Looking For Advice: How Can I License A Large Swahili Voice Dataset For AI Training?

I run a call center in Dar es Salaam, Tanzania, and we have been operating for about five years.

Over that period, we have accumulated thousands of hours of naturally occurring Swahili call-center conversations. The calls are recorded with customer consent, and we can anonymize/redact personal and sensitive information before any potential licensing.

I’m exploring whether this type of data could be valuable for companies developing:
– Speech recognition / ASR
– Conversational AI
– Voice agents
– Multilingual AI models
– Swahili language models

I’ve started looking into companies such as Defined.ai and Appen, but I’d like to hear from people who have experience in AI training-data licensing or selling speech datasets.

What would be the best way to approach this market?

Specifically:
Which companies or marketplaces should I contact?
Is licensing the dataset non-exclusively generally better than selling it outright?
What documentation/metadata would buyers expect?
What kind of pricing model is typical for large speech datasets?

I’m not looking to publish any of the actual recordings publicly. I’m mainly looking for advice from people who understand the AI data marketplace.
Thanks.