CART captioning for live training events: when to use real-time human captions instead of ASR, how to procure CART services, and ADA accommodation workflow

What CART is and how it differs from ASR live captions

Communication Access Real-Time Translation (CART) is a service in which a human stenographer — using a steno machine and Computer-Aided Translation (CAT) software — transcribes spoken language into text in near real time, typically with a latency of one to four seconds from speech to caption display. CART was developed as an ADA accommodation for Deaf and hard-of-hearing individuals and is widely used in legal proceedings, university courses, medical appointments, and professional conferences. In the training context, CART is the accommodation response to a specific employee’s request for accessible live training.

The distinction from ASR (Automatic Speech Recognition) is fundamental: CART is human-produced captioning delivered in real time, while ASR captions (Zoom’s live transcription, Microsoft Teams auto-captions, Google Meet captions) are machine-produced in real time. The practical consequence is accuracy: CART captionists routinely deliver 97–99%+ accuracy because they are trained professionals who prepare with advance materials, adapt in real time to speaker voice, accent, and vocabulary, and use specialised stenography techniques to maintain pace. ASR systems deliver 85–93% accuracy on clear, general-content speech and drop to 73–85% on technical, medical, or domain-specific content where proper nouns, acronyms, and specialised vocabulary are concentrated.

This accuracy gap matters in three specific ways for live training:

First, at 85–90% accuracy, a Deaf or hard-of-hearing attendee following training exclusively via captions encounters errors at a rate that breaks comprehension. On a 30-minute training session containing 3,000 words, 90% accuracy means 300 incorrect words. Those 300 errors are not uniformly distributed: they cluster on the proper nouns, product names, regulatory citations, and technical terms that carry the most semantic weight in the training. An employee following a compliance training module via captions that read “sox violations” instead of “SOX violations,” “hipaa safeguards” instead of “HIPAA safeguards,” or “the energize” instead of “de-energize” is not receiving the training content accurately.

Second, live events have no post-production correction window. When errors appear in recorded training video captions, the process for remediation is to correct the SRT/VTT file and re-upload to the LMS. The employee who reviewed the faulty captions has already consumed incorrect information, but future viewers receive the corrected version. In a live training event, there is no correction window: the CART or ASR output appears in real time and cannot be revised before the employee reads it. Whatever accuracy the caption source provides is the accuracy the employee experiences.

Third, the employee’s effective participation depends on the caption quality at the moment questions are being asked, concepts are being demonstrated, and comprehension-checks are occurring. A training facilitator who pauses to ask “any questions on the LOTO de-energisation protocol?” and receives no response from the accommodated employee may interpret silence as understanding when the employee’s caption stream read “lotto the energisation protocol” — a phrase that conveyed nothing actionable.

Accuracy comparison: CART vs. ASR for live training

The latency column deserves specific attention: CART’s 1–4 second latency is actually longer than ASR’s 0.5–2 seconds, but this difference is irrelevant for training contexts because the content is instructional, not time-critical in the way that a real-time auction or sporting event caption would be. A 2–3 second lag on a training explanation does not materially impair comprehension; a 15% error rate on technical vocabulary does.

CART also differs from ASR in speaker adaptation: an ASR system treats every session identically, applying the same model to every speaker regardless of accent, speech pattern, or vocabulary domain. A CART captionist who works with the same training program over multiple sessions becomes familiar with the speaker’s vocabulary, pacing, and terminology — accuracy often improves over the course of a training series precisely because the captionist accumulates context that ASR cannot.

The legal trigger: individual ADA accommodation vs. general accessibility

Understanding when CART is required requires distinguishing between two legally distinct obligations under the Americans with Disabilities Act: the general accessibility obligation that applies to programs, services, and content broadly; and the individual accommodation obligation triggered when a specific person with a disability requests a specific modification to enable their participation.

General accessibility: captioned recorded content

ADA Title I (29 CFR Part 1630), which applies to employers with 15 or more employees, requires that training materials and programs be accessible to employees with disabilities as a component of equal employment opportunity. This is the obligation that drives the 99% WCAG 2.1 AA accuracy requirement for recorded training video: the captioned version of a training module must provide equivalent access to the audio content for an employee who cannot hear. This is a general program-level obligation — it applies to all recorded training content regardless of whether any specific employee has identified themselves as needing captions.

ADA Title II (28 CFR Part 35), which applies to state and local government entities including most public universities, public school districts, and government agencies, imposes a parallel obligation for general program accessibility. The ADA and WCAG compliance matrix covers the full framework for recorded training content, Section 508, WCAG 2.1 AA, and the interaction with state laws. For public universities and school districts, the Title II general accessibility obligation is why lecture capture and distance-learning content must be captioned before delivery, not after a student requests it.

Individual accommodation: CART at live events

The individual accommodation trigger is different. Under ADA Title I (42 USC § 12112(b)(5)(A)), an employer must provide a “reasonable accommodation” to the known physical or mental limitations of an otherwise qualified individual with a disability unless the accommodation would impose an undue hardship. The interactive process (29 CFR § 1630.2(o)(3)) requires the employer and the employee to engage in a good-faith dialogue to determine what accommodation is appropriate for that individual’s specific disability and needs.

The critical legal structure here is: the accommodation must be effective for the specific individual. Zoom’s built-in live transcription might be adequate for an employee who is slightly hard of hearing and follows technical content without difficulty at 88% accuracy. The same Zoom transcription might be entirely inadequate for a Deaf employee who reads captions as their primary modality for audio access and needs 97%+ accuracy to follow a compliance training session accurately. The ADA standard is effectiveness for the individual, not provision of any captioning service.

