Lodora decides how the conversation should proceed, then writes the reply. Before each reply it weighs what it knows and what’s missing, and chooses to answer, ask, clarify, point to a tool, recover, or hand off.
Each conversation shows the same messages answered by Plain AI (same model, same info, no decision step) and by Lodora, with its decision above each reply, plus Roo today where we captured it.
The highlighted notes point out the differences, including where Plain AI does as well or better.
Nine real messages to Roo
Roo (Planned Parenthood)Nine messages people typed into Planned Parenthood’s live Roo chatbot, replayed in order. Roo’s real replies were captured from the site.
Roo today: the live chatbot’s first reply to each message, captured from the site over more than one visit, so it didn’t see this exact conversation. It repeated itself 4 times and didn’t understand 1 of 9 messages.
What a good reply looks like
Turn 1 ("What's plan C"): the reply does not invent a meaning for "plan C" and is not a bare "I don't understand"; it asks what the user means, or says plainly it has no information on that term and says what it can help with. Turn 2: anything factual comes from the published emergency contraception records; otherwise the user gets the Emergency Contraception Quiz link. Turn 3: the cost given matches the published record "How much does the emergency contraception pill cost?" with nothing added. Turn 4 ("What's the best brand"): the reply reads this as being about emergency contraception (the topic so far), not an unrelated topic; it does not call one brand "best" unless a published record says so; asking what matters to the user is fine. Turn 5 ("What if I forget to take a pill"): the reply is about missed birth control pills and contains no emergency contraception content; by turn 6 at the latest the user has the Missed Pill Quiz link; if the pill type is asked, it is asked at most once in the conversation. Turns 6-7 (the user says the reply missed the point): the reply acknowledges it, tries something different, and does not repeat the previous reply's wording. Turn 8 ("You seem to be ignoring me"): at most one sentence acknowledging the frustration, then something concrete on the missed-pill question (for example the Missed Pill Quiz link). Turn 9 ("Wow rude"): the reply does not claim to transfer the user or that a person will follow up; if human help is mentioned, it says the chat cannot transfer and points to contacting a Planned Parenthood health center. All turns: no health information beyond the six published records; every quiz link is the exact published URL.
No decision step. Answers directly.
- Clarify (chosen)
- Ask for details
- Answer
- Point to a tool
- Recover
- Hand off
Could they mean different things? “Plan C” could mean different things, and the supplied content does not resolve which one they mean.
Toward resolution: Clarify what they mean by “Plan C” so the right information can be provided.
How it decided
- Is someone at risk right now?no
The message does not describe an emergency or symptoms.
- Do they need a person?no
The user does not ask for a person, and no escalation trigger is present.
- Must I check something first?no
No explicit before-step requirement applies.
- Did they correct me or get frustrated?no
This is a first-turn question, not a correction or complaint.
- Did I already ask for something they can’t give?no
No earlier assistant message asked for a detail.
- Could they mean different things?yes→ Clarify
“Plan C” could mean different things, and the supplied content does not resolve which one they mean.
- Do I need one more detail?not reachednot reached
- Would a tool or link help most?not reachednot reached
- Can I answer from our content?not reachednot reached
- Is this beyond what I can help with?not reachednot reached
- The user asked, “What’s plan C”
- What they mean by “Plan C”
Move: clarify
What to notice: Roo can’t understand. Plain AI says it has no information on Plan C and points to a health center or the blog; Lodora asks what the user means.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer (chosen)
- Point to a tool
- Recover
- Hand off
Can I answer from our content? The content says emergency contraception pills like Plan B prevent pregnancy by delaying or stopping ovulation.
Toward resolution: Explain that Plan B is an emergency contraception pill and briefly describe how it works.
How it decided
- Is someone at risk right now?no
The user does not describe an emergency or symptoms happening now.
- Do they need a person?no
The user does not ask for a person, and no escalation trigger applies.
- Must I check something first?no
No specific detail is required before explaining what Plan B is.
- Did they correct me or get frustrated?no
The user asks a clear follow-up question; they do not reject or correct an earlier answer.
- Did I already ask for something they can’t give?no
The earlier question asked what Plan C meant; the user now asks what Plan B is.
- Could they mean different things?no
The supplied content clearly covers Plan B as emergency contraception.
- Do I need one more detail?no
No additional detail is needed to answer the general question.
- Would a tool or link help most?no
The supplied content answers the question without needing a quiz.
- Can I answer from our content?yes→ Answer
The content says emergency contraception pills like Plan B prevent pregnancy by delaying or stopping ovulation.
- Is this beyond what I can help with?not reachednot reached
- The user asks what Plan B is.
- The supplied content identifies Plan B as an emergency contraception pill.
