AI Models in Sally
Sally offers multiple AI models that are optimized for different needs – from fast and cost-efficient meeting documentation to highly accurate, enterprise-grade knowledge capture.
Sally’s AI models are never trained on customer data. Customer audio, transcripts, and summaries are not used for training and are not accessible to us as a provider.
The models are structured in clear tiers:
- Bronze Model (Starter plan)
- Silver Model (Pro plan)
- Gold Model (Enterprise plan)
Each tier builds on the previous one. That means:
- Higher accuracy
- Better speaker recognition
- Smarter summaries
- More robustness in complex real-world scenarios.
Quick Navigation:
- 1. Quick overview
- 2. Detailed overview
- 3. How to choose the right model
- 4. How are our AI models trained?
1. Quick overview
| Feature | Bronze (Starter) | Silver (Pro) | Gold (Enterprise) |
|---|---|---|---|
| Transcription accuracy | Up to 90.3% | Up to 94.1% | Up to 98.8% |
| Processing speed | Up to 3 minutes | Up to 3 minutes | Under 60 seconds |
| Speaker recognition | ●● | ●●● | ●●● |
| Accents & dialects | ●● | ●● | ●●● |
| Technical terminology | ● | ●● | ●●● |
| Best suited for | Simple meetings | Regular business use | Critical & large-scale meetings |
Legend: ● = industry standard | ●● = strong | ●●● = excellent
2. Detailed overview
2.1 Bronze Model (Starter)
The Bronze model is designed for simple and structured meetings where speed and cost-efficiency matter more than absolute precision.
2.1.1. What it’s great at
- Clear speech with little overlap
- Internal syncs, stand-ups, short calls
- Reliable baseline transcription quality
2.1.2 Key characteristics
- Transcription accuracy up to 90.3%
- Basic speaker recognition
- Limited robustness for strong accents or dialects
- Solid recognition of common business terms
- Processing time: up to 3 minutes
2.1.3. Typical use cases
- Daily team check-ins
- Internal updates
- Non-critical documentation
The Bronze model prioritizes efficiency. It works best when audio quality is clean and speakers are clearly distinguishable.
2.2 Silver Model (Pro)
The Silver model is the default choice for most teams. It balances accuracy, speed, and robustness and performs well in typical business environments.
2.2.1 What it’s great at
- Multiple speakers
- Mild accents and regional dialects
- More reliable summaries
2.2.2 Key characteristics
- Transcription accuracy up to 94.1%
- Improved speaker recognition
- Better handling of accents and pronunciation differences
- Strong recognition of domain-specific vocabulary
- Processing time: up to 3 minutes
2.2.3 Typical use cases
- Customer calls
- Team workshops
- Cross-department meetings
If you’re unsure which model to use, Silver is usually the safest and most balanced option.
2.3. Gold Model (Enterprise)
The Gold model is built for high-stakes conversations and large-scale usage where precision, speed, and consistency really matter.
2.3.1. What it’s great at
- Fast-paced discussions
- Overlapping speech
- Technical or industry-specific language
- Parallel meetings across teams
2.3.2. Key characteristics
- Transcription accuracy up to 98.8%
- Excellent speaker recognition
- Full robustness for accents & dialects
- Highly intelligent summaries with strong context awareness
- Processing time: under 60 seconds
2.3.3. Additional strengths
- Better understanding of intent and decisions
- More consistent structure in summaries
- High reliability for action items and documentation
- Designed for organization-wide knowledge capture
2.3.4. Typical use cases
- Sales & negotiation calls
- Strategy meetings
- Expert interviews
- Legal, technical, or regulated environments
- Company-wide documentation initiatives
Gold focuses less on individual meetings and more on systematic, reliable knowledge capture across the organization.
3. How to choose the right model
Choosing the right AI model depends on how critical your meetings are, who relies on the results, and what consequences inaccuracies could have.
A simple rule of thumb:
- You want speed and simplicity → Bronze
- You want reliable, audit-safe results for daily work → Silver
- You want maximum accuracy, scale & governance → Gold
3.1. Bronze (Starter): For learning & low-risk use cases
The Bronze model is ideal if you are just getting started or if transcripts are used mainly for personal reference.
Typical examples:
- Students documenting lectures or study groups
- Individuals learning how to work with AI-generated meeting notes
- Simple internal conversations with low documentation risk
Even at this level, Sally already performs above the general industry standard for transcription accuracy.
Speed and affordability matter more than formal correctness or legal traceability.
3.2. Silver (Pro): For professional, audit-safe daily work
The Silver model is designed for professional business use where transcripts are actively used, shared, and relied upon.
Typical examples:
- Companies documenting customer calls and internal meetings
- Teams that need structured summaries, action items, and decisions
- Organizations that require revision-safe documentation
This model offers a strong balance between accuracy, robustness, and speed and is suitable for most operational business scenarios.
Meeting outcomes have real consequences and documentation must be reliable and defensible.
Our general recommendation
In practice, we recommend the Pro license with the Silver model to most of our customers.
Why?
- It already delivers revision-safe, highly accurate results
- It covers the vast majority of real-world business meetings
- It provides an excellent balance between cost, speed, and reliability
- It leaves room to scale up to Gold for selected high-stakes meetings
For most organizations, Silver is the point where AI-generated meeting documentation becomes truly trustworthy.
3.3. Gold (Enterprise): For high-stakes and regulated environments
The Gold model is built for maximum reliability in environments where every word matters.
Typical examples:
- Management teams of publicly listed companies
- Executive boards documenting strategic decisions
- Organizations in regulated or compliance-driven industries
Here, transcripts are not just notes — they are part of decision justification, accountability, and governance.
Inaccuracies could lead to legal, financial, or reputational risk.
4. How are our AI models trained?
Protecting your data is a top priority at Sally. We never train our AI models using customer data.
Neither audio recordings nor transcripts or summaries from customer systems are used for model training - and they are not technically accessible to us as a provider.
4.1 No use of customer data
In concrete terms, this means:
- Customer data is not stored, analyzed, or reused.
- No training, fine-tuning, or prompt learning is performed using customer data.
- Meeting content remains the exclusive property of the customer.
- Processing is isolated, purpose-bound, and compliant with applicable data protection and security standards.
4.2 Training with proprietary, controlled datasets
Our AI models are trained exclusively on internal, anonymized, and legally compliant datasets, including:
- More than 120,000 hours of self-generated and licensed audio material.
- Over 25 million annotated sentences for transcription, speaker separation, and contextual understanding.
- Thousands of simulated meeting scenarios with varying audio quality, number of speakers, and domain-specific language.
- Controlled datasets covering accents, dialects, and industry-specific terminology.
These datasets are continuously expanded, reviewed, and quality-controlled — without using real customer conversations.
4.3 What this means for you
- Maximum data security with no hidden secondary use.
- Reproducible and explainable model quality.
- A clear separation between product usage and model training.
- Suitable for use in sensitive, regulated, or confidential environments.