Fan-Engagement & Conversational-Support AI Agent

What this agent does

Answers the question, not the incident

Ingesting fan messages alongside the matchday pack, venue map and ticket records the club configures as sources.

Normalising fixture, gate, block and seat references, and carrying each answer forward with its source.

Reason

Answering matchday, wayfinding and account questions from those configured sources.

Identifying itself as an agent when a conversation opens, and handing over on request.

Applying the accessibility regime the venue is under, public or private.

Decide

Routing accessible-seating enquiries onto the venue's dedicated seating path.

Handing a medical or safeguarding report to a named person at once, untriaged.

Out

Retaining the question, the source it was answered from, the disclosure and the handover per conversation.

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

→ Product statement

The agent answers and signposts; a named steward or box-office lead decides anything touching money, a seat or a safety response.

Example workflow

One question, arrival to handover

AgentHuman

1Question receivedMatchday chat, club app, web widget, social message or venue kiosk

2Sources assembledMatchday pack, venue map, fixture data and the ticket record on file, each with its source

3Answer preparedAnswer, wayfinding steps, flagged lines and confidence

4Controls appliedSource checks, accessible-seating routing, safety-intent detection, disclosure check and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is refunded, transferred or allocated — the agent is answering, and the steward's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold

High confidence

Answers the fan and offers a person.

Low confidence

Adds a steward read first.

Steward handover

The conversation is held with its sources, its flagged lines and the confidence.

Approve · Amend · Hand to the box office

Approved — sent to the fan▼

6Support systems updatedOnly where write access and approval policy allow it

7Outcome evaluatedAnswer accuracy, handover reasons, repeat questions and corrections made after the fixture

Amendments

Every steward amendment is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Refunding, crediting or repricing a ticket.

Transferring, reissuing or voiding a ticket.

Allocating or releasing accessible seating.

A medical, crowd-safety or safeguarding report.

Automation boundaryAgent acts unaided

✓Answer matchday and venue questions from the configured sources.

✓Give wayfinding and gate directions from the venue map.

✓Look up a ticket or account where the venue has connected one.

✓Say it is an agent, and hand over to a person on request.

Any write happens inside the boundaries agreed at implementation, never ahead of a person's decision.

Ejection, ban and disputed-entry decisions.

Quoting a price or fee not on the published card.

Committing the club to compensation or goodwill.

Changes to source, escalation or disclosure rules.

Example output

One conversation, annotated

Everything the agent answers is attached to the source it was answered from.

Support output · single conversationIllustrative example

Conversation

Answer given

Gate and block

Source of record

Confidence

Disclosure

Matchday wayfinding

Step-free route to Block 118 from Gate C, lift on the left

Gate C · Block 118

Access map on file

93%

Agent identified at open

As receivedTaken from the venue access map and the fixture pack — nothing on this side is written by the agent.

Sources used Venue access map Matchday operations pack Ticket record

Why this answerIt repeats the route the venue published and offers a person — the line the steward weighs.

ActionApproveAmendHand to the box office

What the score decidesBelow the configured threshold the answer picks up a steward read before it reaches the fan.

Value

Where AI adds value

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

Where the value landsValue 01 – 04

Every questionFrom the fan's channel

03Answering

Answer from the sources

Draw on the matchday pack, the venue map and the records the club configured.

01Approved path

Answer the concourse queue

Routine matchday and wayfinding questions are answered as they arrive.

02Human review

Send the rest to a person

Safety, money and accessible-seating intents go straight to the named person on duty instead.

04Build an evidence trail

The question, the source it answered from and the steward who took it on stay together.

Integrations

Typical integrations

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

Ticketing and accessTicketmaster · SeatGeek
AXS · Paciolan · Tixr

Fan channelsClub app · web chat
WhatsApp · SMS · social DM

Venue operationsSteward and incident systems
Matchday control room

Agent

Fan engagement and support

Reads the sources
Answers the question
Hands to a person

Service desk and CRMZendesk · Salesforce
Freshdesk · HubSpot

Observability & evaluationOpenTelemetry · Langfuse
Supported monitoring/evaluation sources

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

Agent controls

Six layers between the model and the fan

Every control encloses the next. What survives the set is named in the map below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modePull the agent back to signposting-only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, source-pack and escalation-rule changes.Track

L4TraceabilityRecord the question, the source, the disclosure event and the handover time.Record

L3Steward handoverHold safety, money and seating intents for a named person; it governs who answers, not what is right.Gate

L2Policy guardrailsTest answers against configured sources, disclosure and seating-routing rules; a failure holds the answer.Restrict

L1Confidence thresholdsRoute low-confidence answers to a steward read before the fan sees them.Require review

Model coreAnswer produced — reply, wayfinding steps, flagged lines and confidence

L1 – L2Test whether an answer may go

L3Puts the conversation in a steward's hands

L4 – L5Keep the answer and the source behind it

L6Falls back to signposting when signals degrade

How Nestack evaluates it

Evaluate the whole conversation — not only the sentence that closed it.

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

Surface — the answer the fan reads

Depth of coverage ▼

E1Final-output evaluationDid the answer match the source it was drawn from?

E2Step-level evaluationDid the agent read the right fixture, venue map and record?

