From my conversation with Roland Seckler of Ypsomed and Paul Grandy of Sphera. Read the full case study →
When Paul meets a company that wants to track supplier risk, he starts with its problems. His questions make a good self-check for your own team:
Is your risk process manual, and does it take people away from their day jobs?
Can you find the information you need at all?
Do you hear about a problem when it happens, or when the supplier decides to tell you?
Which risks matter most where your suppliers are: geopolitical, man-made or natural disasters?
For Paul, question 3 is often where a company’s problems start. “A lot of companies’ first problem is they’re reactive rather than being proactive to any types of risk,” he says.
I asked Paul how Sphera finds out first. Here are the steps he described.
From the news to your inbox
1. Watch the news. Sphera says it monitors more than 500,000 news sources a day and draws on 15 billion data points.
2. Work out who and where. Paul’s example is a reported factory fire. Sphera’s language models work out which company and which location are involved.
3. Match it to your suppliers. The system checks whether that site connects to your supplier network.
4. Rank it. Each event gets rated as a high, medium or low alert.
5. Check it. Sphera’s risk analysts work around the clock to validate priority alerts. Sphera says the mix of software and people gives it 99.9% accuracy.
6. Send it. The alert says what happened, which suppliers are involved, which products or parts could be affected, and where you may need to act.
At Ypsomed, step 6 arrived five minutes after a fire near a supplier was on the news. I cover what happened next in ‘What Ypsomed does when a supplier risk alert comes in.’
Where teams fall short
Paul says Sphera has run this platform for close to 20 years and has used AI since it started, long before AI became a buzzword. The newer AI work at Sphera goes in two directions: watching more alerts and more suppliers, and helping customers act on what they get.
That second part is where Paul sees teams struggle. “Where a lot of teams fail is not being able to leverage the data,” he says. Responding to an alert takes time, because someone has to contact the supplier and follow up. Paul says Sphera is rolling out AI this year that summarizes alerts, answers questions about the data and suggests next steps with the supplier.
What a customer wants next
Roland wants the tool to look ahead. In one or two years, he wants the software to say what could happen when a crisis starts, as when the conflict in the Middle East began. It should also say which suppliers the crisis could hit. Then Ypsomed can act on that supplier or region before anything breaks. Today, he says, that still takes too much manual work.
Before you apply this
The 500,000 sources, 15 billion data points and 99.9% accuracy are Sphera’s own figures. When you look at any risk tool, ask the vendor how it measures accuracy. Then test it on your own suppliers, as Ypsomed did with a small proof of concept.
More from this conversation:
