AI & Privacystable

Set up the optional AI assistant — choose a provider and model, understand what data is (and isn't) sent, or run AI entirely on your own machine.

AI & Privacy

AI in Kinetex is optional and off until you turn it on. Every tool — Compare, Convert, EIB, Load Errors, Reporting — works fully without it. AI just adds convenience: suggested column mappings, plain-language explanations of differences and load errors, and the Start Here assistant.

Turn on free AI (no API key)

The fastest way to start: on the Start Here screen, click Enable free AI (or open User Menu → AI Providers). This provisions Kinetex's hosted AI in one click — no API key to paste, no account to create. Free AI comes with a daily usage allowance, shown on the Start Here card.

The free tier runs on a private Kinetex server using Ollama, with Google's Gemma (gemma4:26b) as the default model — an open model, not a third-party commercial API. Requests are PII-redacted before they're sent (see Your privacy, below).

You can turn all AI on or off at any time from the User Menu (Enable AI / Disable AI).

Your privacy

Privacy is the default, not an upsell:

• PII is redacted before every request. When AI is used for mapping, it sees

your column names and structure — never the values inside cells (SSNs, IDs, salaries, names). Sharing any cell contents is governed by a single app-level setting, "Include PII in AI requests," which is OFF by default.

• Your data files are never uploaded. All file processing happens locally on

your machine. Only the minimal metadata needed for a given AI feature is sent (and PII-redacted per above).

• You can run AI entirely on your own network (see Local AI below), so

nothing leaves your machine at all.

Choose a provider and model

Open User Menu → AI Providers to:

• See your active provider (the free hosted one by default).
• Add your own provider — paste an API key for a provider you already use.
• Pick a model — use Test to verify the connection and list the models

available to your key, then select one.

You can keep multiple providers configured and switch between them.

Run AI fully local (Ollama)

For maximum privacy — or to work offline / behind a strict corporate network — point Kinetex at your own Ollama instance. Inference then runs on your own hardware and never leaves your network. Add it as a provider in AI Providers, set the model, and Test the connection.

Prefer containers? Ollama also ships as an official Docker image, so you can run it in a container — locally or on a server your team shares — instead of installing it directly. See the Ollama Docker image, then point Kinetex at that container's address the same way.

Where AI shows up

• Map with AI — suggests column mappings between two sources (Compare,

Convert, EIB). You review and accept; nothing is applied without you.

• Analyze with AI — explains the differences on the Diffs tab, or diagnoses a

Workday load-error report in Load Errors.

• Start Here — an assistant that helps route you to the right tool.

Limits and fallback

The free hosted AI has a daily allowance (weighted by usage, shown on the Start Here card). If the daily limit is reached, Kinetex can automatically roll over to a fallback provider if you've configured one, so your workflow isn't interrupted. Bring-your-own-key and local Ollama have no Kinetex-imposed limit.

Pro AI (planned)

A future Pro AI tier will add Workday-domain grounding (Reference IDs + the EIB knowledge base) and a mapping model tuned for Workday data. This is on the roadmap.

Troubleshooting

"AI is unavailable" or a feature is greyed out:

1. Check AI is enabled — User Menu → Enable AI.
2. Check a provider is configured and the model is set — AI Providers → the

provider should pass Test.

3. If you're on the free hosted AI, you may have hit the daily limit — try

again tomorrow, add a fallback provider, or connect your own key/Ollama.

4. On a corporate network? Strict firewalls/proxies can block requests to a

hosted AI endpoint. Use a local Ollama provider, or ask your IT team to allow the endpoint.

Need help?

Email support@kinetex.ai and we'll help you get set up.