Configure AI Providers

BayWise Scheduler uses external AI providers to power its intelligent features: equipment detection from bay photos, technician scheduling suggestions, job duration estimation, repair analysis, and incentive formula generation. These features rely on models from providers you choose and configure yourself — BayWise does not bundle any AI model of its own.

This means you are always in control of which provider you use, which model you pay for, and how your workshop data is handled. It also means that before any AI feature becomes active, you need to supply at least one provider’s API key.

Who can configure AI providers: Account Owner, Org Admin.

The five AI functions

BayWise treats each AI capability as a separately configurable function. You can route each one to a different provider and model, or point all five at the same provider.

Function	What it does	Recommended model type
Vision	Detects equipment from bay photos	Vision-capable multimodal model
Scheduling Suggestions	Recommends which technician to assign to a job	Text reasoning model
Duration Estimation	Estimates how long a job will take based on service type, vehicle, and history	Text reasoning model
Repair Analysis	Analyses repair requirements and recommends approaches based on vehicle and fault data	Text reasoning model
Incentive Formula Generation	Drafts a starter incentive formula based on your inputs	Text reasoning model

Vision requires a model that can process images. The four reasoning functions work with any capable text model.

You can use different providers for different functions. For example, use Google’s Gemini for vision (equipment detection) and Anthropic’s Claude for scheduling and incentive formula generation. This lets you optimise for cost and performance per task.

Before you start

You will need at least one API key from a supported AI provider. At the time of writing, BayWise supports:

Gemini (Google) — models such as gemini-2.0-flash, gemini-1.5-pro, gemini-1.5-flash
OpenAI — models such as gpt-4o, gpt-4o-mini, gpt-4-turbo
Claude (Anthropic) — models such as claude-sonnet-4-20250514, claude-3-5-haiku-20241022

Obtain your API key from the provider’s developer console before opening BayWise settings.

Configure a provider for a function

Open AI Settings

From the top navigation, click the Settings icon (gear, top right). In the settings panel, select the AI tab.

You will see five function cards:

Vision
Scheduling Suggestions
Duration Estimation
Repair Analysis
Incentive Formula Generation

Each card shows its current configuration status: Not configured, Configured, or Error.

Select the function to configure

Click the card for the function you want to set up — for example, Scheduling Suggestions.

The card expands to show the configuration fields for that function.

Choose a provider

From the Provider dropdown, select the AI provider you want to use for this function:

Gemini (Google) — for Gemini-family models
OpenAI — for GPT-family models
Claude (Anthropic) — for Claude-family models

Enter the API key

Paste your API key into the API Key field.

API keys are stored encrypted. Once saved, the key is never displayed in full — you will see only the last four characters (for example, ...k9mZ). If you need to update the key, you replace it entirely.

Do not share your API key. Do not paste it into any field outside BayWise settings.

Enter the model name

Type the exact model name into the Model field. Use the name exactly as the provider specifies it in their API reference:

Vision function: gemini-2.0-flash (Google) or gpt-4o (OpenAI)
Scheduling Suggestions: claude-sonnet-4-20250514 (Anthropic) or gpt-4o-mini (OpenAI)
Duration Estimation: claude-sonnet-4-20250514 (Anthropic) or gpt-4o-mini (OpenAI)
Repair Analysis: claude-sonnet-4-20250514 (Anthropic) or gpt-4o (OpenAI)
Incentive Formula Generation: claude-3-5-haiku-20241022 (Anthropic) or gpt-4o-mini (OpenAI)

If you are unsure of the exact model name, check your provider’s API documentation — not their marketing pages, which sometimes use display names that differ from the API identifier.

Test the connection

Click Test Connection. BayWise sends a minimal request to the provider using your key and model name to verify:

The API key is valid and active under your account
The specified model exists and is accessible
For the Vision function only: the model can accept image inputs

The test takes 3–8 seconds. A green tick confirms the connection is working. A red error message tells you what failed.

