25 years of expert coaching,
compressed into patterns.
Illustrative trajectories drawn from Heather's methodology — composite shapes that recur across the practice. Each inflection reflects the coaching frameworks behind it.
Everything shown on this page is aggregate, pattern-level, and anonymized. No individual client conversation, note, or trajectory is reproduced, attributed, or used in any way that could identify someone. What HeatherAI delivers is Heather's coaching methodology — the frameworks and the voice — not anyone's personal record.
From moments to playbooks.
One real moment at a time. Pattern matching extracted from 1M+ coaching conversations, mapped to the play that worked.
A real moment fires. The network routes it to the matching pattern — extracted from 1M+ coaching conversations — and back out as the play that worked. Continuously, in real time.
When coaching needs to show up.
When people need coaching most.
Friday evening. Saturday afternoon. Late Sunday. The patterns reveal when most people need coaching — so HeatherAI meets you there.
Five named frameworks, thousands of playbooks.
Coaching Archetypes
The Daily Logger, Executive, Returner, Under Pressure, Post-GLP1, Maintainer. The AI identifies which archetype you match and adjusts its coaching posture to fit.
Trigger Playbooks
Every trigger (business dinner, scale bump, post-vacation, stress-eating, 10pm craving) paired with the coaching response that worked.
Recovery Strategies
The Protein Day reset. "Today is a new day." Zero judgment on return. Extracted from thousands of real comebacks.
The Coaching Arc
Intensive → autonomy-building → maintenance → self-managing → re-engagement. The AI knows where you are and adjusts intensity.
Weight Trajectory Intelligence
Plateau → breakthrough → regain shapes. Every inflection point tied to the coaching conversation that produced it.
The academic record catches up.
The published evidence is catching up to what Heather's clients have known for years. Here's what the research says about the approach HeatherAI is built on.
AI coaching matched human coaching outcomes over 12 months.
A phase 3 randomized controlled trial from Johns Hopkins demonstrated that a fully automated AI lifestyle intervention was noninferior to human coaching across weight loss, physical activity, and metabolic health markers over a full year. This is the gold standard of clinical evidence — a phase 3 RCT published in the most authoritative medical journal — and it validates the entire category HeatherAI occupies.
Mathioudakis et al. (2025). JAMA, 334(23):2079-2089. doi:10.1001/jama.2025.19563
Methodology-grounded AI significantly outperforms generic LLMs.
A study from Google's Verily life sciences team found that multi-agent AI workflows grounded in behavioral science frameworks were strongly preferred by expert evaluators over baseline LLM coaching. The key differentiator: specialized systems that identify root causes of dietary barriers before offering strategies. This directly validates HeatherAI's architecture — Heather's methodology embedded in the coaching logic, not a generic nutrition prompt.
Yang et al. (2025). JMIR Formative Research, 9:e75421. doi:10.2196/75421
Hybrid AI + human coaching retains 57–92% of users.
A systematic review of 21 studies including 8 RCTs found that hybrid AI + human coaching platforms consistently achieve the highest retention rates in digital health — ranging from 57% to 92%. This supports HeatherAI's expert-in-the-loop design, and the AI + Human Coach tier, as both clinically and commercially optimal.
Dunmore et al. (2025). Current Cardiovascular Risk Reports (Springer). doi:10.1007/s12170-025-00776-7
Personalized AI coaching increased engagement by 26%.
A 3-month longitudinal study of an AI-powered coaching assistant found that personalized interactions increased user engagement by 26% over non-personalized ones, and that conversational style increased average message length by more than 4×. Longer, more detailed messages mean richer context for the AI, better memory, and a stronger coaching relationship — the engagement flywheel HeatherAI is designed to create.
Garcia-Zapirain et al. (2025). Journal of Healthcare Informatics Research (Springer). doi:10.1007/s41666-025-00217-5
What the research doesn't yet prove.
RCT evidence specifically for NLP chatbots improving diet behavior is still thin — small samples, short durations. That's a reason to be precise about claims, and it's why HeatherAI is building its own outcome data alongside the published literature.
Chen et al. (2025). JMIR mHealth and uHealth, 13:e66403. doi:10.2196/66403
Want the full picture?
Investors, journalists, and partners — we're happy to walk you through the corpus, the architecture, and the roadmap.
Contact the team