Digital Human BY NEUROLOGIC AILet’s talk
Digital HumanINSIGHTS · FIELD NOTES

A different way
to see the work.

Essays on tacit knowledge, specialist models and the controls enterprise autonomy actually requires.

Explore the essays
THREE IDEAS TO RECONSIDER01The average isn’t the expert.02Reliability is the whole workflow.03Knowing goes beyond saying.

Perspectives on the knowledge, evaluation and controls that specialist AI requires.

Field notes from the work of building Digital Human
01
RESEARCH NOTE · 8 min READ

The averaging fallacy

Why your automation is exactly as good as your average employee, forever.

More mixed-skill data does not recover excellence. Compliance-gated, expert-weighted traces do.

Read the note
THE AVERAGING FALLACY
Mean: 12 minNobody works that way.
Illustrative task times. Quality and compliance come before speed.

Five task times. A mean of 12 minutes that describes none of them.

Explore each step
02
RESEARCH NOTE · 9 min READ

Why enterprise AI stalls between experiment and scale

Reliability compounds across every step of a workflow.

Scaling needs a learning asset, a workflow-level evaluation and a control model, not a better demo.

Read the note
RELIABILITY COMPOUNDS
98%36%
One step50 steps
1 step50 steps
Illustrative arithmetic: 0.9850 ≈ 36%, assuming independent steps. Not a product result.
03
RESEARCH NOTE · 10 min READ

Polanyi and the last automation frontier

We can know more than we can tell.

The frontier is no longer better retrieval from documentation. It is recovering expertise from execution.

Read the note
KNOWLEDGE HAS TWO LAYERS
WHAT IS WRITTENThe procedure.
WHAT EXPERIENCE ADDSTiming. Context.
Exceptions. Judgment.
Observe the work to connect the steps with the decisions behind them.

The written procedure preserves the visible steps.

Explore each step
FROM THESIS TO EVIDENCE

The test is not whether the idea sounds right.

It is whether the model carries one bounded workflow, with the customer holding the test set.

Explore a design partnership Talk to us
↗