Under ADA Title I, the employer can only deny the accommodation or substitute a different accommodation if (1) the specific accommodation requested would impose an undue hardship on the organisation, or (2) an alternative accommodation would be equally effective for the employee. For most training events, CART at $100–$200/hour for a half-day training is not an undue hardship for any employer with 15+ employees — the cost of a single CART day is $400–$800, which courts have consistently found to not constitute undue hardship except for the smallest covered employers. The litigation costs of refusing a reasonable accommodation request are substantially higher.

ADA Title II and the program access standard

For Title II entities (state and local governments), the obligation is framed as “program access” (28 CFR § 35.150): the entity must ensure that its programs, services, and activities, when viewed in their entirety, are readily accessible to and usable by individuals with disabilities. In the training context, this means that a public university or state agency that provides training programs must be able to accommodate a training participant who requests CART for a live session. The Title II framework also requires that communications with participants with disabilities be “as effective as communications with others” (28 CFR § 35.160(a)), and that the entity furnish auxiliary aids and services where necessary — of which CART is an explicitly listed example.

Section 504 and federal contractor obligations

Section 504 of the Rehabilitation Act (29 USC § 794) applies to any program or activity receiving federal financial assistance — which includes federally funded training programs, workforce development, and vocational rehabilitation training. The Section 504 standard is essentially parallel to ADA Title II for the recipient programs. For federal contractors subject to Section 503 (29 USC § 793, administered by OFCCP), the affirmative action and nondiscrimination obligations similarly require accessible training.

The employer-size threshold

ADA Title I applies to employers with 15 or more employees (42 USC § 12111(5)). Below that threshold, ADA Title I does not apply — but state disability laws often do. California’s Fair Employment and Housing Act (FEHA) applies to employers with 5 or more employees. New York State’s Human Rights Law applies to employers with 4 or more employees. New Jersey’s Law Against Discrimination has no employee minimum. Most training organisations that are large enough to have an L&D function are covered by ADA Title I and at least one state law.

The key operational implication of the individual accommodation framework: CART is not a policy decision the L&D team makes for all future live training; it is a response to a specific request from a specific employee for a specific event, coordinated between HR and the training team, and documented in the accommodation record. This is why the accessibility coordinator role is operationally important: someone needs to own the workflow between an incoming accommodation request, HR, the training scheduling function, and the CART provider. Without a documented workflow, the request gets lost, the training event occurs without CART, and the employer has a Title I violation on record.

Live training event types and CART applicability

Not every live training event poses the same CART logistics challenge, and the event type determines which integration method is appropriate and what lead time is realistic.

In-person instructor-led training (ILT)

In-person ILT is the original context for which CART was developed. A credentialed CART captionist attends the training in person, positioned near the accommodated participant, using a steno machine connected to a laptop. The CAT (Computer-Aided Translation) software on the laptop receives the steno input and renders it as text in near real time. The output is displayed on a secondary monitor positioned so the participant can read it, or projected for the full room if the accommodation does not need to be discrete.

In-person ILT is the most straightforward CART scenario because audio quality is within the captionist’s direct control: they can position themselves near a speaker, request that the facilitator use a lapel microphone, and adjust in real time to acoustic conditions. The primary logistics challenge is travel: an onsite captionist in a specific city requires a provider with availability in that market. For organisations in major metropolitan areas, this is usually achievable with 2-week lead time. For remote or suburban locations, lead time extends to 3–4 weeks.

Virtual instructor-led training (vILT) via Zoom or Teams

Virtual ILT delivered via Zoom, Microsoft Teams, or Webex is where most corporate training organisations now deliver live instructional content, and where CART requests have become more common because the ASR quality of platform-native captions is both visible and objectively insufficient for technical training. The ILT and virtual classroom captioning playbook covers the general approach to captioning vILT, including how Zoom’s ASR compares to dedicated captioning tools. The CART integration layer — what happens when an employee requests CART specifically, not just better ASR — is covered in the technical integration section below.

For vILT, the CART captionist operates remotely: they join the session via computer audio or phone-in, use their steno equipment at their own location, and push output to the Zoom CC channel or a streaming URL. The audio quality is the primary concern: if the Zoom session has poor audio quality (attendees on phone audio, low bandwidth connections, background noise from participants who don’t mute), the captionist experiences the same audio degradation as any other attendee, and accuracy falls. This is one reason the vILT captioning playbook recommends audio quality standards for training facilitators regardless of whether CART or ASR is providing captions.

Hybrid events

Hybrid training — where some participants are in a physical room and others attend remotely via video conferencing — creates the most complex CART logistics because the accommodation may need to serve the in-person participant, the remote participant, or both. A typical configuration is: CART captionist joins the Zoom session remotely and pushes output to a StreamText URL; the in-room participant with the accommodation opens the StreamText URL on a tablet positioned beside the meeting display; remote participants with the same accommodation open the same URL on their own devices. This works but requires that the captionist can hear the in-room audio clearly, which depends on how well the room audio feeds into the conference system.

Large conferences, all-hands training, and seminars

Multi-hour or multi-day training events — annual all-hands meetings, compliance training days, new-hire orientation weeks — require planning for captionist scheduling across the full event duration. A single CART captionist can typically work for 20–30 minutes at a time before requiring a break; for events longer than one hour, CART agencies typically provide two captionists who rotate on 20–30 minute cycles. A full-day training event (8 hours) requires a two-captionist team, and the per-day cost increases accordingly. For multi-day events, captionist teams are usually available for the same event under a single booking.

