SAMI Golf AI vs GOATY — Agentic AI Coaching, Compared

In April 2026, Bryson DeChambeau led an eight-figure acquisition of Sportsbox AI and announced SAMI — the “next generation of agentic AI coaching.” The announcement generated enormous buzz in the golf world. Sportsbox has genuine 3D biomechanics credentials, and DeChambeau’s involvement brings serious visibility to the concept of AI coaching in golf.

But what is agentic AI coaching, exactly? And is there already a system doing it right now that golfers can use today? This article answers both questions honestly.

What Is Agentic AI Coaching?

The word “agentic” in AI means the system acts proactively rather than reactively. A standard AI tool waits for you to ask a question and then answers it. An agentic AI system observes, decides what action is needed, takes that action, and verifies the result — all without requiring you to prompt it at each step.

Applied to golf coaching, agentic AI means:

• Observes your swing in real time — not just after you upload a video, but as it happens
• Diagnoses the fault — identifies which specific mechanical gate is failing
• Selects the coaching intervention — chooses the cue most likely to help based on your history and population-level learning data
• Delivers the coaching — speaks it to you before your next rep
• Verifies the outcome — checks whether the advice produced measurable improvement
• Feeds outcomes back into the system — so the coaching gets better over time

This is the full agentic loop. SAMI has announced the intent to deliver it. GOATY has been running this loop since 2025.

What SAMI Is (and Is Not Yet)

SAMI is Sportsbox AI’s planned agentic coaching assistant, powered by Google’s Gemini AI models. The architecture: Sportsbox’s patented 3D motion capture technology extracts biomechanical data from your swing, and SAMI interprets that data and delivers personalized coaching recommendations in natural language.

The technology stack is credible. Sportsbox’s 3D reconstruction from 2D video is genuinely sophisticated, and Google Cloud’s AI infrastructure is world-class. The vision — a proactive AI agent that translates biomechanical measurements into actionable coaching — is exactly right.

What GOATY Is Doing Right Now

GOATY has been operating the full agentic coaching loop since 2025. Here is what the live system actually does:

1. Real-time swing observation — GOATY watches your swing through your phone camera using server-side MediaPipe pose detection at 30fps, with body position tracked across 33 landmark points
2. Gate-by-gate diagnosis — every rep is evaluated against 7 biomechanical gates (ENGINE, ANCHOR, WHIP categories) and a GOAT Score from 0–100 is computed per rep
3. Cue selection from population learning — a contextual bandit system ranks coaching cues by their verified effectiveness for your specific gate failure, skill tier, and personal history
4. Voice delivery between reps — the selected cue is spoken through your phone in the window between swings, keeping you in your practice routine
5. Outcome verification — the Recursive Self-Improvement system tracks whether each recommendation produced improvement in subsequent swing data
6. Weekly coaching updates — the system prompt is updated every Sunday with proven insights from the week’s outcome data, making the coaching smarter for everyone

These are not projections or demos. They are production numbers from the live system, updated continuously.

Feature Comparison

Why “Agentic” Is the Right Word — And Why It Matters

The golf AI market has been dominated by analysis tools: apps that record your swing, measure things about it, and present data on a screen. These tools are useful but they are not coaches. They describe. They do not act.

The shift to agentic AI coaching is the shift from description to intervention. Instead of “your hip rotation measured 42°,” an agentic coach says: “Your trail hip slid instead of coiling — feel the socket stay fixed while the body winds around it. Try again.” And then it watches the next swing to see if the coaching worked.

This is why SAMI’s announcement matters — it signals that the broader golf tech industry is recognizing that measurement without coaching is incomplete. The agentic model is the right direction. The question is whether you want to wait for a system that does not exist yet, or use one that has been running it live for a year.

What Sportsbox’s 3D Advantage Actually Means

To be fair to SAMI, Sportsbox does something GOATY does not: it reconstructs a full 3D body model from 2D video. This means it can measure joint angles, rotation velocities, and kinematic sequencing in three dimensions — not just the 2D landmark positions that phone-based pose detection produces.

For biomechanics researchers and instructors who want raw 3D data, this is a genuine advantage. The question is whether that additional dimensional precision changes the coaching outcomes for recreational golfers — and the honest answer is: we do not know yet, because SAMI has not run enough coached sessions to produce outcome data.

What we do know from GOATY’s 152,000 coached reps is that 2D pose detection at 30fps with server-side MediaPipe processing is sufficient to identify the mechanical gates that matter for coaching — trail hip coil depth, lead side stability, wrist structure, spinal separation — and to produce 1,840 verified mechanical improvements. Whether 3D data would improve those outcomes is an interesting research question. It is not an answered one.