E3Tool evaluationDid a lookup hit the correct ticket and the correct account?

E4Confidence calibrationDo low-confidence answers actually attract more steward amendments?

E5Slice evaluationHow does performance change across specific question types?

E6Business outcomeHow many conversations needed a steward or a second attempt?

Floor — the exposure the venue carries

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed where it first appears.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

FE-03

Superseded matchday source

Gate or access detail read from an earlier fixture's pack.

Stage gathersFan message, matchday pack, venue map and ticket record

02 · Reasoning2 modes

FE-04

Accessible seat read as inventory

An accessible-seating enquiry is answered on the general ticket path.

FE-06

Safety intent read as a query

An incident report is answered instead of handed to a person.

Stage proposesAnswer, wayfinding steps, flagged lines and confidence

03 · Tool / write2 modes

FE-02

Answer beyond the sources

The agent asserts what the configured pack does not carry.

FE-05

Reply in the wrong thread

An answer lands on another fan's conversation.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

FE-01

Disclosure missed at the open

A conversation starts without the agent naming itself as an agent.

Stage returnsThe answer the fan reads and the venue answers for

05 · Change / Version1 mode

FE-07

Silent escalation regression

A model or rule change narrows what the agent hands over.

Stage tracksModel, prompt, source pack and escalation rules

Sev-1 · acts outside the escalation boundary Sev-2 · a wrong answer reaches the fan Sev-3 · source degrades, answer routes to a steward

See the complete Sports & Fitness failure-mode catalogue →

Affected slices

The total is calm; one question type is not

Across every conversation the steward-amendment rate looks settled; inside accessible-seating enquiries it runs several times that. Nestack reports the rate by slice, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Accessible-seating enquiries	4.0%	3.8×	Review
Refund and transfer questions	2.8%	2.7×	Review
Away-end and visiting fans	2.0%	1.9×	Watch
Routine wayfinding questions	0.7%	0.7×	Normal

Bar: steward-amendment-rate lift vs. routine-wayfinding baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

Nothing closes until a case exists

The loop closes on a case the next release must pass, not on an explanation. That suite is what the next answer given on a matchday is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Steward amendments rise in one question type.

02Diagnose

The steward who took the conversation over reads it back until one cause is left.

03Improve

The fix is versioned with the conversations that motivated it.

04Verify

Nothing ships while an affected case is failing.

05Learn

The case is added for good, and the matchday playbook changes.

Learn → DetectThe return edge. Every following fixture is measured against a longer suite.

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, conversation workflow, evaluation, integration, then production validation and handover.

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

01Support workflow discovery and boundary definition.

02Ticketing and venue-source assessment.

03Disclosure, escalation and accessible-seating mapping.

04Source ingestion and matchday-pack.

05Answering logic and source binding.

06Confidence scoring and handover routing.

07Steward handover workflow.

08Ticketing and channel integration.

09Answer-accuracy cases.

10Guardrails and escalation controls.

11Conversation-trail instrumentation.

12Deployment, documentation and Agent Care handover.

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 tier PilotOne venue, one channel ProductionProduction fan channels AdvancedMultiple venues / clubs

Introduced at Pilot

Answers from your configured sources✓✓✓

Steward handover✓✓✓

Answer-quality baseline✓✓✓

Introduced at Production

Reporting by matchday and channel—✓✓

Handover workflow in your systems—✓✓

Approved write-back—✓✓

Ticketing-account integration—✓✓

Introduced at Advanced

Multi-competition rule sets——✓

Multi-stage venue approvals——✓

High matchday volume——✓

Multi-venue support controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.

Separate from buildBuild 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.

You bringWe build with it

01Your matchday sources and the questions fans actually ask → Source ingestion and question mappingWeek 1

02Representative answered conversations → Answering baseline, source binding and disclosure wordingWeek 2

03Which accessibility regime your venue is under → Disclosure, escalation and accessible-seating mappingWeek 1

04Access to relevant APIs, feeds or exports → Ticketing, channel and venue-source assessment, then integration setupWeek 2

05Answers you would not want repeated → Accuracy cases and the evaluation suiteWeek 4

06Where an answer must hand over to a steward → Confidence scoring, handover routing, guardrails and escalation controlsWeek 3

07Named stewards and box-office staff to take handovers → Steward handover workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each phase occupies the weeks it needs, so the fifth week carries two kinds of work.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Support workflow discovery, escalation mapping and the automation boundary W2Channel integration and the answering baseline W3Conversation workflow, confidence logic and handover controls W4Evaluation suite, disclosure checks and failure-mode testing W5Ticketing integration, pilot fixtures and targeted corrections W6A run of fixtures supported under the matchday team, then handover

Reading the bandA band covers the weeks its work is named in, and no others. The week 5 overlap is real, not padding.

At the end of W6The fixtures close validation and Agent Care assumes monitoring.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Sports & Fitness AI agent

Build a fan-engagement agent around the people who answer on matchday.

Show us your matchday sources, your channels and who takes a handover. Get a matchday answer wrong and it is repeated to thousands of fans inside the hour, so we set the boundary first.

Discuss this agent →

Nestack Agents · Fan engagement and conversational supportAGT-SF-12 · Agent Care available after launch