Panel discussions and multi-speaker workshops

Training events with rapid speaker turnover — panel Q&A, workshop facilitation with group participation, scenario role-plays — are the most demanding CART context because the captionist must transition between speakers quickly and often without advance warning of who is speaking next. Providing the captionist with a speaker roster in advance (with name pronunciations noted) helps; briefing the facilitator to use call-outs (“Raj, your thoughts on this?” rather than simply pointing) helps further. In sessions where audience members ask questions verbally without microphones, the captionist may miss questions that aren’t captured clearly in the conference audio stream.

How CART integrates with Zoom, Teams, and in-person events

The technical integration method determines what the accommodated employee experiences: whether captions appear in their existing video window, whether they need to open a separate device or browser tab, and whether captions are visible to all attendees or only to the employee with the accommodation. Training coordinators who have not worked with CART before often assume the captionist simply types into Zoom’s caption bar directly — and while that is one option, it is not always the best option. Understanding the available methods before a request arrives allows the coordinator to discuss options intelligently with the employee during the interactive process.

Method 1: Zoom built-in closed captioning (CC channel)

Zoom has a built-in closed captioning feature (separate from its ASR-generated live transcription) that allows a human typist to submit captions directly into the meeting’s caption feed. The workflow is:

1. The meeting host enables “Closed Captioning” in Zoom account settings (Account Management → Account Settings → In Meeting Advanced).
2. During the meeting, the host opens the CC panel and selects “Assign someone to type” rather than “Enable auto-transcription.”
3. The host enters the CART captionist’s email address; Zoom sends them a unique token URL that opens the CC typing interface.
4. The captionist opens the token URL in a browser, receives the meeting audio via their device, and types into the browser-based CC interface. Their input appears in the Zoom caption bar visible to all attendees who have captions enabled.

Advantages: simple for the employee (no separate device needed, captions appear in their existing Zoom window), captions are available to all attendees (not just the accommodated employee), and the captionist doesn’t need to be a Zoom participant taking up a seat or appearing in the participant list.

Limitations: Zoom’s CC interface has a 1,000-character display buffer; text older than approximately 3–5 minutes is not scrollable during the session. For long training sessions, the employee cannot scroll back to re-read something they missed. Additionally, the CC display in Zoom is typically limited to 1–3 lines of text in the video window, which creates a reading flow issue for dense instructional content. The captionist must also maintain a stable internet connection to the CC token URL, separate from their audio connection to the meeting.

Method 2: StreamText real-time streaming

StreamText (streamtext.net) is a hosted service that allows CART captionists to push their CAT software output to a hosted web URL that any device can open in a browser. The captionist connects their CAT software (CaseCatalyst, Eclipse, Stenograph) to a StreamText “channel,” and all text produced by the steno machine appears on the StreamText URL in real time with a scrolling transcript view. The URL is shared with the employee before the event starts; they open it on a laptop, tablet, or phone alongside the Zoom window.

Advantages: the StreamText view provides a full scrolling transcript, allowing the employee to scroll back during the session if they need to re-read a section. The text is larger and more readable than the Zoom caption bar. StreamText also produces a post-event transcript export (plain text or RTF) that can be corrected by the captionist and delivered as the post-session documentation artifact. The URL can be shared with multiple employees simultaneously, making it appropriate for situations where more than one attendee needs CART-quality captions.

Limitations: the employee needs a second device or browser tab; if they are watching Zoom on their laptop, they either split-screen or use a tablet for StreamText. For employees with low vision who are also using screen magnification software, managing two windows simultaneously is a usability challenge. Confirm the employee’s preference before selecting this method.

Method 3: CART to Microsoft Teams

Microsoft Teams has a built-in “Assign sign language interpreter” accessibility setting, but this is designed for visual language interpreters who need to be visible in a video window, not for text CART. For text CART in Teams, the two practical options are: (1) the captionist types into the Teams meeting chat (visible to all participants, not ideal because it mixes CART output with chat messages); or (2) the captionist provides CART via StreamText URL alongside the Teams meeting. StreamText is generally the preferred method for Teams CART because Teams does not have an equivalent to Zoom’s human-typist CC channel. Some third-party Teams CART integration tools exist (Caption Access, MyClearText), but they are platform-specific and require advance configuration. Confirm with the CART provider which method they support for Teams-based sessions.

Method 4: In-person display

For in-person ILT, the captionist’s CAT software output is displayed on a secondary monitor positioned near the accommodated participant. Typical setups:

• Private display: A laptop or tablet beside the participant’s seat, showing the captionist’s output via a local connection or the StreamText URL. This is discrete and does not draw attention to the accommodation.
• Room display: CART output projected onto a second screen or section of the main display, visible to all attendees. This is common in conference and seminar settings where general audience accessibility is also desired.
• Wireless streaming to participant device: The captionist connects via local Wi-Fi to the training venue network; the participant opens the StreamText URL on their own device. This is the most flexible setup because the participant can position their device as needed.

For in-person events, confirm the venue has adequate Wi-Fi if the streaming method is planned. In facilities with restricted network access (secure government facilities, medical research campuses), Wi-Fi-based CART streaming may not be possible, and a direct HDMI connection from the captionist’s laptop to a display may be the only option. Discuss venue constraints with the CART provider during booking.

