Deep dive
How Lucy learns what works.
“Lucy learns what works” is a precise claim, not a slogan. Our product design keeps five steps separate, because blurring them is how AI products overclaim.
In short
Lucy observes (with permission), forms a hypothesis about the goal, asks the person to confirm or correct it, measures completion, adoption, capability and value separately, and only reuses what has been qualified, across people only after privacy checks. The asset that compounds is knowledge of what helps, whom it helps and when.
- Observe
With permission, Lucy notes what actually happened: the apps, steps and timing of real work.
Example: Someone is researching a company. - Interpret
She forms a best guess about the goal, notes how sure she is and what else it could be, and stays quiet when evidence is weak.
Example: Meeting prep? A supplier check? Just learning? - Confirm
She asks. The person confirms or corrects her, and a correction always wins for that person.
Example: “It’s a supplier review on Friday.” - Measure
Completion, adoption, capability and value are measured separately. A finished task is not proof of skill.
Example: The review prep they built is used again next week. - Reuse, carefully
What worked is vetted before it helps anyone else, and reaches other people only after privacy checks, as protected aggregates or solutions their builders choose to share, never your business processes or trade secrets.
Example: A supplier-review workflow, offered to the next person who needs it.
The rules that keep it honest
Four commitments in the design.
A correction always wins
For that person’s intent. It is never used to train other companies’ AI models.
Results are measured separately
A time saving alone is not ROI.
Reuse is qualified and gated
One person’s success is not proof of general usefulness.
The person does the work
Nothing in the design lets Lucy click, type, install or deploy on someone’s behalf.
Plans work backwards from a goal through its prerequisites, so a novice and an expert get different paths to the same outcome. Success is capability, not usage: needing Lucy less for a task can be the win. This is the design our product team is implementing now.
Simple on the surface
Science underneath.
We plan to hire across six disciplines, each owning a different question.
Workflow research
What are people really doing, including handoffs and exceptions?
Applied science
What is this person probably trying to do, and when should Lucy ask?
Behavioral science
Which coaching changes habits, retention and independence?
Knowledge engineering
What does a goal require: tasks, tools, skills, prerequisites?
AI evaluation
Is the inference right, checked against independent examples?
Product data science
Did capability and the work actually improve, and why?
We also plan to hire for trust-focused design, security, reliability and integrated quality engineering, in stages.
Why it matters
Every win makes Lucy a better coach.
With each win, Lucy learns how to help people build better solutions, adapt to different preferences and understand different industries.
What helps, whom and when
Evidence, interpretation, outcome, then careful reuse
Capability, adoption, the relationship
Real work, a useful fix, a visible result, trust, the next problem
Productive capacity and coverage
Revenue, reserves, new hires, more customers served
Questions
Questions about how Lucy learns
What happens when Lucy guesses wrong about what I’m doing?
You correct her, and your correction wins. By design, Lucy forms a best guess about your goal, notes how sure she is and what else it could be, and stays quiet when the evidence is weak. When she asks and you tell her, for example, that the spreadsheet you’re cleaning is for a board pack, not a customer report, that becomes her understanding of your intent. The correction doesn’t rewrite the raw evidence of what happened.
Do beginners and experts get the same coaching?
No. By design, plans work backwards from your goal through its prerequisites: your current skill, the tools you’re allowed to use and the feasible next step. A beginner and an expert can aim at the same outcome by different paths, and a first plan can start from a goal you state, without waiting for weeks of observation.
Is a time saving the same as ROI?
Not on its own. Saved minutes become value only when they go somewhere: more customers served, faster decisions, fewer errors, less spend. That’s why Lucy measures completion, adoption, capability and business value separately, against a baseline. A finished task doesn’t prove a skill, and a skill doesn’t prove business value until the work itself improves.
Is all of this working today?
These five steps are the product design our team is implementing now. Today the Lucy desktop app observes how you work, with your permission, and coaches when you ask, and leaders have the Executive Dashboard. Proactive coaching is planned, and the six research disciplines described here are roles we plan to hire for. The current list of what’s here today and what’s coming is on How Lucy Works.
Keep exploring
More deep dives
What’s here today vs. what’s coming
Some features described on this site, like proactive coaching, are still being built.
