taking input management to a new level with the cognitive classifier

With its diverse range of insur­ance and finan­cial ser­vices, the Debe­ka Group is one of the top five in the Ger­man insur­ance and home sav­ings indus­try. Every year, Debe­ka receives around 16 mil­lion doc­u­ment pages via a wide vari­ety of input chan­nels: clas­si­cal­ly by mail or fax, by e‑mail, as an upload via the cus­tomer por­tal, or via sales. In addi­tion to the sheer vol­ume of doc­u­ments, the com­plex­i­ty of incom­ing doc­u­ment types is a chal­lenge due to the many dif­fer­ent insur­ance lines.

Debe­ka has been using our smart FIX solu­tion very suc­cess­ful­ly for sev­er­al years. Essen­tial­ly, two use cas­es were imple­ment­ed here: iden­ti­fy­ing the respon­si­ble divi­sion and ana­lyz­ing the doc­u­ment struc­ture and con­tent. Based on the iden­ti­fied doc­u­ment class, rout­ing to the appro­pri­ate ser­vice depart­ment takes place and the cor­re­spond­ing work­flows are start­ed.

The incom­ing mail of the car insur­ance and build­ing soci­ety divi­sions is char­ac­ter­ized by many incom­ing doc­u­ments with images such as acci­dent pic­tures, real estate plans, or ID doc­u­ments. These doc­u­ments offer lit­tle poten­tial for pure­ly text-based clas­si­fi­ca­tion and thus had to be indexed almost entire­ly man­u­al­ly. Thus, the idea arose to extend the text clas­si­fi­er used so far with an image clas­si­fi­ca­tion. The new Cog­ni­tive Clas­si­fi­er now com­bines the best of both worlds: the proven excel­lent text recog­ni­tion and now, new­ly, the clas­si­fi­ca­tion of images with Deep Learn­ing. The pre­vi­ous work­flows remained unchanged and the front end for users also remained the same. The AI con­tin­ues to work reli­ably in the back­ground and decides whether text recog­ni­tion, image recog­ni­tion, or both are to be used.

The automa­tion effect of the com­bined clas­si­fi­er is enor­mous. While the major­i­ty of doc­u­ments con­tain­ing images had to be indexed man­u­al­ly before the intro­duc­tion of the new process, this is now less than 10 per­cent.

“Togeth­er with our part­ner Insid­ers Tech­nolo­gies, we have deployed an AI solu­tion that is absolute­ly at the cut­ting edge for an impor­tant use case.This makes work­ing with Insid­ers fun!”

says Patrick Schnei­der, Head of Fron­tend Depart­ment at Debe­ka Kranken­ver­sicherungsvere­in a. G.

With the Cog­ni­tive Clas­si­fi­er, a uni­ver­sal solu­tion for the clas­si­fi­ca­tion of doc­u­ments and pho­tos is now avail­able that puts all exist­ing approach­es in the shade. This allows proven process­es to be used even more effec­tive­ly and the lev­el of automa­tion to be fur­ther increased. The insurer’s and Insid­ers’ project for the Cog­ni­tive Clas­si­fi­er is already attract­ing a great deal of inter­est in the indus­try.