Audio feed to the captionist

Regardless of the output method, the captionist needs high-quality audio of all speakers. For vILT, this means the session’s audio quality standards apply to the captionist the same as to any attendee. For in-person events, the captionist should be positioned where they can hear directly and should have access to a PA feed or digital audio mixer output if one is available. Facilitators using lapel microphones significantly improve captionist accuracy because the microphone is close to the speaker’s mouth and reduces room echo. For panel discussions, individual microphones for each panelist are strongly recommended.

Finding and procuring CART providers

CART providers are not the same as professional captioning vendors like Rev, 3Play Media, or Verbit. Those services produce captions for recorded video content using a combination of ASR and human review. CART providers are human stenographers — trained court reporters or CART-specialist captionists — who deliver real-time human captions for live events. The procurement process, credentials, and contracting are accordingly different.

Where to find CART providers

NCRA ProLink directory: The National Court Reporters Association (NCRA) maintains a member directory at ncra.org with a provider-search function. Filtering for “CART” as a specialty within the directory returns credentialed practitioners. The NCRA directory is the most reliable starting point because it covers the full United States and many international markets, and lists the specific certifications each practitioner holds.

CART agencies: National CART agencies staff and schedule captionists and handle the logistics of booking, backup coverage, and advance materials distribution. Working with an agency rather than a freelance captionist provides more scheduling reliability (if the primary captionist is unavailable, the agency provides a backup) and typically a single point of contact for billing. Examples of national CART agencies include CaptionFirst, Aberdeen Caption Bureau, and Advantage Communications Group; regional agencies operate in most major metropolitan markets. For organisations with recurring CART needs across multiple training programs, an agency relationship is generally more sustainable than managing a roster of individual freelance captionists.

Vocational rehabilitation provider lists: State vocational rehabilitation offices maintain lists of approved accommodation vendors, including CART providers, for individuals receiving VR services. If your employee is a VR client, their case manager may have an approved provider relationship that covers CART cost under the VR plan. Confirming this early in the interactive process can eliminate the employer’s cost for the specific accommodation.

Registry of Interpreters for the Deaf (RID): RID (rid.org) is the primary credentialing organisation for signed language interpreters, but some of its members are dual-certified as both sign language interpreters and CART captionists. RID’s member search can surface captionists who work in both modalities, which is useful for organisations that may also need sign language interpretation for some training events.

Credentials to look for

The NCRA’s certification structure for realtime captioning:

For training content with specialised vocabulary (medical, legal, financial, engineering), ask the provider about domain experience specifically. A CRC with 10 years of experience captioning court depositions in patent cases will handle technical engineering vocabulary more accurately than a CRC who primarily captions general-audience conferences. The credential tells you baseline competency; the domain experience tells you performance on your specific content type.

What to evaluate in a CART procurement

When engaging a CART provider for a training programme (rather than a one-off event), run a structured evaluation covering:

• Accuracy guarantee: What is the captionist’s stated accuracy on content similar to your training? The baseline for CART should be 97%+ on general content and 96%+ on technical content with adequate advance materials. Ask for sample output from similar events.
• Post-event deliverables: Does the provider deliver a corrected transcript after the event? What is the turnaround time? What format (plain text, RTF, SRT)? For training documentation purposes, a corrected transcript is more useful than the draft CART output.
• Advance materials process: How does the captionist handle advance materials? Is there a materials-intake process? What lead time do they need? Is prep time billed separately?
• Backup coverage: For recurring training programs, what happens if the scheduled captionist is unavailable? Does the provider guarantee a replacement, or does the training coordinator need to find backup independently?
• Remote delivery capability: Does the captionist support Zoom CC token integration? StreamText? Teams? What is their preferred output method for virtual events?
• Specialized vocabulary support: Can the captionist use a custom dictionary in their CAT software? Most professional CAT software (CaseCatalyst, Eclipse) supports custom dictionaries; the captionist entering your product names, acronyms, and regulatory citations into their dictionary before the event is the mechanism that produces near-100% accuracy on those specific terms.

For ongoing CART relationships, the caption vendor SLA and contract review checklist covers the contractual provisions that matter most — accuracy guarantees, delivery obligations, substitution procedures, and data handling. While that checklist was written for recorded-content captioning vendors, the provisions for accuracy measurement, corrected-file delivery, and glossary ownership apply equally to CART service agreements.

Rate structures and cost planning

CART rates vary by geography, provider type (agency vs. freelance), event type, content domain specialisation, advance notice available, and session length. Understanding the rate structure in advance is essential for accurate cost planning in the accommodation budget and for the undue hardship assessment that determines whether CART is a required accommodation for your organisation.

Typical rate ranges

Total cost examples for common training scenarios

2-hour virtual compliance training (remote CART): Remote CART at $120/hour × 2 hours = $240. One captionist. Post-event transcript included. Total: $240.

Full-day in-person new-hire orientation (8 hours, two captionists): Onsite CART at $150/hour × 8 hours × 2 captionists = $2,400 plus travel. A realistic all-in cost for a full-day ILT in a major metropolitan market is $2,500–$3,000.

Quarterly training series (four 3-hour virtual sessions per year): Remote CART at $120/hour × 3 hours × 4 sessions = $1,440/year. At volume, an agency retainer arrangement often reduces this to $1,000–$1,200/year for a recurring quarterly booking.

The undue hardship calculation

Under ADA Title I, an employer can deny a requested accommodation if it would impose an “undue hardship” — defined as a significant difficulty or expense (42 USC § 12111(10)). The undue hardship factors include the cost of the accommodation, the financial resources of the facility and the overall organisation, the type of operation, and the impact of the accommodation on operations. For most employers with 15+ employees delivering corporate training, CART at $240–$800 per event is not an undue hardship. Courts have found undue hardship only in narrow circumstances involving very small employers with extremely limited financial resources and ongoing daily CART needs. A quarterly training session, an annual all-hands, or a monthly onboarding programme does not approach undue hardship for most covered employers.

