Two decisions that point in opposite directions.

Your plan holds many decisions next to each other, such as two goals under the same strategy. When two decisions, neither of which sits above the other, point in opposite directions, Helen reads both, writes down why they conflict, and flags them for a person to decide. The check is called Semantic Contradiction.

How it works
AI judgment, with a person deciding
Where it stops
It compares only the business decisions at the top of your plan, and never a pair where one sits above the other. Helen stays quiet when her confidence is below a fixed level of 0.6 on a scale of 0 to 1.

What the check catches.

Helen's Foundry keeps the line from your strategy down to each piece of work as linked records. In the product, this line is called the context chain. This check reads the business decisions on that line, such as your mission, vision, strategy, goals, the customers you serve, and their problems. It skips drafts, closed decisions, and archived ones.

Helen picks pairs where neither decision sits above the other. That covers decisions under the same parent and decisions on separate branches. A pair qualifies when the two texts cover the same topic, share key words, or sit under the same parent. She reads each pair side by side and decides whether the two agree, repeat each other, contradict each other, or are unrelated. She raises this flag only for a contradiction she is confident in.

Chain Contradiction is the companion check. It compares a decision with one above it on the same line. Semantic Contradiction never compares those pairs, so the two checks never flag the same pair. When two texts are almost identical, Helen leaves the pair to a separate check for duplicates.

Helen keeps her judgment of each pair. She reads the pair again only when the title, type, or text of either decision changes, or when her instructions change.

What Helen records.

When two things no longer agree, Helen raises a flag. In the product, this flag is called a finding. Each Semantic Contradiction finding records both decisions, Helen's judgment, and how she found the pair. Her written reason is an AI judgment, and a person decides whether it holds.

  • Both decisions The title and type of each one.
  • Her judgment Her verdict, how confident she is, and her reason in plain language.
  • How she found the pair How similar the two texts are, which searches found the pair, and the version of the instructions she judged it with.

What a person does about it.

The flag goes to Reconcile, the place where a person decides what to do about a flag. Reconcile names both decisions and asks which one reflects your intent.

If one decision drifted, revise or close it. If both can hold at once, dismiss the flag and record why. Helen keeps a dismissed flag closed while the evidence behind it stays the same. Reconcile offers no one-step fix for this flag, so you make the change on the decision itself.

Reconcile shows this flag at Major, the highest of its three attention levels. Helen does not label it as blocking, and she does not stop anyone from working. Acknowledging the flag hides it from the default view until the finding changes. Helen never changes a decision on her own.

Research basis.

Locke, E. A., Smith, K. G., Erez, M., Chah, D. O., & Schneider, A. (1994). The effects of intra-individual goal conflict on performance. Journal of Management, 20(1), 67–91.

See what Helen flags in your own plans.

Plans start at $20 a month.