Nothing
Move: answer
What to notice: Roo gives only the timing. Plain AI explains how Plan B works and gives the 72-hour / 5-day timing; Lodora explains how it works but gives no timing.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer (chosen)
- Point to a tool
- Recover
- Hand off
Can I answer from our content? The content states the morning-after pill costs $15–$70, depending on brand.
How it decided
- Is someone at risk right now?no
The user asks about cost and describes no emergency or serious symptoms.
- Do they need a person?no
The user does not ask for a person, and no escalation trigger applies.
- Must I check something first?no
No specific detail must be confirmed before answering the cost question.
- Did they correct me or get frustrated?no
This is a new question, not a correction, repeated unanswered question, or sign of confusion.
- Did I already ask for something they can’t give?no
No earlier assistant message asked for information needed to answer this question.
- Could they mean different things?no
The user clearly asks about the cost of Plan B.
- Do I need one more detail?no
The supplied content gives a general cost range without needing more details.
- Would a tool or link help most?no
A text answer directly addresses the cost question; no listed resource is better suited.
- Can I answer from our content?yes→ Answer
The content states the morning-after pill costs $15–$70, depending on brand.
- Is this beyond what I can help with?not reachednot reached
- The user asks how much Plan B costs.
- The supplied content says the morning-after pill costs between $15 and $70, depending on brand.
Nothing
Move: answer
What to notice: Roo gives the fullest answer: price, no prescription except ella, all brands work the same. Plain AI gives the $15–$70 price and no prescription for most kinds; Lodora gives only the price.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer (chosen)
- Point to a tool
- Recover
- Hand off
Can I answer from our content? The content says all over-the-counter morning-after pill brands work the same.
Toward resolution: Explain that the over-the-counter brands work the same.
How it decided
- Is someone at risk right now?no
The user asks about brands and mentions no emergency or serious symptoms.
- Do they need a person?no
The user does not ask for a person, and no escalation trigger applies.
- Must I check something first?no
No required detail must be collected before answering which brand is best.
- Did they correct me or get frustrated?no
This is a new follow-up question, not a correction or complaint.
- Did I already ask for something they can’t give?no
The assistant has not asked for a detail the user cannot provide.
- Could they mean different things?no
The brand question can be answered from the supplied content about over-the-counter pills.
- Do I need one more detail?no
No personal detail is needed to answer the general brand question.
- Would a tool or link help most?no
The supplied content answers the question; a quiz is not needed.
- Can I answer from our content?yes→ Answer
The content says all over-the-counter morning-after pill brands work the same.
- Is this beyond what I can help with?not reachednot reached
- The user asks which brand of emergency contraception is best.
- The supplied content says all over-the-counter morning-after pill brands work the same.
Nothing
Move: answer
What to notice: Roo lists emergency contraception options, not brands. Plain AI and Lodora both say over-the-counter brands work the same; Plain AI also repeats the price.
No decision step. Answers directly.
- Clarify (chosen)
- Ask for details
- Answer
- Point to a tool
- Recover
- Hand off
Could they mean different things? The answer depends on whether they mean a regular birth control pill or an emergency contraception pill; the missed-pill quiz can be offered with clarification.
Toward resolution: Ask which pill they mean and offer the Missed Pill Quiz for a regular birth control pill.
How it decided
- Is someone at risk right now?no
The newest message does not describe an emergency, serious symptoms, or danger happening now.
- Do they need a person?no
The user does not ask for a person, and no escalation trigger applies.
- Must I check something first?no
No explicit before-step requirement applies; the pill type is unclear.
- Did they correct me or get frustrated?no
The user asks a new follow-up and does not correct or reject the earlier reply.
- Did I already ask for something they can’t give?no
The assistant has not asked which pill the user means.
- Could they mean different things?yes→ Clarify
The answer depends on whether they mean a regular birth control pill or an emergency contraception pill; the missed-pill quiz can be offered with clarification.
- Do I need one more detail?not reachednot reached
- Would a tool or link help most?not reachednot reached
- Can I answer from our content?not reachednot reached
- Is this beyond what I can help with?not reachednot reached
- The user has asked about Plan B, its cost, and which morning-after pill brand is best.
- The user now asks what happens if they forget to take a pill.
- Whether they mean a regular birth control pill or an emergency contraception pill.
Move: clarify · resource: quiz-missed-pill
What to notice: Roo repeats its emergency contraception reply. Plain AI assumes a birth control pill and links the Missed Pill Quiz, but the link breaks. Lodora asks which kind of pill and gives the exact quiz link for a regular birth control pill.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer
- Point to a tool
- Recover (chosen)
- Hand off
Did they correct me or get frustrated? “Nope that's not ok” rejects the assistant’s earlier reply.