For the caption programme budget planning guide’s framework for building an accommodation line item, the CART cost should be estimated based on the number of live training events per year, the expected frequency of accommodation requests (which depends on your workforce composition and the nature of your training programme), and a per-event rate appropriate to your markets.

Comparison with alternatives

CART is more expensive per hour than ASR caption services, but the comparison is not like-for-like. Zoom’s built-in live transcription is included in the subscription but delivers 73–85% accuracy on technical content — insufficient for an employee with an active CART accommodation. Rev’s live captioning service costs approximately $1.50–$2.25/minute (=$90–$135/hour) and uses ASR with the option of human review for post-event files, but does not guarantee CART-level real-time accuracy for live events. For the purpose of satisfying an individual CART accommodation request, no ASR-based service can substitute for human CART unless the employee agrees that the alternative is equally effective for their individual needs.

The accommodation request workflow

The workflow for a CART accommodation request spans HR (which owns the ADA interactive process), the training team (which schedules events and coordinates logistics), and the CART provider (which delivers the service). Without a documented workflow that specifies who does what by when, accommodation requests get routed incorrectly, deadlines are missed, and training events occur without the accommodation in place. The workflow below covers a standard corporate training context under ADA Title I; Title II entities (universities, government agencies) have equivalent workflow requirements under their program access obligations.

Step 1: Request received

The employee submits an accommodation request to HR (typically via the HR system, an accommodation request form, or a direct communication to their HR business partner). The request may be explicit (“I need CART for the compliance training on July 15th”) or general (“I need help accessing the live training”). Either form triggers the same workflow: HR acknowledges receipt, logs the request with a date-stamp, and begins the interactive process.

Target acknowledgment timeline: 3–5 business days from receipt. For requests where a specific training event is approaching and the lead time is already tight, acknowledgment should be same-day or next-day.

Step 2: Interactive process

HR schedules a conversation with the employee to discuss their disability-related needs, the training events for which they need accommodation, and the available accommodation options (CART, alternative modalities, advance written materials, etc.). The interactive process is a good-faith dialogue; HR cannot simply unilaterally determine the accommodation without consulting the employee.

Medical documentation: the ADA permits employers to request documentation that establishes the existence of a disability and the functional limitations that necessitate accommodation. However, the employer is not entitled to the employee’s diagnosis; documentation that confirms a hearing impairment requiring real-time captioning access is sufficient. Some employees with longstanding accommodations may already have documentation on file; do not request documentation again for a previously documented disability.

Determining the appropriate accommodation: for most CART requests from employees with hearing disabilities who attend live training, CART is the appropriate accommodation. HR should confirm (1) the events requiring accommodation (specific sessions on the training calendar), (2) the preferred delivery method (Zoom CC vs. StreamText), and (3) any content-domain preferences from the employee (e.g., whether they want the captionist briefed on specific terminology the training will cover).

Step 3: Approval and notification

HR documents the accommodation determination and notifies the employee and the training coordinator simultaneously. The notification to the training coordinator should include: which training sessions require CART, the employee’s preferred delivery method, whether there are content materials the captionist should receive, and the deadline for confirming CART coverage. It should not include the employee’s diagnosis or disability documentation (accommodation details are confidential at the level of “this employee requires CART captioning” — the underlying disability is the employee’s private information).

Step 4: CART provider booking

The training coordinator contacts the CART provider (or agency), confirms the date, time, duration, and delivery method, and initiates the advance materials process. Key information to provide at booking:

• Event date, time (with timezone), and expected duration
• Number of captionists needed (one for sessions under 2 hours; two for sessions 2+ hours)
• Delivery method (Zoom CC token, StreamText URL, in-person)
• Content domain and any known vocabulary specialisation requirements
• Advance materials availability and timeline for delivery
• Whether a post-event corrected transcript is needed

Booking confirmation in writing (email or booking confirmation from the provider) is important: it establishes the record that the accommodation was arranged, which matters if a dispute arises later about whether the employer met its accommodation obligation.

Step 5: Advance materials preparation and delivery

The training team prepares and delivers advance materials to the captionist at least 24–48 hours before the event (more for specialised technical content). See the advance preparation section below for the full materials list.

Step 6: Pre-event logistics check

Within 24 hours of the event, confirm:

• Captionist has received advance materials and has reviewed them
• Zoom CC token (if applicable) has been generated and shared with the captionist
• StreamText channel (if applicable) has been set up and the URL confirmed with the captionist and shared with the employee
• Tech test scheduled (ideally 15–30 minutes before the event) to confirm audio, CC feed or StreamText URL, and display

Step 7: Event delivery and monitoring

During the event, the training coordinator or a designated logistics person monitors that the CART feed is active and the employee’s access to captions is working. If the captionist’s connection drops or the CC feed goes dark, there should be a backup plan: a direct phone number for the captionist, a procedure for restarting the CC token, and the employee’s awareness that they can flag a caption outage. For long sessions, the captionist team rotation (if two captionists) should be seamless — output continues while one captionist rests.

Step 8: Post-event follow-up

After the event: collect corrected transcript from the provider (typically within 24–48 hours), deliver to the employee if requested, file in the accommodation record. Collect feedback from the employee on the CART quality: this information improves the arrangement for future sessions. Document the accommodation in HR’s system with date, event, provider, and employee confirmation that the accommodation was provided. For recurring training calendars, establish the standing booking before the next session’s advance materials lead time begins.

