I talk to supply chain and IT leaders every week about where AI actually delivers value versus where it just adds noise. So when IDC shared research surveying over 2,000 supply chain professionals across every industry and geography, I wanted to see if the data matched what I'm seeing on the ground. Mostly, it did, and it sharpened a few things I've believed for a while.
1. Adoption is easy. Accountability to an outcome is hard.
Nearly every company I talk to has AI touching their supply chain somewhere: the IDC data puts it at 98%. But only 12% consider themselves leaders. That tracks with what I see constantly: it's not hard to stand up a pilot anymore. AI has democratized access to data and to modeling. What's hard is building something that's actually accountable to a business outcome, with the guardrails and governance to back it up.
2. Trust is a design problem, not a data problem.
Over half of the IDC survey respondents pointed to trust in their data and systems as their biggest blocker. I'd go further: trust is a function of whether the system was built with governance in mind from day one, or bolted on after the fact. Most of the fast proofs of concept I see get built with no logging, no security model, no defined guardrails, which is exactly why they stall before scaling. You don't retrofit trust. You architect for it from the start!
3. Perfect data is a myth. Use AI to get closer to good data.
True, 62% said better data quality would accelerate their AI efforts. I'd have guessed higher. But I don't buy data readiness as a gate you have to fully clear before you start. In my experience, the tools themselves like AI-detected seasonality patterns, automated lead-time correction, etc., are actually some of the best ways to improve data quality. Don't wait for clean data. Use AI to help get you there.
4. The real risk isn't slow decisions, it's confidently wrong ones.
This is where I think a lot of organizations are miscalibrated. Everyone wants speed. Fewer people are asking whether the faster answer is actually the better one. A decision delay to catch a bad recommendation costs nothing compared to a wrong million-dollar purchase order. The magnitude of the decision has to determine how much autonomy you give it. It should be a core design principle for agentic AI in planning.
5. Your best people will thank you for this, if you frame it right.
The stat that didn't surprise me: 80%+ of respondents see AI as an opportunity rather than a threat to their role. I've watched this happen firsthand with our own customers and teams. Once AI takes the manual, repetitive work off someone's plate, they get to spend their time on judgment calls that actually require their expertise. Technology doesn’t cause adoption to fail – the failure comes from leaders who deploy AI as a mandate instead of an empowerment story.
The bottom line I keep coming back to: this isn't about how much AI you've deployed. It's about whether you've built the trust, the data foundation, and the governance to make that AI accountable to outcomes that actually matter to your business. That's the gap worth closing, and it's the one most companies haven't started on yet.