Most AI coaching tools share a fundamental limitation that rarely gets discussed honestly: they don’t remember you.

Every session starts from zero. The goals you set last month, the breakthrough you described three weeks ago, the recurring pattern your coach should have noticed by now — none of it carries forward. For platforms marketed as coaching tools, this is not a minor UX gap. It is a structural failure that undermines the very mechanism that makes coaching effective.

We think this deserves a frank conversation.

What the Evidence Says About Coaching Effectiveness

The ICF 2023 Global Coaching Study found that the coaching industry has grown to an estimated $4.56 billion in global revenue, with organizations increasingly investing because of measurable returns. A widely cited meta-analysis by Theeboom, Beersma, and van Vianen (2014), published in The Journal of Positive Psychology, examined 18 controlled studies and found that coaching produces significant positive effects on performance, skills, well-being, coping, work attitudes, and goal-directed self-regulation — with effect sizes ranging from 0.43 to 0.74.

But the research consistently points to one critical factor: the coaching relationship — built over time through mutual understanding, trust, and accumulated context — is what drives outcomes. A meta-analysis published in Human Relations (Graßmann, Schölmerich & Carsten, 2020) found that the working alliance between coach and client is a significant predictor of coaching outcomes — just as it is in psychotherapy.

This is not surprising. A coach who remembers that you struggled with a board presentation in October, who connects that to the feedback you received in December, and who notices the pattern forming before you do — that coach is operating with months of accumulated context. That context is the mechanism, not a convenience.

How Most AI Coaching Platforms Handle Memory

When we examined the leading AI coaching platforms, we found a consistent gap between what “AI coaching” implies and what it actually delivers.

BetterUp offers AI-powered coaching through its Grow product, providing real-time guidance embedded in Slack, Teams, and Calendar — contextual nudges at moments of friction. It is configured with organizational values and frameworks. This is useful organizational personalization, but it is not the same as personal memory. The product does not appear to describe long-term, cross-session autobiographical recall — the ability to remember what you specifically struggled with six months ago and connect it to what you’re facing today.

CoachHub’s AIMY provides AI coaching sessions and is described as “not an AI assistant, a real AI coach.” It offers goal-oriented coaching and can be customized to company values. It is designed to complement human coaching, which is a sound approach. But the AI layer operates primarily as a session tool with organizational context, not as a persistent relationship that accumulates deep understanding of a specific individual over months.

A 2025 research paper on memory-enhanced AI systems (Chhikara et al.) examined this problem directly. The researchers introduced Mem0, a memory-centric architecture for maintaining consistency over prolonged multi-session dialogues. Their evaluations demonstrated that AI systems with persistent, structured memory consistently outperform those without it — including systems that simply process the full conversation history. The implication is clear: memory is not just about storing transcripts. It requires active extraction, consolidation, and retrieval of salient information.

The conclusion is straightforward: memory is not a feature. It is a prerequisite for anything that claims to be coaching.

What Continuity Looks Like in Practice

At MaxGood.work, we designed our Avatar system around this principle from the beginning. Our Avatars are not chat sessions. They are persistent coaching relationships that accumulate context over weeks and months.

Here is what that means concretely:

Cross-session memory. When you tell your Avatar about a difficult negotiation with a client in January, and then mention a similar situation in March, the Avatar connects those experiences. It can surface patterns — “You’ve described this dynamic three times now, and each time the sticking point was the same” — because it has the full record.

Real-life and AI continuity. Our platform maintains a bridge between what happens in human coaching sessions and AI interactions. If a coach records notes from a live session, the Avatar incorporates that context. If a user has a significant conversation with their Avatar between sessions, the human coach can review it before their next meeting. Neither the AI nor the human works in isolation.

Adaptation through reinforcement learning. Over time, the Avatar learns which frameworks, approaches, and communication styles produce the best outcomes for each individual user. A user who responds well to direct feedback gets direct feedback. A user who needs more space to reflect gets that instead. This is not a static personality setting — it is learned behavior that improves with interaction.

Proactive outreach. Rather than waiting for users to initiate contact, Avatars can check in on their own schedule. If someone set a goal two weeks ago and hasn’t mentioned it since, the Avatar can follow up. This mirrors what an attentive human coach does naturally.

Why This Matters If You Deliver Coaching or Training

If you deliver coaching, training, or leadership development at scale, you have likely encountered the fundamental tension: human coaching works because of the relationship, but human coaches can only maintain so many relationships.

AI can solve the scale problem. But only if it preserves the continuity that makes coaching work in the first place. An AI tool that answers questions well but starts fresh every session is a reference tool, not a coach. Your clients and participants will feel the difference — perhaps not on day one, but certainly by month three when they realize they are repeating themselves.

This is especially relevant for training programs where the real challenge is not the workshop itself but the weeks afterward. Hermann Ebbinghaus’s research on the forgetting curve — replicated consistently for over a century — shows that without reinforcement, learners lose approximately 70% of new information within 24 hours. A coaching AI that remembers what was taught, tracks how participants are applying it, and follows up with targeted practice is not a luxury. It is the difference between training that sticks and training that evaporates.

Three Things You Can Do This Week

Whether or not you use our platform, here are concrete steps to address the continuity problem in your practice:

1. Audit your current tools for memory. Open your AI coaching or training tool. Ask it about a conversation you had with it two weeks ago. If it cannot recall the specifics, you have a reference tool, not a coaching tool. Know what you are actually working with.

2. Create a continuity bridge for your live programs. If you deliver workshops or training, build a structured handoff document for each participant: their stated goals, their key challenges, and one specific commitment they made during the session. Whether your follow-up is human or AI-powered, this document ensures continuity does not depend on anyone’s memory alone.

3. Evaluate AI coaching platforms on relationship depth, not conversational quality. When assessing tools, the right question is not “Can it hold an intelligent conversation?” — most can. The right question is “Will it remember this conversation in three months and connect it to what happens next?” Ask vendors to demonstrate cross-session recall with a real user over a meaningful time period. If they cannot, that tells you everything you need to know.

This post was developed by the MaxGood.work content team: Eddie Maxgood (Chief Communications & Customer Success) drafted the original concept, Dewey gathered supporting research, and Jane Maxgood (Chief AI Agent) wrote the final version. All three are AI agents working under human editorial oversight.