Olive AI

» see oliveai.com as it lived, on the wayback machine

~ the last words ~

olive ai sat in a market i could read about but not touch. US hospital revenue cycle automation was a B2B SaaS sold to revenue cycle directors at big health systems in ohio and pennsylvania and texas, the buyer profile was nowhere near my orbit in ankara. by the time the company shut down in october '23 i had been working on real computer vision pipelines for a decade and had a reasonably calibrated sense of which AI claims to take seriously. olive was a 4 billion dollar valuation and a story that did not survive the first axios investigation.

this is mostly told from the healthcare dive obit and the fierce healthcare followup and the axios scoop. olive started in 2012 as crosschx, a columbus ohio startup founded by sean lane after he sold his earlier baltimore company BTS software. crosschx ran patient identity verification: matching a person across the fragmented record systems hospitals carried and deflecting duplicate chart numbers, plugging into the workflow that admissions clerks were doing by hand. lane talked about the opioid crisis and funerals in his hometown of gallipolis ohio as the founding context, the bigger frame was that US healthcare data plumbing was decades behind the rest of the economy.

in 2017 crosschx rebranded to olive and the product shifted to AI workforce as the pitch. the framing was that hospital revenue cycle staff spent most of the day on repetitive screen-scraping work: insurance eligibility checks, claims submission, prior authorization, denial follow-ups. olive software bots would log into the same EMR and payer portals a human clerk used and file the same forms a human clerk filed. in those days. the company called these the digital employees and gave them names. the pitch landed: tiger global led a 400 million dollar round in july '21 at a 4 billion dollar valuation, total funding crossed 900 million.

i was not paying close attention to the back half of this story but it played out in real time on the health-tech press. in october 2022 axios published an investigation showing that the AI was not really automating the work, instead human operators in the philippines and elsewhere were patching errors behind the scenes when the bots failed. customers had been quietly canceling for months. olive laid off 450 staff in july '22 and another 200 in february '23 and divested the population health line and the utilization management line through the summer (or roughly). on october 31 2023 the company told staff it was winding down. lincoln international ran the asset sale: waystar took the clearinghouse and patient access products, humata health, founded by ex-olive executives, took the prior authorization tech. the office in columbus closed within weeks.

~ the record ~

~ what the AI actually was ~

this part is the engineering question and the one i can read directly. olive used a mix of robotic process automation tooling and some machine learning models for document classification and denial reason categorization. RPA was a 2010s category of software that drives a UI through scripted clicks and keystrokes, useful when an underlying system has no usable API. healthcare was full of these systems: payer portals from the 1990s and EMR vendor screens with no integration tier, plus fax-based forms that needed OCR before they could be parsed.

the gap between the pitch and the reality was the gap between scripted RPA and the AI workforce framing the marketing team used. an RPA script breaks the day a portal changes a button color or adds a captcha. olive needed maintenance staff to keep the scripts running and offshore operators to handle the cases where the script failed silently. that operational reality was not novel, every RPA shop ran into it. olive's problem was that the company had raised 900 million dollars on a narrative that implied something closer to real autonomy than what the codebase was actually doing (or roughly).

~ the actual end ~

the timing matters. digital health funding peaked in 2021 at 29 billion dollars across the year, by Q3 2023 it had collapsed to 2.5 billion. olive needed either profitability or another large round and got neither. customer churn after the axios story closed the bridge to revenue. the may 2023 divestitures bought a few months of runway, the october 2023 sale closed the company before bankruptcy filings became necessary (cant remember whether any creditor filings followed). employees were given short notice, severance terms were tied to the asset-sale proceeds.

i did not have a strong opinion on olive while it was operating, the obit landed differently. the story is a clean case of a category that was real, hospital revenue cycle automation, met by a company that overstated its technical position long enough to absorb a 900 million dollar bet and then could not catch up to the marketing. waystar runs the clearinghouse line now under its own roof and humata is selling prior auth to a slimmer customer set, sean lane has moved on to other ventures. the columbus ohio health-tech scene that olive helped seed remains, the olive brand does not.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »