AI Providers and Models¶
FieldOps uses a provider-neutral AI architecture, allowing organizations to use AI models from different providers without changing how the AI Assistant is used.
This gives organizations flexibility to choose the providers and models that best fit their requirements.
Bring Your Own Key¶
FieldOps follows a Bring Your Own Key (BYOK) approach for AI providers.
Your organization provides its own API credentials for the AI provider it chooses to use.
The AI request is then made using that organization's provider connection.
This means:
- Your organization controls its AI provider account.
- Your organization manages its provider credentials.
- Your organization pays the AI provider directly.
- FieldOps does not act as an AI usage billing intermediary.
Supported AI Providers¶
FieldOps is designed to support multiple AI providers.
Supported providers can include:
- OpenAI
- Anthropic
- Google Gemini
- DeepSeek
- Qwen
- Meta Llama
- Grok
- Mistral
- Kimi
The providers available in your FieldOps environment depend on the providers and models configured by your FieldOps administrator.
Why multiple providers?¶
Different AI providers offer different models, capabilities, pricing, and performance characteristics.
A provider-neutral architecture allows FieldOps to work with a broad range of models instead of requiring organizations to use a single AI vendor.
For example, an organization may choose one provider for general operational questions and another for more complex analytical tasks.
The FieldOps AI Assistant workflow remains the same.
Models¶
Each provider can offer multiple AI models.
Models can differ in areas such as:
- Reasoning capability
- Response quality
- Context capacity
- Response speed
- Output length
- Cost
- Specialized capabilities
FieldOps maintains a model catalogue so that available models can be configured and associated with their providers.
Model Guidance¶
Choosing an AI model can be difficult, especially when different providers offer multiple models with different capabilities.
FieldOps provides model guidance directly in the AI connection configuration to help you make an informed choice.
When you select a model, FieldOps displays guidance about that model, including:
- Tier — the general capability level of the model
- Description — a summary of the model's characteristics
- Strengths — areas where the model performs particularly well
- Best for — the types of AI work the model is most suitable for
This guidance is displayed when you select a model, so you do not need to be an AI expert to choose an appropriate model.
For example, when deciding between models, you can use the guidance to determine whether a model is better suited for:
- Fast, straightforward questions
- General operational analysis
- Complex reasoning
- Longer analytical tasks
- Large amounts of contextual information
Use the Guidance When You Are Unsure¶
If you are not sure which model to choose, review the model guidance displayed in FieldOps and consider the type of work you expect the AI Assistant to perform.
For example, if your organization primarily asks questions such as:
How many sites do we have?
a model optimized for speed and general-purpose use may be appropriate.
For more complex questions such as:
Why has Kapsabet Sub-County Hospital been underperforming for the last three months?
you may want to choose a model whose guidance indicates stronger reasoning or analytical capabilities.
The guidance is intended to help you make a practical choice without requiring detailed knowledge of individual AI models.
Tip
You do not need to choose the most powerful model for every use case. Choose a model based on the type of work you expect FieldOps to perform, using the guidance provided for each model.
Active Models¶
A provider becomes available for use when it has active models configured.
This allows FieldOps administrators to control which models are available without changing the AI Assistant itself.
For example, a provider may have several models available from the provider, while only selected models are enabled for use in FieldOps.
Choosing a Model¶
The appropriate model depends on the type of question being asked.
For routine questions such as:
How many sites do we have?
a fast general-purpose model may be sufficient.
For more complex questions such as:
Why has Kapsabet Sub-County Hospital been underperforming for the last three months?
a model with stronger reasoning and analytical capabilities may produce a more useful response.
For longer investigations, models with larger context capacity can also be useful.
Provider Connections¶
AI providers are configured through provider connections.
A connection contains the credentials required for FieldOps to communicate with the selected provider.
Provider connections are organization-specific.
Do not share provider API keys through:
- AI Assistant conversations
- Project descriptions
- Visit notes
- Findings
- Corrective actions
- Support requests
- Other operational data
API credentials should only be entered through the appropriate provider configuration interface.
Provider Billing¶
AI provider usage is billed directly by the selected provider.
FieldOps does not purchase AI credits on behalf of your organization or resell AI model usage.
Your organization is responsible for:
- Creating the provider account
- Adding payment details where required
- Maintaining provider credits or billing
- Managing provider usage limits
- Monitoring provider-side costs
This gives organizations direct control over their AI provider relationship.
AI Usage Tracking¶
Although FieldOps does not handle AI provider billing, FieldOps records usage information for AI requests made through the platform.
This gives your organization visibility into how its AI connections are being used.
Depending on the request and configuration, usage information can include:
- Provider
- Model
- Feature
- Request
- Token usage
- Request duration
- Request status
- Error information
Authorized users can view AI usage information from the FieldOps dashboard.
This can help organizations:
- Monitor AI usage across the organization
- Understand which providers and models are being used
- Review token consumption
- Monitor request performance
- Investigate failed AI requests
- Understand how AI features are being used
Provider Billing¶
FieldOps usage tracking is separate from AI provider billing.
Your organization continues to pay the selected AI provider directly. The provider's own dashboard and billing records remain the authoritative source for:
- Charges
- Credits
- Provider-side usage
- Billing limits
- Payment status
FieldOps usage information is provided for operational monitoring and visibility and should not be interpreted as an AI provider invoice.
Provider Credentials and Security¶
Provider API credentials should be treated as sensitive credentials.
Do not expose API keys in application content or operational records.
If a provider connection stops working, common causes include:
- Invalid credentials
- Expired credentials
- Provider account restrictions
- Exhausted credits
- Rate limits
- Provider-side outages
- Model availability changes
Check the provider connection and the provider's own account status when troubleshooting these issues.
Changing Providers¶
Because the AI Assistant is provider-neutral, changing the configured provider does not require users to learn a different way of asking questions.
For example, the same question can be asked regardless of the configured provider:
Which sites require attention?
The provider and model used to answer the question may change, but the FieldOps workflow remains the same.
Comparing Models¶
Different models can produce different answers to the same question.
For example, two models may:
- Identify the same underlying issues but explain them differently.
- Produce different levels of detail.
- Focus on different aspects of the available data.
- Ask different follow-up questions.
- Reach different interpretations when the data is ambiguous.
The underlying FieldOps records remain the source of truth.
For important operational decisions, verify AI-generated conclusions against the relevant FieldOps records.
Provider Independence¶
The purpose of the provider-neutral architecture is to keep the AI Assistant independent from any single AI vendor.
This allows FieldOps to evolve as the AI ecosystem changes while preserving the same core workflow:
Ask → retrieve relevant FieldOps data → analyze → respond
Organizations can therefore evaluate different providers and models without changing how users interact with FieldOps.