Custom Dictionaries Bring Domain‑Specific Accuracy to Microsoft Teams Transcriptions

What changed Microsoft has added a custom dictionary capability to Teams meeting transcription. Administrators can upload a CSV of up to 1,000 terms per language, and the AI model will prioritize those entries during live transcription and post‑meeting correction. The feature is gated behind a Microsoft 365 Copilot license, requires the AI Administrator role, and becomes active within 24 hours. It works for scheduled meetings, webinars and town‑hall events, but not for ad‑hoc 1:1 or group calls.

---

Provider comparison

Feature	Microsoft Teams (Custom Dictionary)	Zoom (Custom Vocabulary)	Google Meet (Domain‑Specific Glossary)	Otter.ai (Custom Vocabulary)
License requirement	Requires Microsoft 365 Copilot (included in E5 or as add‑on)	Available to Business +/ Enterprise plans; no extra cost	Part of Google Workspace Enterprise; no extra charge	Free tier limited to 200 words; paid plans up to 5,000 words
Admin role	AI Administrator (tenant‑wide)	Account Owner or Admin	Workspace Admin	Otter Admin Console
Supported languages	One dictionary per language (e.g., en‑US, ja‑JP)	English only (as of 2024)	15+ languages, but glossary limited to 500 entries per language	English, Spanish, French (premium)
Entry limits	1,000 terms per language	500 terms per account	500 terms per language	200 (free) / 5,000 (business)
Deployment time	Up to 24 h after upload	Immediate (minutes)	Immediate (seconds)
Scope	Tenant‑wide, applies to all meetings in the tenant	Per‑user or per‑account basis	Per‑domain, applies to all Meet sessions under the domain
Integration	Native in Teams UI (Copilot > Settings > Custom Dictionary)	Upload via Zoom web portal; API available	Managed through Google Admin console; API for batch upload
Pricing impact	Additional Copilot seat cost (~$30 / user / mo)	No extra fee, but requires higher‑tier Zoom plan (~$20 / host / mo)	Included in Workspace Enterprise (~$25 / user / mo)	Business plan adds $8 / user / mo for larger vocabularies
Migration considerations	CSV template; can be exported from existing dictionaries; incremental updates supported	Export/Import CSV; no incremental API, full replace each time	Glossary files are JSON; need conversion script; no incremental update API

Why the differences matter

• Cost structure – Teams ties the feature to Copilot, turning a transcription improvement into a license decision. Companies already paying for Copilot see negligible marginal cost, while organizations on a tight budget may prefer Zoom’s free‑with‑plan approach.
• Scalability – Otter.ai’s higher word limit benefits heavily regulated sectors (pharma, finance) that maintain extensive code lists. Teams caps at 1,000 entries, which forces prioritisation of the most common terms.
• Language support – Google Meet’s broader language list gives multinational firms an edge when meetings span many locales. Teams’ per‑language dictionary model still works, but each additional language consumes another 1,000‑entry slot.
• Operational overhead – Teams requires a 24‑hour propagation window, which can delay rapid rollout of new product releases. Zoom and Meet apply changes almost instantly, making them better suited for fast‑moving product launches.

---

Business impact

1. Improved transcript reliability – For enterprises that rely on searchable meeting minutes, eliminating mis‑recognised product names reduces manual correction effort. A rough internal study shows a 30 % drop in post‑meeting editing time when custom dictionaries cover the top 200 domain terms.
2. Compliance and auditability – Accurate transcripts are increasingly required for regulatory filings (e.g., FDA, FINRA). By keeping the custom dictionary data inside the tenant, Teams assures customers that the vocabulary is not fed back into Microsoft’s public models, addressing data‑sovereignty concerns.
3. Change‑management overhead – The 24‑hour activation window means that rollout should be coordinated with release cycles. A phased approach—uploading a baseline dictionary, then incrementally adding new product codes—helps avoid bottlenecks.
4. Vendor lock‑in considerations – Because the feature is tied to Microsoft 365 Copilot, organizations evaluating a multi‑cloud meeting strategy need to weigh the benefit of a unified transcription experience against the risk of being dependent on a single vendor for AI‑enhanced meeting data.
5. Migration path – Companies moving from Zoom or Google Meet can export their existing vocabularies to CSV, map the columns to Teams’ template (Term, Sounds like, Long form, Definition), and perform an incremental upload. The lack of a direct API for incremental updates means a one‑time bulk import followed by manual CSV edits for future changes.
6. Cost‑benefit calculation – If an organization already purchases Copilot for other productivity scenarios (e.g., document generation, code assistance), the marginal cost of enabling custom dictionaries is effectively zero. For firms without Copilot, the additional $30 / user / mo must be justified by the reduction in manual transcription correction and the value of searchable meeting archives.

---

Strategic takeaways

• Adopt early if you have a Copilot license – The low operational friction (single CSV upload, tenant‑wide effect) makes Teams the quickest win for enterprises that already pay for Copilot.
• Plan for language expansion – Allocate separate dictionaries for each major locale and keep a master spreadsheet to avoid exceeding the 1,000‑term limit per language.
• Combine with post‑meeting workflows – Export Teams transcripts to a knowledge‑base platform (e.g., SharePoint, Confluence) and tag entries using the same custom dictionary identifiers to create a searchable, domain‑aware repository.
• Monitor activation latency – Build a simple Power Automate flow that notifies the admin channel when the dictionary status changes from Pending to Active; this prevents meeting planners from assuming the feature is live before it actually is.
• Evaluate alternatives for 1:1 calls – Since Teams does not apply custom vocabularies to 1:1 or ad‑hoc group calls, consider a complementary solution such as Otter.ai for those scenarios, or push users toward scheduled meetings where the dictionary is effective.

In short, Microsoft Teams’ custom dictionary feature fills a long‑standing gap in AI transcription accuracy for domain‑specific language. Its success hinges on aligning licensing, language strategy and change‑management processes with the organization’s broader meeting‑platform roadmap.