How Intact and SoftwareOne engineered a 40-fold increase in productivity

About Intact

Intact is the world leader in software solutions for the audit and certification chain of compliance and integrity-dependent sectors such as food, aviation, healthcare, and many others. Founded in Austria, Intact has offices in the other DACH markets, as well as in the UK, Australia, and the US.

After more than two decades of success and growth, Intact was determined to be a pioneer in how AI will shape the future of certification. It partnered with SoftwareOne to explore the incredible possibilities of this technology.

Why SoftwareOne

Intact looked at alternative collaborators for its AI project but selected SoftwareOne because it had the widest field of expertise, and also the manpower to keep up with the speed of AI development. Because Intact looked for breakthrough innovation, it needed more than a competent yet passive partner, one that brought its vision to the table. This is what SoftwareOne delivered, both during the research phase where it often guided the discussions and in the development of the platform.

Trust was a key element. “We were super happy with our SoftwareOne experience and want to build further on what we already have. We have now a product that works well, that is recognized in the industry. We now want to go next level with SoftwareOne.”

Intact's vision and partnership with SoftwareOne

Intact was founded in Austria in 2001, partly in response to a series of food safety scandals that rocked the industry. At that time, the certification and audit processes around food production were largely manual, and this is what Intact digitalized. Two decades later, Intact is a leader in this sector; its integrity audit platform is used by certification bodies and standard setters all over the world in many industries, including aviation, healthcare, automotive manufacturing, and of course food.

As the technologies of AI and data analytics synergize to drive the next wave of transformational change, Intact is determined to remain ahead of the pack. Much of the “new” thinking that goes on at Intact is led by Jochen Kleboth, Head of Data, Analytics, and AI.

“My team and I have been talking about the future direction of our industry for a long time, some eight years,” he says. “And at the core of our research was data. Certification and audit processes gather vast amounts of what you could call problematic data: It’s very hierarchical, it has a lot of noise, and includes a lot of free text. But it’s valuable of course. So the question arises, ‘What do we do with it? And what can we do with it?’ What our customers need above all is the ability to automate cognitive tasks, detect anomalies, analyze free text, and predict non-conformities. This is where SoftwareOne comes in.”

Seeking a trustworthy AI partner

The success of a next-generation analytics platform depended on a deep understanding of Artificial Intelligence (AI) and Natural Language Processing (NLP). “This was our biggest problem,” says Kleboth.

“We are a small team and to keep up with the speed of AI development requires critical mass and manpower. So, we looked for a partner. The choice of SoftwareOne was a no-brainer for us because SoftwareOne had the widest field of expertise, the manpower, and the agility to execute fast.”

Validating the algorithm and leveraging NLP

The project's first phase was to validate the chosen algorithmic approach for anomaly detection in audit data. A study was carried out jointly by Intact and SoftwareOne, with data provided by a certification body. The dataset contained the results of 4,562 worldwide recertification audits in the food sector from January 2018 to March 2021.

The design and results of this study have been published in the scientific journal Food Control. For the purposes of our discussion, it is sufficient to note that the algorithm was highly successful in spotting outliers that turned out to be real anomalies, with few false positives. Tobias Rechberger, Senior Data Scientist at SoftwareOne and one of the domain experts who verified the findings of the analytics engine commented: “We see this analysis as a positive validation of our approach, and we discovered inconsistencies in the data that we have not even thought of occurring and would not have found otherwise.”

“SoftwareOne’s contribution to this validation was pivotal,” comments Kleboth, “and not only on the technology side of things. The certification industry is built on trust, and in SoftwareOne we had a trustworthy partner that understood our needs, yet had a confident and independent view of its own. SoftwareOne led many of the workshops that produced such encouraging results, and eventually, the analytics platform it helped us build.”

The second pillar of SoftwareOne’s contribution was to exploit NLP techniques to neutralize the problem of free text in the audit chain.

“Auditors give ratings to a checklist of requirements and if they spot an outlier – or, what in their view, is an outlier – they explain their judgment in what we call a non-conformity description. This is free text. And we all know it is not really possible to analyze free text fields.”

What SoftwareOne developed for the Intact Analytics Platform was the ability to spot patterns in free text and make this actionable. The free text functionality observes if certain critical control points – around hygiene, for example – regularly pose a ranking problem. “These are things the Analytics Platform makes really explicit,” Kleboth says, “and that has a big effect because once you know there is a consistent problem you can revise the standard or phrase your question less ambiguously.

“The checklists of the standards were not always very clear because for some control points the Analytics Platform identified a huge difference in the findings. Once you know there is a problem you can remedy it. You can train the auditors better and say: ‘Look, this is always happening here. Focus on this especially.’ This is transformative for our customers.”

Revolutionizing productivity and auditing standards

The domain experts who participated in the Intact/ SoftwareOne validation study observed significant gains in productivity which Kleboth now quantifies as a 40-fold increase. “With this caveat,” he adds. “Yes, on average, you can carry out an audit 40 times faster with our software, but this doesn’t mean our customers now work one hour a week instead of 40! Audits take a lot of time – from 20 minutes to anywhere up to four hours. And one of our customers handles 50,000 audits a year! This means a lot of certification bodies do random checks because anything else just isn’t possible.

“But now, with our Analytics Platform, it is possible to scan 100% of your audits and then really focus on the ones that are suspicious.”

Kleboth’s instinct is that this will drive up standards generally, but – ever the statistician – admits that it would take further study to prove this.

“But 100% is better than a random sample, no matter how sound the statistics.”