IQ Bot ships with an easy-to-use graphical training tool where users upload samples of a given document type (invoices, for example) and then define data fields they want to be identified and extracted.

An initial processing pass identifies fields automatically, and provides a graphical editor for users to review the results and manually highlight fields that haven’t been automatically identified – in which case these manually specified fields are used to further train IQ Bot.

The trained IQ Bot is then deployed as a centrally-hosted service that ‘regular’ TaskBots can call – and when IQ Bot can’t determine the information it should, humans make corrections that feed back into the IQ Bot’s underlying models.