Financial services · Automation · 2015–2017

KYC from 800+ sources, after two attempts were abandoned

Pulling know-your-customer data from more than 800 registry and regulator sites, accurately and continuously, while the sites changed underneath it. Two previous attempts had been abandoned — not because the automation could not be written, but because keeping it alive cost more than the data was worth.

KYC data pulled from 800+ sites into the client's existing BPM system More than eight hundred registry and regulator sites around the world are read by one bot each, generated with ZenVerify. Sixty bots run in parallel, and the results are handed to the client's existing BPM application through an API rather than replacing it. Two earlier attempts had been abandoned because maintaining hand-written bots cost more than the data was worth. The constraint Maintenance, not authorship Two earlier attempts were abandoned. Writing 800 bots was never the hard part — keeping them alive as 800+ sites changed was. 800+ source sites registries and regulators, worldwide each with its own markup each one changes on its own high accuracy is non-negotiable ZenVerify one generated bot per site 100+ bots built in weeks 800+ bots live by month 6 60 running in parallel Client’s own BPM untouched, reached through an API Nothing the client already ran had to be replaced, which is most of why this attempt was allowed to proceed. 90% lower cost to build and maintain than the attempts before
KYC for a Big 4 accounting firm. The win is in the right-hand box: the cost of keeping 800 bots alive, which is what killed the two attempts before ours.

What was wrong

  • KYC data had to be retrieved continuously from 800+ sites worldwide, at very high accuracy.
  • Those sites ran in different languages and changed their formats without warning, so anything built against them was out of date almost immediately.
  • Two previous attempts had failed because the cost of building and maintaining the bots was prohibitive.
  • Whatever was built had to integrate with the firm’s existing in-house BPM application, which was not going to be replaced.

What we did

  • Used ZenVerify to generate one bot per site, so adding the eight hundredth site cost roughly what the first one did.
  • Attacked maintenance rather than authorship — the cost that had killed both earlier attempts.
  • Integrated through ZenVerify’s APIs into the existing BPM application, leaving the client’s own systems untouched.
  • Handled the language differences and the format changes in the generated bots, so a site changing its layout was a regeneration rather than a rewrite.

What came out

  • 100+ bots built within weeks and integrated with the client’s BPM application.
  • In production within six months with 800+ bots and 60 parallel executions.
  • Cost of building and maintaining the bots reduced by 90%.
“Their expertise in building adaptable, cost-effective solutions allowed us to deploy a robust system in record time. They delivered beyond our expectations, and we are thrilled with the results and the long-term impact on our operations.”
The client — a Big 4 accounting firm, name withheld by agreement

Built with

  • ZenVerify
  • JavaScript
  • AWS

Client names withheld by agreement. References available on request. ZenVerify later led to EverTest, our AI test-automation company.

Does your automation cost more to maintain than it saves?

That is the cost that kills many large suites, and it is the one AI coding tools make easier to ignore. Show us the suite and we will tell you honestly whether it is worth keeping.