The caption compliance program build post covers the programme management infrastructure that supports accommodation workflows at scale: the policy document, the coordinator role, the audit trail, and the governance structure that ensures accommodation commitments are met across a training year.

Preparing for a CART session: advance materials and logistics

Advance preparation is the single largest accuracy driver for CART sessions. A human CART captionist who receives complete advance materials 24–48 hours before a technical training event and enters the relevant vocabulary into their CAT software custom dictionary will consistently deliver 98–99%+ accuracy on that content. The same captionist, joining the session cold with no materials, will hit accuracy problems on product names, acronyms, and regulatory citations that are indistinguishable from the problems ASR has on the same content — because proper nouns and domain-specific terms are exactly where human steno has the same coverage gaps as machine ASR unless the dictionary is pre-loaded.

Advance materials checklist

Speaker deck and presenter notes: The slide deck in its current form, sent as early as possible before the event. If the presenter is modifying the deck up to the day before, send a draft early and a final version as soon as available. The captionist reads the deck to anticipate vocabulary, not to follow along live (they will be listening to the speaker, not reading slides).

Speaker roster: Full names and titles for all speakers, including correct spellings of first names that could be ambiguous (Aarav vs. Arav, Siobhan pronounced “Shuh-vawn,” Lior vs. Liore). For panel discussions, the order in which speakers will present helps the captionist anticipate transitions.

Vocabulary glossary: A list of all proper nouns, product names, acronyms, regulatory citations, and technical terms that appear in the training. Format: one term per line, with expansions for acronyms. Example list for a compliance training:

• FINRA — Financial Industry Regulatory Authority
• SOX — Sarbanes–Oxley Act
• BSA — Bank Secrecy Act
• AML — anti-money laundering
• KYC — know-your-customer
• MiFID II — Markets in Financial Instruments Directive II
• CFTC — Commodity Futures Trading Commission
• [Product names, internal system names, org unit names]

For technical engineering training, the vocabulary list might include API names, SDK symbols, CLI commands, and product internal identifiers. For medical training, drug generic and brand names, diagnostic codes, and clinical procedure names. The captionist enters these into their CAT software’s custom dictionary, typically as a “job dictionary” that applies only to this specific session, so it doesn’t conflict with their general dictionary. GlossCap’s glossary output format (term + phonetic rendering + context) is structured to work directly with CART captionist prep workflows — the same glossary that drives glossary-biased decoding for recorded content can be repurposed as CART advance materials.

Regulatory citation formats: If the training references specific regulatory sections, provide the exact citation formats. “OSHA 29 CFR 1910.147” is a specific string; a captionist who hasn’t pre-loaded CFR citation formats may render it as “29 C-F-R 1910 point 147” or miss the decimal. Same for SEC Rule 10b-5, HIPAA 45 CFR 164.312, WCAG 2.1 SC 1.2.2 — providing the exact citation strings in advance eliminates the variability.

Pronunciation guide for unusual terms: For any term whose pronunciation is non-obvious, include a phonetic note. For proper nouns from non-English languages, geographic names, or brand names with unusual pronunciation, a brief pronunciation key helps. Example: “Boehringer Ingelheim = BUR-ring-er ING-ul-heim.”

Event structure and timing: Approximate agenda with timing (10 min intro, 40 min presentation, 10 min Q&A) helps the captionist plan for speaker-transition density. If there is a structured Q&A period where audience members may speak without microphones, flag this so the captionist can prepare for lower-audio-quality content.

Technology setup checklist

Before the day of the event, confirm:

• Zoom CC token (if using Zoom CC method): host account has “Closed Captioning” enabled in settings; host knows how to open the CC panel and select “Assign someone to type”
• StreamText channel (if using StreamText): channel created in StreamText.net, captionist has the stream URL, employee has the viewer URL, both URLs tested in a non-event browser window
• Captionist’s audio source confirmed: will they use Zoom computer audio, phone dial-in, or a hybrid (phone audio + Zoom for visual)? Phone dial-in to Zoom typically provides better audio quality than computer audio for captionists because it bypasses Zoom’s audio processing
• Pre-event tech test scheduled: 15 minutes before start time, captionist connects, audio confirmed, CC or StreamText output visible to employee
• Backup contact: captionist’s direct phone number in the training coordinator’s hand during the event

For in-person events, additional checklist items:

• Display location for captionist output confirmed with venue
• Wi-Fi access confirmed for captionist (or direct network port available)
• Extension cord and power strip for captionist’s equipment
• Seating arrangement: captionist positioned where they can see and hear the speaker directly
• Microphone system: confirm lapel or handheld microphones for facilitators; request a PA feed output for the captionist if the room has a PA system

Eight failure modes

1. Conflating ASR captions with CART accommodation

The most common failure: an employee requests CART, and the training coordinator confirms that “we have live captions enabled in Zoom” as if that satisfies the request. It does not. Zoom’s auto-transcription is ASR; CART is human steno captioning. When an employee has specifically requested CART as their ADA accommodation, the employer’s obligation is to provide CART — not a substitute the employer finds more convenient. Providing ASR captions instead of requested CART may constitute a failure to provide a reasonable accommodation under Title I. The question the training coordinator needs to ask is not “do we have captions?” but “did the employee agree that ASR captions are adequate for their individual needs?” If they requested CART, the answer to that second question is no.

