How Coaching Mode Works

Coaching Mode is a live observation mode. Instead of following a structured sequence of steps, the worker performs freely while Atanzo AI watches and fires tips at regular intervals. The AI is not waiting for the worker to confirm each step — it monitors continuously and comments as the work unfolds.

This mode is for anyone choosing between coaching and guided steps for a task (managers, trainers) and for the worker being coached. Skip it if you just need step-by-step task guidance — see Guided Tasks Overview instead.

How It Differs from Guided Tasks

Both guided tasks and coaching mode use the same phone-based interface, but they are designed for different situations.

Guided Tasks present a defined sequence of steps. The worker confirms each step before the system advances to the next. The system controls the pace. This is the right choice when sequence and completeness matter — inspections, installations, maintenance procedures.

Coaching Mode has no fixed sequence. The worker performs the task as they would naturally, and the AI observes and comments continuously. This is the right choice when you want to improve technique, check procedure under real conditions, or review a task where the exact order of sub-steps may vary.

In short: guided tasks drive the worker through a process; coaching mode watches and advises while they work. See Guided Tasks Overview for more on the guided task approach.

Accessing Coaching Mode

In the console. Coaching experiences appear as branded entries in your dashboard's agent picker — alongside your other agents — when your company has the matching coaching module enabled. Pick one and start a session the same way you'd start any other agent. If an experience you expect to see is missing, the module isn't enabled for your account yet — ask your account administrator.

Through the API & SDK. Coaching experiences can also be launched from a partner application or integration built on the Atanzo AI API. That path is covered separately — see Coaching Presets for Developers.

What the Coach Watches For

If you picked a branded coaching experience from the agent picker, its settings are fixed and already tuned for that experience — there's nothing to configure. The rest of this section applies to a coaching session your administrator has set up manually.

What it's watching for. Every coaching session is built around a plain-language description of what the AI should evaluate — for example, "Check the worker's lifting posture and flag any lower-back rounding" or "Review the assembly sequence and note deviations from the standard procedure." The more specific this description is, the more targeted the tips will be.

How many cues per pass. Each time the AI comments, it can deliver between 1 and 5 tips at once. Fewer tips keep feedback focused on one thing; more tips are useful when you want the AI to comment on several aspects of performance in the same pass.

How often. The AI can be set to comment automatically on a timer, anywhere from every few seconds up to every 10 minutes (600 seconds). Setting the interval to 0 turns off the automatic timer — tips will only fire when someone asks for one.

When Cues Arrive

A coaching session can deliver its feedback in one of two ways:

  • Live, during the session — the worker hears or sees tips as they perform, in real time.
  • Held for the end-of-session report — the AI still observes throughout, but tips are saved and surfaced together afterward instead of interrupting the work.

Which of these applies is part of how the session was set up — check with whoever configured it if you're not sure which your worker will experience.

Example Use Cases

Form check A worker performs a series of lifts. The AI is configured to watch lifting technique and fires 2 tips every 8 seconds throughout the set. Workers receive corrections in real time without stopping to consult a trainer.

Swing analysis A worker takes repeated golf swings or tennis strokes. The AI is configured to fire 3 tips per cycle via video input, commenting on stance, contact point, and follow-through after each pass.

Procedure review A trainer watches a technician perform a maintenance task. The AI monitors the procedure against the expected steps and annotates deviations as they occur — surfacing what was skipped, out of order, or performed incorrectly — without interrupting the flow.

Session Results

At the end of every coaching session, Atanzo AI generates a session report automatically. The report covers the full session — not just a list of tips fired, but a structured assessment of performance.

A coaching session report includes:

  • Performance metrics — measurements captured continuously throughout the session (for example: pace, consistency, or technique, depending on the coaching task configured)
  • Coach's notes — a written AI summary of the session: what the worker did well, what needs work, and any patterns observed
  • Scores — rated dimensions relevant to the coaching task, each with a brief rationale

The report appears immediately when the session ends. Observers and administrators can review it in the Atanzo AI console by selecting the completed run.

See Session Reports for a full guide on reading and using session results.