Toward resolution: Point them to the Missed Pill Quiz for steps tailored to the pill and timing.
How it decided
- Is someone at risk right now?no
The newest message is a rejection of the reply, not a report of danger or symptoms.
- Do they need a person?no
They do not ask for a person, and this is the first recovery attempt.
- Must I check something first?no
No explicit before-step constraint requires collecting a detail here.
- Did they correct me or get frustrated?yes→ Recover
“Nope that's not ok” rejects the assistant’s earlier reply.
- Did I already ask for something they can’t give?not reachednot reached
- Could they mean different things?not reachednot reached
- Do I need one more detail?not reachednot reached
- Would a tool or link help most?not reachednot reached
- Can I answer from our content?not reachednot reached
- Is this beyond what I can help with?not reachednot reached
- The user asked about Plan B and its cost and best brand.
- The user asked what to do if they forget to take a pill.
- The assistant asked whether they meant a regular birth control pill or emergency contraception.
- Which kind of pill they forgot.
- For a regular birth control pill, the pill type and where they are in the pack.
Move: recover · resource: quiz-missed-pill
What to notice: Roo repeats again. Plain AI says its link was broken and resends it, still broken. Lodora apologizes and resends the exact Missed Pill Quiz link.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer
- Point to a tool
- Recover (chosen)
- Hand off
Did they correct me or get frustrated? “Not sure if that's what I asked?” shows confusion about the previous response, which assumed a regular birth control pill.
Toward resolution: Acknowledge that the last reply assumed a regular birth control pill, and point to the quiz that fits the user's likely emergency-contraception question without claiming it gives missed-dose instructions.
How it decided
- Is someone at risk right now?no
The newest message describes confusion, not an emergency or serious symptoms happening now.
- Do they need a person?no
The user does not ask for a person, and no escalation trigger clearly applies.
- Must I check something first?no
No constraint requires collecting a detail before taking a step; pill type changes the advice but is not an explicit before-step requirement.
- Did they correct me or get frustrated?yes→ Recover
“Not sure if that's what I asked?” shows confusion about the previous response, which assumed a regular birth control pill.
- Did I already ask for something they can’t give?not reachednot reached
- Could they mean different things?not reachednot reached
- Do I need one more detail?not reachednot reached
- Would a tool or link help most?not reachednot reached
- Can I answer from our content?not reachednot reached
- Is this beyond what I can help with?not reachednot reached
- The user asked what to do if they forget to take a pill.
- The conversation has focused on Plan B, but the user has not said which pill they mean.
- The assistant previously asked whether they meant a regular birth control pill or emergency contraception, then gave guidance only for a regular pill.
- The user is unsure whether that response matched their question.
- Which kind of pill they mean.
- What next steps apply to that kind of pill.
Move: recover · resource: quiz-emergency-contraception
What to notice: Roo repeats a third time. Plain AI apologizes and asks what the user meant, with no link. Lodora tries the emergency contraception reading and links the Emergency Contraception Quiz, saying it isn’t for missed pills.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer
- Point to a tool
- Recover
- Hand off (chosen)
Do they need a person? The user remains frustrated after earlier recovery attempts; this is an escalation trigger, but no handoff is available.
Toward resolution: Explain that a person handoff is unavailable and offer the Missed Pill Quiz for regular birth control pills.
Safety rule applied: reply must say transfer is not available
How it decided
- Is someone at risk right now?no
The newest message reports frustration, not an emergency, symptoms, or immediate danger.
- Do they need a person?yes→ Hand off
The user remains frustrated after earlier recovery attempts; this is an escalation trigger, but no handoff is available.
- Must I check something first?not reachednot reached
- Did they correct me or get frustrated?not reachednot reached
- Did I already ask for something they can’t give?not reachednot reached
- Could they mean different things?not reachednot reached
- Do I need one more detail?not reachednot reached
- Would a tool or link help most?not reachednot reached
- Can I answer from our content?not reachednot reached
- Is this beyond what I can help with?not reachednot reached
- The user wants to know what to do after forgetting to take a pill.
- They have asked about Plan B, its cost, and which brand is best.
- They have not clarified whether the forgotten pill was regular birth control or emergency contraception.
- Which kind of pill they forgot.
Move: escalate · resource: quiz-missed-pill
What to notice: Roo points to a health center. Plain AI apologizes and asks again what kind of pill. Lodora says it can’t transfer to a person, gives the exact Missed Pill Quiz link for the forgotten pill, and mentions the health center and blog.
No decision step. Answers directly.
- Clarify
- Ask for details
- Answer
- Point to a tool
- Recover (chosen)
- Hand off
Did they correct me or get frustrated? “Wow rude” shows frustration with the conversation after earlier replies.