2. No advance materials sent to the captionist

A CART captionist who joins a technical training event without advance materials will hit vocabulary failures on proper nouns, acronyms, and regulatory citations at the same rate as ASR. The proper-noun failure problem in captioning — covered in detail in the proper noun failure modes post — is as real for human CART as for machine ASR when the captionist has no prior exposure to the vocabulary. Advance materials eliminate this vulnerability. Sending advance materials is not a favour to the captionist; it is a step in the workflow that the training coordinator owns and must complete within the agreed lead time.

3. Booking too late for specialised content

The standard lead time for CART in most markets is 2 weeks for general-content training. For training with specialised vocabulary (medical, legal, financial, engineering), the lead time extends to 3–4 weeks because captionists with domain experience are in higher demand and need more prep time. Accommodation requests that arrive 3 business days before a technical training event create a logistics crisis that may not be resolvable: a captionist may be available but unable to adequately prepare for the content. Booking too late also increases cost (rush surcharges of 25–50%) and may result in a captionist who is not the best fit for the vocabulary domain. The training calendar should have a standing rule: CART confirmation for any live event must be finalized 15 business days before the event date.

4. Wrong integration method configured at event time

For virtual CART via Zoom, there are two failure configurations: the training coordinator enables Zoom’s built-in auto-transcription (“live transcription”) instead of the human-typist CC channel, or the host never generates and shares the CC typing token with the captionist. The captionist joins the session with nowhere to type, the employee sees ASR auto-captions (the thing they requested CART instead of), and the session proceeds with the wrong captioning source. These failures are entirely preventable by testing the CC channel or StreamText connection in a practice session before the event. Do not assume the technology works; test it.

5. No post-event corrected transcript

CART draft output is 97–99% accurate but not 100%. A corrected transcript — in which the captionist reviews their draft output after the event and corrects any remaining errors — is the documentation artifact that serves as the accurate record of what was said in the training. For regulated industries (financial services, healthcare, legal), an accurate training record is a compliance documentation requirement. For any organisation, the corrected transcript is what the employee can use to review content they may have missed or want to reference later. Many CART providers include corrected transcript delivery as a standard deliverable; some bill it separately. Confirm whether a corrected transcript is expected at booking, not after the event.

6. Instructor not briefed before the event

Facilitator habits that undermine CART quality: speaking while writing on a whiteboard (captionist hears a muffled voice), reading verbatim from slides without indicating they are reading (captionist cannot tell whether to caption the spoken content or the slide text), not pausing between speakers during Q&A (captionist cannot resolve attribution or pace recovery between speakers), and speaking while other attendees are also speaking (overlapping audio is difficult even for human CART). A 5-minute facilitator briefing before the event — covering these specific habits and the presence of the captionist — significantly improves output quality. The briefing does not need to identify the employee with the accommodation; the facilitator can simply be told that a professional caption service is active for the session and be given the specific speaking habits that help.

7. Standing accommodation not converted to recurring booking

An employee receives CART accommodation approval for a monthly compliance training series. CART is booked for the first session. The session goes well. The second session comes up two weeks later, no one has re-confirmed with the CART provider, the captionist is on another assignment, and the session occurs without CART. This failure happens because the accommodation approval is a one-time HR document that does not automatically translate into a recurring calendar booking. The training coordinator must establish a standing booking with the CART provider for the full series of sessions at the time the accommodation is approved — not just for the next session. Provide the complete training calendar to the CART agency, confirm dates one by one, and build the CART booking into the same scheduling system as the training itself.

8. Tech check skipped before the event

The first technical check for CART integration should not occur while 50 training participants are already in the Zoom lobby waiting for the session to start. A 15-minute pre-event tech check — scheduled and confirmed with the captionist — allows the coordinator to verify the audio connection, the CC token or StreamText URL, and the employee’s access to the caption feed before the event is live. Common issues found in tech checks: the Zoom host forgot to enable “Closed Captioning” in account settings (requires a settings change that takes 24 hours to propagate in some configurations); the StreamText URL the captionist provided differs from the URL the employee is using; the captionist’s computer audio is feeding back through the meeting. All of these are 5-minute fixes in a pre-event check and 15-minute crises when discovered at session start.

FAQ

Does Zoom’s automatic captions feature satisfy a CART accommodation request?

No. Zoom’s live transcription feature uses ASR (Azure Cognitive Services or similar) and typically achieves 87–93% accuracy on clear, general-content speech. On technical, medical, or legal training content, accuracy drops to 73–85% — the same range documented in the LMS native auto-caption accuracy comparison. When an employee has specifically requested CART as their ADA accommodation, the employer’s obligation is to provide CART unless the employee agrees that ASR captions are equally effective for their individual needs. The test is effectiveness for the individual, not provision of any captioning. An employer who responds to a CART request by pointing to Zoom’s auto-captions has not engaged in the interactive process; they have unilaterally substituted a lower-quality alternative without the employee’s agreement. If the employee later demonstrates that ASR captions were inadequate for their access and files an ADA Title I charge with the EEOC, the employer’s position will be difficult. The correct approach: if the employee requests CART, either provide CART or discuss with the employee whether a specific alternative would meet their needs equally well — and document that conversation.

How far in advance should we book a CART captionist for a live training event?

