The Victorian government stopped using an IBM analytics tool as part of its digital contact tracing system after its requirements changed and the system could not be reconfigured.
IBM was one of two large vendors - the other being Salesforce - that was brought in by the government in July to digitise contact tracing just prior to a surge in coronavirus cases in the state.
However, a parliamentary inquiry into the re-developed system [pdf] has revealed that use of the IBM component was discontinued a month later.
The committee branded the IBM engagement “a misguided and costly mistake” in its report.
The contract value for IBM i2 advanced analytics software was almost $4.2 million. However, a Department of Health and Human Services spokesperson said only a fraction of that was actually spent.
The committee said the government approached IBM in mid July “to determine whether its i2 analyst platform could assist Victoria with Covid-19 data analysis.”
IBM was then awarded a tender at the end of August for “enhanced health tracing”, with the committee describing the tool as being useful “for predictive modelling of Covid-19 data to inform policy and public health decisions.”
Media reports in September branded i2 as an “artificial intelligence” tool, and there appears to have been some confusion about what i2 could and could not do.
IBM global technology services general manager Charles Agee told a public hearing of the committee “that the i2 platform is designed as ‘an analyst tool, intended to assist a user in performing their analysis’, further stating ‘it is not an AI tool’, because it does not have capability to automate the analysis or to predict an alert.”
This caused problems as the project progressed.
After a “two-day period of use” in early October, the government “directed IBM to consider developing capabilities for predictive analysis and automated alerts using the i2 platform and to provide indicative timelines for completing that process.”
“According to IBM Services, the Victorian government needed these new capabilities because the situation had changed since IBM was first hired,” the committee said.
“While IBM did put forward some solutions to address the requested automation, ultimately, IBM’s i2 platform capabilities did not extend to predictive analysis or automated alerts.”
Agee said that by the week of October 12, “it became clear to IBM that the DHHS [Department of Health and Human Services] data analysis team required a system capable of alerting them to significant events or conditions that should be followed up by an analyst, in contrast to an analyst using the system to discover these conditions or events.”
It is not clear from the committee report where this misunderstanding might have occurred.
However, it ultimately led to the government calling time on its use of i2.
“On October 14, the DHHS project manager requested IBM stop work on the i2 platform deployment,” Agee said.
“DHHS subsequently indicated their intention not to proceed to the next stage using the i2 platform’s capabilities as its requirements had evolved.”
The committee added that it had received additional evidence that “the platform is not in use by the Victorian government today.”
The Salesforce software, by contrast, had worked as intended, even if the state government had rebuffed an earlier attempt to license it.
“The committee notes that Salesforce first offered its contact tracing end to end software to the Victorian Government on March 24 2020, but this offer was declined,” it said.
“On July 24, following an invitation by the Victorian government, Salesforce presented its contact tracing capabilities to the Department of Health and Human Services.”
The Salesforce platform “established an end-to-end contact tracing system … replacing the version of contact tracing deployed during Victoria’s second wave which relied on manual data collection and case interviewing, several companies responsible for different components and a highly centralised public health response sitting within [the DHHS]”.
Updated, 3.55pm: DHHS advised iTnews that it spent only a fraction of the $4.2 million allocated to the IBM project on a proof-of-concept, and did not progress this further.