Common causes of test failure:

Invalid API key — check for leading or trailing spaces when pasting
Model not found — double-check the model identifier against the provider’s API reference
Insufficient quota or billing issue — your provider account may need a payment method or credit top-up
Rate limit — you have exceeded your provider’s request limit; wait a moment and retry

Fix any errors before saving. BayWise will not save an untested or failing configuration.

Save the configuration

Click Save. The function card updates to show:

The provider name
The last four characters of the key (e.g., ...k9mZ)
The model name
Status: Configured

Repeat steps 2–7 for each function you want to enable. You do not need to configure all five to use BayWise — configure only the functions you want to use.

Using different providers per function

There is no requirement to use the same provider for all five functions. A common setup at multi-make workshops is:

Vision → Google gemini-2.0-flash — fast, cost-effective multimodal model with strong image recognition for workshop equipment
Scheduling Suggestions → Anthropic claude-sonnet-4-20250514 — strong structured reasoning over technician profiles and job data
Duration Estimation → Anthropic claude-sonnet-4-20250514 — same provider and key, no extra credential needed
Repair Analysis → Anthropic claude-sonnet-4-20250514 — detailed reasoning over vehicle fault data and repair approaches
Incentive Formula Generation → Anthropic claude-3-5-haiku-20241022 — fast and cost-effective for a one-off generation task

Each function stores its own provider, key, and model independently. Changing one does not affect the others.

What happens if AI is not configured

BayWise is fully functional without any AI provider connected. All core scheduling, job management, bay assignment, technician management, and reporting features work as normal. No data leaves your workshop unless you have configured and tested a provider.

When AI functions are not configured, the affected controls are visually disabled:

The AI Suggest button in the technician assignment panel is greyed out
The Detect Equipment from Photo button in bay settings is hidden
The Generate option in incentive scheme creation does not appear
Repair analysis suggestions are unavailable
Duration estimation falls back to the manually-set default durations from your service catalog

Your advisors and controllers can run all standard operations — they simply make assignments and estimates manually, as they always have. Many workshops run in manual mode for their first few weeks, letting job history accumulate, then enable AI once there is enough operational data for the suggestions to be meaningful.

Updating or removing a key

To rotate a key: Open the function card, click Update Key, paste the new key, run the test, and save. The old key is replaced immediately.

To disable a function: Open its card and click Remove Configuration. The function reverts to disabled state. This does not delete any job history or settings — it only removes the AI connection for that function.

Common questions

Does BayWise send customer or vehicle data to the AI provider? No. BayWise sends structured job attributes — skill requirements, WIP counts, service type, estimated durations, technician efficiency — not customer names, vehicle registration numbers, or any personally identifiable information. The data sent is operational metadata. Review your provider’s data processing terms if your workshop has specific data residency requirements.

Can I use the same API key for all four functions? Yes, if all five functions use the same provider. Enter the same key for each function that uses that provider. BayWise stores each independently, so the key is not shared across functions in a way that exposes it — it is simply the same credential stored five times.

The connection test passes but the AI Suggest button is still greyed out. Why? A passing test confirms the API key and model are reachable, but also check that you configured the Scheduling Suggestions function specifically — not just Vision or Duration Estimation. Each function must be configured independently. Open the AI settings tab and confirm the Scheduling Suggestions card shows Configured.

Which provider and model gives the best scheduling suggestions? This depends on your job mix. In general, larger reasoning models produce better-justified suggestions when jobs are complex or technician profiles are nuanced. For a workshop with a straightforward service menu and 5–10 technicians, a smaller model such as gpt-4o-mini or claude-3-5-haiku-20241022 is typically sufficient and more cost-effective.

My provider is not in the dropdown. Can I still use it? Only the listed providers are currently supported. Contact BayWise support if your preferred provider is absent — provider support is expanded based on customer demand.

What does “model not found” mean if I copied the name from the provider’s website? Provider marketing pages sometimes use display names that differ from API identifiers. For example, a provider’s website may say “Gemini 1.5 Pro” while the API expects gemini-1.5-pro. Always use the identifier from the provider’s API reference documentation.