Two weeks is the target for general-content training in most metropolitan markets. For specialised content — medical, legal, financial services, technical engineering, or any training with high proper-noun density — allow 3–4 weeks because captionists with domain expertise are in higher demand and need time to prepare their custom dictionaries. Minimum practical lead time in most markets for general content is 72 hours; below 72 hours, availability is limited and rush surcharges apply (typically +25–50%). For accommodation requests that arrive with less than 72 hours until the event, agencies may be able to fill the slot but cannot guarantee the same captionist quality level as a 2-week booking with advance materials preparation. In this situation, contact the CART agency immediately, accept the rush surcharge, and document the short-notice circumstance in the accommodation record. The training coordinator should also assess whether the event can be rescheduled if CART cannot be confirmed — proceeding with a training event without the accommodation when the accommodation was requested may constitute a Title I failure.

What advance materials should we send to the CART captionist?

At minimum: the presenter’s slide deck and a glossary of any proper nouns, product names, acronyms, and technical terms that appear in the content. More complete materials: speaker notes or a detailed outline, the full speaker roster with name spellings and pronunciations, regulatory citation formats used in the training, and — for any acronyms — the full expansion (BSA = Bank Secrecy Act, LOTO = Lockout/Tagout, not just the abbreviation). Send materials at least 24–48 hours before the event; for specialised content, 72 hours or more gives the captionist time to build a comprehensive job dictionary. If materials will be available late (presenter is finalising the deck until the morning of the event), send a draft early and a final version as soon as it is available. The captionist benefits from the draft even if it changes, because they can begin vocabulary preparation from the draft and update from the final version. Confirm with the captionist that they received and reviewed the materials before the pre-event tech check.

What does remote CART look like technically for a Zoom virtual training?

Two configurations are most common. Zoom CC channel: The meeting host enables “Closed Captioning” in Zoom account settings, then during the meeting opens the CC panel and selects “Assign someone to type.” They enter the captionist’s email; Zoom sends a token URL. The captionist opens that URL in a browser, receives meeting audio via their Zoom client or phone dial-in, and types into the browser-based CC interface. Their output appears in the Zoom caption bar for all attendees. StreamText: The captionist creates a StreamText channel, connects their CAT software (CaseCatalyst, Eclipse) to the channel, and provides the streaming viewer URL to the employee and training coordinator. The employee opens that URL on a device alongside Zoom and reads a full scrolling transcript. The StreamText method provides a better reading experience for long sessions (full scrollable transcript vs. Zoom’s limited caption buffer) and produces a post-event file. The Zoom CC method is simpler for the employee (no second device). Confirm the employee’s preference during the interactive process. Both methods work; the deciding factor is the employee’s reading preference and their device setup. For Teams-based training, StreamText is generally the more reliable option because Teams does not have an equivalent to Zoom’s human-typist CC channel.

We have an employee who attends a monthly training series. Do we need to book CART each month separately?

No — and booking separately each month is a failure mode. Once the accommodation is approved and the first session’s CART is confirmed, convert immediately to a standing booking with the CART agency for the full training calendar. Provide the agency with all future dates and times. Most agencies that work with recurring training programmes will hold a preferred captionist (or captionist team) for the series, which improves vocabulary familiarity and accuracy over time as the captionist builds context about the training programme. The practical mechanics: share the full annual training calendar with the agency, ask for written confirmation of each date on the calendar, and confirm each session approximately 5 business days in advance. If dates change, notify the agency as early as possible — cancellation within 24 hours may incur a cancellation fee. The captionist who works with the same training series over 6–12 months will develop vocabulary expertise on your specific content domain that compounds into better accuracy, just as a GlossCap customer glossary compounds accuracy on recorded content over time.

Is CART required for training events with fewer than 15 employees?

ADA Title I applies to employers with 15 or more employees. Below that threshold, ADA Title I does not apply — but several other frameworks may. California’s Fair Employment and Housing Act (FEHA) applies to employers with 5 or more employees and requires reasonable accommodation for employees with disabilities. New York State’s Human Rights Law applies to employers with 4 or more employees. New Jersey’s Law Against Discrimination has no employee minimum. Many state disability laws track ADA Title I’s substantive provisions while lowering the employee-count threshold. Additionally, if the employer receives federal financial assistance (grants, contracts), Section 504 of the Rehabilitation Act applies regardless of employee count, and Section 503 applies to federal contractors and subcontractors who meet the contract-value threshold ($15,000 for non-construction, $10,000 for construction). An employer below 15 employees should confirm with employment counsel whether ADA exemption applies in their specific state and funding context. The accommodation obligation for CART does not disappear at 14 employees in many jurisdictions.

What credentials should we look for when hiring a CART provider?

For live training CART, prioritise providers with NCRA certifications specific to realtime captioning: the CRC (Certified Realtime Captioner) is the most directly relevant credential, indicating that the captionist has passed NCRA’s realtime-specific examination. The CCP (Certified CART Provider) is a specialty overlay that indicates the captionist specialises in CART delivery specifically. The CRR (Certified Realtime Reporter) demonstrates realtime stenography at court-reporting speed and accuracy and indicates strong baseline performance. Beyond credentials, ask about domain experience in your content type: a CRC with 5 years of experience captioning medical conferences will perform differently on technical medical training than a CRC who primarily captions legal depositions. Ask for a sample of CART output from training content similar to yours and assess the accuracy on proper nouns and technical terms. For ongoing training programme CART, establish the relationship with an agency rather than an individual freelancer so that backup coverage and quality consistency are managed at the agency level. The caption vendor accuracy evaluation methodology post covers how to score and compare captioning output; the same DCMP scoring principles apply to evaluating CART provider output quality. For your broader caption programme annual review, include CART provider performance in the vendor review cycle alongside your recorded-content caption provider.