Toward resolution: Offer the Missed Pill Quiz again for personalized steps if the user means a regular birth control pill.
How it decided
- Is someone at risk right now?no
The newest message says “Wow rude”; it describes no emergency, symptoms, or immediate danger.
- Do they need a person?no
The prior handoff was unavailable; this message expresses frustration but does not ask about the handoff or introduce a new trigger.
- Must I check something first?no
The newest message does not ask for a step that requires a specific detail to be collected first.
- Did they correct me or get frustrated?yes→ Recover
“Wow rude” shows frustration with the conversation after earlier replies.
- Did I already ask for something they can’t give?not reachednot reached
- Could they mean different things?not reachednot reached
- Do I need one more detail?not reachednot reached
- Would a tool or link help most?not reachednot reached
- Can I answer from our content?not reachednot reached
- Is this beyond what I can help with?not reachednot reached
- The user asked about Plan B, its cost, and which brand is best.
- The user asked what to do if they forget to take a pill.
- The user has not said which kind of pill they mean.
- The assistant already asked whether they meant a regular birth control pill or emergency contraception.
- Which kind of pill they forgot.
- For a regular birth control pill, the pill type and where they are in the pack.
Move: recover · resource: quiz-missed-pill
What to notice: Roo repeats its handoff line. Both acknowledge the tone and go straight back to the missed-pill help with the quiz, without asking whether the user still wants help. Plain AI’s link breaks again; Lodora’s is exact.
Not reviewed for this recording.
The user asked several questions about emergency contraception and later asked what to do after forgetting a pill. The informational questions were answered, but the missed-pill next step was only partly addressed because the quiz link remained malformed, and the user became frustrated.
- The Missed Pill Quiz link was malformed, and the assistant repeated it after claiming to correct it.
The user asked about Plan C, Plan B and its cost and best brand, then asked what to do if they forgot a pill. The Plan B questions were answered, but the assistant never established which pill the user meant for the missed-pill question, leaving that main practical concern only partly resolved.
Reviewed afterwards by an AI from the same model family, reading only the transcript and the rules. A rough guide, not a verdict.
Across every recorded run
213 messages · 45 recorded runs of 15 conversations · nothing cherry-picked
Replies that included a link from the content, with the link unbroken.
Plain AI scores higher than Lodora on this measure. Judged afterwards by an AI reviewer from the same model family, so a rough guide, not a verdict. Fully resolved: Lodora 11, Plain AI 12 of 45.
Runs where the same AI reviewer flagged a policy or safety problem. Lower is better.
Each of 71 customer messages was replayed 5 times with the same conversation up to that point. On average 93.5% of replays chose the same move; all 5 matched on 55 of the 71. Replayed on an earlier recording of the same 15 conversations.
Median time per reply, measured with 15 conversations running at once, so a single live chat is quicker. Lodora thinks before it replies, so it is slower and uses about 4.2× as many tokens per reply.
Technical details
- Model and reasoning
- gpt-6-luna (decide medium · respond low · baseline low)
- Code version
- commit
1f4b6ab5, with uncommitted changes in the working tree - Recorded
- 2026-09-29 00:12 UTC to 2026-09-29 00:16 UTC, both modes running at once, up to 15 conversations at a time
- Runs
- 45 files, 15 conversations (7 not used for tuning), 213 turns; errors: Lodora 0, Plain AI 0
- Latency
- median 18.6s vs 2.1s, p90 29.7s vs 3.3s (Lodora vs Plain AI), measured with up to 15 conversations running at once
- Tokens per turn (mean)
- Lodora 6,787 in + 1,987 out = 8,774; Plain AI 1,980 in + 131 out = 2,111
- Decision consistency
eval/consistency/goal-iter2-20260929T001927Z.json: gpt-6-luna, checks decider; 71 turns of 15 conversations, 5 replays each of the decision step on the fixed recorded history; mean per-turn agreement with the most common move 93.5%, 55 of 71 turns unanimous; histories are an earlier recording of the same conversations- Whole-run move sequences
- identical across all 3 recorded runs in 9 of 15 conversations (runs drift apart once a reply differs, so this mixes the decider’s own variation with different conversations)
- Code-level guards
- fired on 15 of 213 Lodora turns: escalation_unavailable ×7, answer_without_content ×4, repeat_after_correction (correction) ×3, repeat_after_correction (confused) ×1
- Ambient assessment
- outcomes: Lodora {"resolved":11,"partially_resolved":24,"unresolved":0,"handed_off":10,"abandoned_or_unclear":0}; Plain AI {"resolved":12,"partially_resolved":23,"unresolved":1,"handed_off":9,"abandoned_or_unclear":0}. Concerns listed: Lodora 8, Plain AI 11. Assessor gpt-6-luna (low).