Main site

Sports & Fitness AI agent · Safeguarding

Safeguarding Report Evidence AI Agent

Capture the disclosure word for word, run a separate clock to each body that must receive a report, and hand the record to the named person who reports it themselves.

Typical delivery
4–6 weeks
Deployment
Your stack
Named reporter
Record only
After launch
Agent Care

What this agent does

Builds the file, not the decision to report

In

  1. 01

    A disclosure is made, and it is held word for word, write-once, from whichever channel it arrived on.

  2. 02

    A moment of first knowledge is entered by the person who had it, and the clocks run from that moment.

Reason

  1. 03

    A clock starts for each body that must receive a report, on the length that body sets, not one shared one.

  2. 04

    A destination register is configured for your state, your sport and your governing body.

  3. 05

    A roster position is read across for training and screening currency, from the systems already holding it.

Decide

  1. 06

    A screening cadence comes from your state, your charter or your governing body, never from the federal statute.

  2. 07

    A pack goes to the named reporter, who sends it personally; nothing is smoothed, filled in or inferred.

Out

  1. 08

    A retention position runs on your own clock, with the fact of a report held apart from what was said.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent assembles the record and holds the clocks; a named individual makes the report, and a mis-started clock becomes the evidence of itself.

Example workflow

One disclosure, intake to report

  1. 1Disclosure receivedCoach, front desk, parent, athlete, a written form or a phone call
  2. 2Words heldWhat was said, word for word, and the moment the reporter names as first knowledge
  3. 3Clocks startedOne clock per destination, each on the length its own rule-maker sets
  4. 4Controls appliedWrite-once capture, destination-register checks, roster currency and destination-match confidence

No human action required

Stages 1 to 4 run unaided, and nothing is reported at any of them — the agent is assembling, and the reporter's lane opens at the destination gate.

5DecisionBranches at the destination match
Clear match

Goes to the named reporter to send.

Any doubt

Adds a safeguarding-lead read first.

Reporter review

The pack is held with the words, the clocks and the destinations.

Confirm the moment · Append to the file · Log the report made

Reported — logged by the reporter

  1. 6Roster and case file updatedOnly where write access and retention policy allow it
  2. 7Outcome evaluatedDestination completeness, clock margin, lead corrections and what a later review found

Corrections

Each safeguarding-lead correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Deciding whether a matter is reportable.
    • Rating, scoring or tiering a disclosure.
    • Transmitting a report to any destination.
    • Searching across matters, or linking by name.

  • Automation boundary

    Agent acts unaided

    • ✓Hold what was said word for word, write-once and timestamped.
    • ✓Run the clocks from the moment the reporter says they.
    • ✓Name each destination that must receive one, with its own clock.
    • ✓Show the roster position on training and screening currency.

    Nothing leaves the record except by the named reporter, inside the agreed boundaries.

    • Investigating, or resolving a matter locally.
    • Editing, tidying or summarising what was said.
    • Removing anyone from a roster or a session.
  • Changes to clocks, destinations or retention.

Example output

One disclosure, annotated

Nothing here rates a disclosure — the threshold is suspicion, and the answer is almost always report.

Intake record · single disclosure

Illustrative example
Intake
Volunteer coach, travel club
Held as
Word for word as spoken, write-once, with no edit path
First knew
21:05 local
Destinations
State agency and law enforcement
Confidence
Held unsent
Sent by
The named reporter, in person

As received

Taken from the account given and the moment the reporter names — nothing on this side is summarised.

What is held

  • The words as spoken
  • Moment of first knowledge
  • Each read, logged

Why no rating here

No score, no tier, no similar-matter panel; the named person decides.

What the score decides

Below the configured threshold the destination set gets a safeguarding-lead read.

Action

Confirm the momentAppend to the fileLog the report made

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Each disclosure

From the words as spoken

03Assembly

Work one matter, alone

No search across matters, no linking by name — that is investigation, and the Center holds it.

01Approved path

There is no triage here

A triage feature that cannot ship disabled should not ship at all, so this one was never built.

02Human review

A clock for each destination

Each runs from the moment of first knowledge — and where the federal reach is unsettled, the state clock still runs.

04

Build an evidence trail

The words as spoken, the clock they started and the person who reported stay on the record.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Club and roster systems

    • Mindbody
    • Glofox
    • ABC Fitness
    • Perfect Gym

  • Training records

    • SafeSport Trained records
    • Governing-body member portals

  • Screening providers

    • JDP
    • Sterling Volunteers
    • NCSI
    • state clearance portals

Agent

Safeguarding report evidence

  • Holds the words
  • Runs the clocks
  • Hands to the reporter

  • Case and document store

    • Write-once object store
    • Case file
    • retention policy

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six checks between the model and the record

Six checks, each assuming the last held. What defeats them all is drawn in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeHold the intake at unsent when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, clock-rule and destination-register changes.Track
L4TraceabilityRecord the words, the clocks, the destinations and each read of the file.Record
L3Reporter hand-offHand the pack to the named reporter; Texas and California both make the duty non-delegable in terms.Gate
L2Policy guardrailsTest the pack against the destination register and the write-once rule; a failure returns it.Restrict
L1Match confidenceSend a low-confidence destination set to the safeguarding lead before the reporter sees it.Require review
Model corePack assembled — the words, the clocks, the destinations and match confidence
L1 – L2
Test whether a pack may stand
L3
Puts the sending in a person's hands
L4 – L5
Keep the words and the clock behind them
L6
Holds the intake at unsent when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the pack that comes out.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the record a review reads

  1. E1Final-output evaluationDid the pack hold the words exactly as they were given?
  2. E2Step-level evaluationDid the agent use the right register, the right clocks and the current roster?
  3. E3Tool evaluationDid it read and write the correct file and the correct field?
  4. E4Confidence calibrationDo low-confidence destination sets actually attract more lead corrections?
  5. E5Slice evaluationHow does performance change across specific intake types?
  6. E6Business outcomeHow many packs needed a correction before the reporter sent them?

