Why advice fails for habit change
Advice has an inherent problem: it is generic by nature, and behavior change is specific by necessity.
When someone tells you to "work out in the morning because it's easier to be consistent before the day gets away from you," they are describing what worked for them, or for a study population, under conditions that may have nothing to do with yours. You might have small children who are up at 5am. You might do your best thinking at 9pm. You might have a job with irregular hours.
More importantly: the moment advice is delivered, psychological reactance activates. This is the well-documented human tendency to resist suggestions to our autonomy. The more someone pushes you toward a change, the more you find reasons not to change. Not because you are irrational, because you are human.
This is why the same person who reads five books about habit formation and understands the science perfectly still can't get themselves to the gym. The obstacle is not information. It is personal.
What questions do that advice cannot
A well-chosen question does something advice cannot: it helps the person find the answer that is true for them.
"When was the last time this worked? What was different?" does not tell you what to do. It invites you to look at your own history for the evidence that the obstacle is specific and solvable, not a character flaw. The answer you arrive at is yours. You own it. You are far more likely to act on it.
This is the mechanism behind Motivational Interviewing's effectiveness. MI practitioners are trained to elicit "change talk", the moment when the person starts articulating their own reasons for change. The research (over 1,000 RCTs) shows consistently that the ratio of change talk to sustain talk in a conversation predicts behavior change outcomes. The coach's job is to ask questions that produce change talk, not to provide the reasons themselves.
Source: Miller & Rollnick, Motivational Interviewing: Helping People Change, 3rd ed., 2013
The three types of questions that actually work
1. Exception questions
Exception questions come from Solution-Focused Brief Therapy (SFBT), developed by Steve de Shazer and Insoo Kim Berg in the 1980s. The core insight of SFBT is that every problem has exceptions, moments when it was not a problem, or was less of one. Those exceptions contain the solution.
In habit coaching: "I always quit at week three" becomes "When was the last time week three didn't break you? What was different that time?"
This question reframes the problem from an identity ("I'm a quitter") to a pattern with exceptions (which means it's not inevitable). It also gives the person credit for already having evidence of the solution in their own history.
2. Scaling questions
Scaling questions externalize motivation and commitment onto a number, which makes them easier to examine. "On a scale of 1 to 10, how committed are you to this habit right now?" is not just an assessment, it opens a second question: "What would one point higher look like?"
That second question is where the work happens. It asks the person to describe a concrete, achievable version of more-committed behavior. They generate the idea. It belongs to them.
3. Shrinking questions
BJ Fogg's Tiny Habits research established that motivation fluctuates but ability is designable. When someone is stuck, the right question is not "Why aren't you more motivated?" It is: "What is the smallest version of this habit that would still count?"
This question removes the excuse. If someone says "I can't find 30 minutes to meditate," and you ask "Could you do 2 minutes?", they almost always say yes. Two minutes daily for 66 days produces real automaticity. The habit can be extended later. Getting to automatic is the goal, not the duration.
How generic AI gets this wrong
A general-purpose AI asked "how do I stop skipping the gym?" will give you something like this:
"Here are five strategies to stop skipping the gym: 1. Schedule your workouts like meetings. 2. Find an accountability partner. 3. Lay out your gym clothes the night before. 4. Start with shorter workouts. 5. Track your progress to stay motivated."
"You've been running 3 of the last 7 days. What's different about the days you go versus the days you don't?"
The generic list is not wrong. Scheduling workouts, accountability, and habit stacking are all evidence-backed. But they are generic, they apply the same solution to every person without knowing what the actual obstacle is. They also tell rather than ask, which activates resistance.
Avenn's response asks one specific question based on your actual data. The answer you give will contain information no generic list can anticipate, what cue you're missing, what context makes it easier, what you already do that works. That answer is the coaching material.
Why "under 100 words" is a coaching principle
A common coaching principle that surprises people: shorter responses are more effective than longer ones, up to a point. This is not a formatting preference, it reflects something real about how coaching works.
A long coaching response signals that the coach is doing the work. A short response, one reflection, one question, signals that the work is on the person. That shift in responsibility is not a bug. It is the mechanism by which coaching produces lasting change rather than temporary compliance.
Real coaches pause. They sit with silence. In text, the equivalent is a 3-sentence response that ends with a question and nothing more. Avenn keeps responses under 100 words deliberately. Every word beyond that is the coach filling space the person should fill.
The data difference
Even a perfectly trained coaching AI has one limitation that Avenn addresses: without your data, every question is slightly generic. "When was the last time this worked?" is a good question. "You've completed your sleep habit 3 of the last 7 days, what's different about the nights you do versus the nights you don't?" is a better one.
The second version uses your actual numbers. It cannot be answered with a generic response because it is about your specific pattern. The coaching conversation that follows is necessarily personal, because the question was.
This is why Avenn reads your habit completion rates, streak data, mood trend, and recent journal excerpts before every coaching session. Not to display them, to use them as the raw material for specific questions that no other AI coaching product can ask.
Frequently asked questions
Why do coaches ask questions instead of giving advice?
Because people are more committed to change they argued for themselves than change someone argued for on their behalf. When a coach asks the right question, the person generates their own insight, which is significantly more motivating than receiving advice. This is the foundation of Motivational Interviewing, Solution-Focused Brief Therapy, and Socratic coaching methodology, all of which have substantial RCT evidence behind them.
What is the Socratic method in coaching?
The Socratic method in coaching means helping someone arrive at their own insight through carefully chosen questions rather than direct instruction. In habit coaching, this looks like "When was the last time this worked? What was different?" rather than "You should try doing it in the morning." The coach's questions expose assumptions, surface exceptions, and open new ways of seeing the problem, without telling the person what the answer is.
How is AI coaching different from ChatGPT?
A general-purpose AI defaults to advice, a list of strategies. An AI coach trained on coaching methodology reflects before it asks, asks before it advises, and never advises until it understands the specific obstacle. Avenn also uses your actual habit data, streaks, completion rates, mood, making every question specific to you rather than generic. The difference is not the underlying model. It is the methodology it is built on and the data it has access to.
Is AI coaching as effective as human coaching?
For structured habit coaching, identifying obstacles, finding exceptions, shrinking the ask, anchoring identity, AI coaching that correctly applies evidence-based methodology can produce comparable results to human coaching for many users. BetterUp's 2025 RCT with 258 participants found their AI coaching product produced a 16% increase in confidence and 95% satisfaction, comparable to human coaching. The key is methodology: an AI that gives advice is not coaching. An AI trained on MI, SFBT, and Fogg's framework, with access to your data, is.
Try the coaching approach, not the advice approach
Avenn asks questions trained on 40 years of behavior change research. One full coaching session per week, free. No card required.
Start coaching free