later.
- Can the engagement remain confidential?
- Yes. Publication of an anonymized case file requires explicit permission.
A direct application
Describe the practice you want examined.
The form replaces an exploratory sales call. Concrete, unpolished answers are welcome.