Floor — the outcome the organisation answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed at the stage where it begins.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • HX-03

      Stale currency record

      Training or screening shown current from an expired record.

    Stage gathers

    The words, the register, the clocks and the roster

  2. 02 · Reasoning

    2 modes

    • HX-04

      Single shared countdown

      One clock is rendered where several are running.

    • HX-06

      Clock started late

      The clock runs from the form, not from first knowledge.

    Stage proposes

    The clocks, the destinations and match confidence

  3. 03 · Tool / write

    2 modes

    • HX-02

      Destination missed

      A body that must receive a report is not listed.

    • HX-05

      Fragmented disclosure

      Three channels, three fragments, no starting clock.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • HX-01

      Drafted, never sent

      Nothing distinguishes a pack drafted from a report made.

    Stage returns

    The pack the reporter sends and a review reads

  5. 05 · Change / Version

    1 mode

    • HX-07

      Silent clock regression

      A rule change quietly lengthens or drops a clock.

    Stage tracks

    Model, prompt, clock rules and the register

  • Sev-1 · a report is late or never sent
  • Sev-2 · a mis-started clock reaches the file
  • Sev-3 · source degrades, pack holds unsent
See the complete Sports & Fitness failure-mode catalogue

Affected slices

One intake type absorbs the corrections

A programme-level verbatim-capture figure can read clean while one intake type absorbs most of the corrections. Nestack reports the lead-correction rate by intake type, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Disclosures across channels8.8%3.7×Review
Multi-body destination sets6.4%2.7×Review
Third-party or anonymous intake4.0%1.7×Watch
Single-destination intakes1.8%0.8×Normal

Bar: lead-correction-rate lift vs. single-destination baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What a missed clock leaves behind

A cycle closes when the late notification is a regression case. That suite is what the next intake held is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Lead-correction rate rises in one intake type.

  2. 02

    Diagnose

    The conversation written up from memory the next morning is read back until one cause is left.

  3. 03

    Improve

    The fix carries a number, and the intakes that forced it travel with it.

  4. 04

    Verify

    Nothing goes out until every touched intake case has passed again.

  5. 05

    Learn

    It is kept permanently, and the clock rules are amended alongside it.

  6. Learn → Detect

    The return edge. The next intake is measured against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, assembly workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Intake workflow discovery and boundary definition.Week 1
  2. 02Roster, training and screening sources.Week 1, Week 2
  3. 03Destination register, clocks and retention mapping.Week 1, Week 2
  4. 04Intake capture and write-once storage.Week 2
  5. 05Clock logic and first-knowledge binding.Week 2, Week 3
  6. 06Destination matching and lead routing.Week 3
  7. 07Reporter hand-off workflow.Week 3, Week 4
  8. 08Roster and case-store integration.Week 5
  9. 09Clock and destination cases.Week 4, Week 5
  10. 10Guardrails and refusal controls.Week 4
  11. 11Intake-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne site, one governing bodyProductionProduction roster systemsAdvancedMultiple bodies / states
Introduced at Pilot
Capture and clocks to your rules✓✓✓
Named-reporter hand-off✓✓✓
Verbatim-capture baseline✓✓✓
Introduced at Production
Reporting by destination—✓✓
Hand-off workflow in your systems—✓✓
Approved write-back—✓✓
Roster-and-training integration—✓✓
Introduced at Advanced
Multi-state clock rules——✓
Cross-body destination sets——✓
High intake volume——✓
Multi-body notification controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your intake channels and record structure

    Intake capture and write-once storage

    Week 1

  2. 02Representative past intakes, redacted

    Capture baseline, clock binding and destination matching

    Week 2

  3. 03Your destinations, by state and governing body

    Destination register, clock rules and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Roster, training and screening assessment, then integration setup

    Week 2

  5. 05Records you would not want edited

    Destination cases and failure testing

    Week 4

  6. 06What no intake may ever become

    Destination matching, lead routing, guardrails and refusal controls

    Week 3

  7. 07Named reporters and a safeguarding lead

    Reporter hand-off workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

The chart is drawn from the weeks actually spent, which is the reason the fifth one is doubled.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Intake discovery, destination mapping and the automation boundary

W2Source integration and the capture baseline

W3Assembly workflow, clock logic and hand-off controls

W4Evaluation suite, refusal checks and failure-mode testing

W5Roster integration, pilot intakes and targeted corrections

W6One season run under the named reporter, then Agent Care handover

Reading the band
A bar spans the weeks its own work is named in and no others. The doubled fifth week is real work, not padding.
At the end of W6
The final checks clear on live intakes and monitoring moves to Agent Care.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Sports & Fitness agent

Build a safeguarding-report agent around the person who has to report.

Show us how a disclosure reaches you today and who would have to report it. The record has to survive a state agency, a governing body and a proceeding reading it back, word for word, years later.

Typical delivery
4–6 weeks
Deployment
Your roster systems
Named reporter
Record only
After launch
Agent Care available

Nestack Agents · Safeguarding report evidence

AGT-SF-17 · Agent Care available after launch

Build a safeguarding-report agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your programme

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.