The Path to Human-Machine Harmony

A few weeks ago, OpenAI CEO Sam Altman warned that the AI frenzy is spilling past reason. Adding fuel to the fire was a recent MIT study claiming 95% of companies see no returns from genAI. The market response? A tech sell-off that has rattled investors. 

Enthusiasm for AI is wavering, but is AI truly overhyped? Yes and no. Altman’s perspective describes the paradox of tech bubbles: they often form around technologies of genuine impact. 

While broad AI investments may be hitting turbulence, sourcing and procurement still represent fertile ground where AI's impact is measurable and immediate. Research by HFS shows that AI-driven sourcing already delivers 20% cost savings, and with the introduction of AI agents, sourcing – one of the enterprise's greatest cost centers – stands poised for fundamental transformation. 

The 95% Failure Rate Isn’t About Technology 

The infamous MIT study is a tad nebulous, but it does point to a fundamental issue. The failure point here isn't technological – it's strategic. Companies treating AI as a silver bullet are more likely to be disappointed by overinvestment, while those focusing on specific, measurable use cases can see dramatic results. 

Think about the success stories already emerging. Novartis reduced supply disruptions by 35% using AI to evaluate and assess risk across 60,000 suppliers. In the 24 months after implementation, it cut single-source dependencies from 15% to 5% of spend and reduced quality-related losses by 25% annually. Walmart's AI-powered sourcing agent handles 2,000 simultaneous negotiations with long-tail suppliers and delivers 3% average savings across deals worth up to $5 million. These aren't moonshot experiments – they're operational realities generating measurable value. 

The difference? These companies didn't ask "How can AI transform everything?" They asked: "Which specific tasks should machines own, and which require human judgment?" 

The Great Work Divide 

Success with AI depends on the clarity of task ownership between humans and machines. But how do you divide the work? In our own discussions with Zero100 members, the answer lies in the division of labor along end-to-end workflows or “jobs to be done.” 

Early implementations demonstrate that machines shine when owning data- and document-heavy tasks, like finding new suppliers, managing contract compliance, or monitoring operational performance. Humans, in contrast, will lean further into tasks that require emotional intelligence, judgment, and ingenuity. This includes shaping the product portfolio, engaging suppliers on corporate objectives, and supporting co-invention with key partners. 

And companies are already applying AI to machine fit-for-purpose tasks:

• Belgian pharmaceutical company UCB implemented an AI agent to assist with supplier qualification and selection. The agent can identify best-fit suppliers, generate negotiation insights, and even facilitate guided negotiation rounds with suppliers and compare offers.
• Lenovo’s Supply Chain Intelligence (SCI) platform integrates over 800 data sources to evaluate the performance of its supply chain, including key supplier metrics like quality and procurement cost trends. The platform automatically generates real-time alerts when KPIs decline. This allows Lenovo to take pre-emptive action, such as recommending a human to decide if or what corrective measures need to be initiated, before performance issues disrupt operations.
• A Unilever AI model integrates real-time sales and forecast data with customers, providing suppliers with precise demand signals and enabling more efficient production and logistics. This allows Unilever to continuously monitor and assess supplier performance based on metrics derived from optimized order accuracy, reduced overproduction, and efficient delivery routes. 

Shared activities between humans and machines might include contract negotiations (AI assists, humans decide), supplier selection (AI narrows, humans choose), risk assessment (AI flags, humans interpret). But the lesson is clear: for AI to deliver ROI, the key is task-level application. 

The Rise of the Digital Employee 

At the moment, despite their autonomy, AI agents aren't likely to take over entire sourcing positions. That said, agentic capabilities are evolving quickly. 

Velon AI, for example. has unveiled three “AI employees” that leaders will be able to plug into their sourcing and procurement organization. Natalie is a category expert – she sets category strategies and runs complex multi-supplier negotiations. Bob is a tactical buyer and manages SAP objects, purchase orders (POs), change orders, and compliance. Sophie is a supplier manager – scanning markets while also qualifying, vetting and onboarding new suppliers. These three roles are just the beginning; the company has more than 20 roles currently in development.  

These advancements mean that sourcing employees need to evolve, now. Fast movers will invest in digital skills, of course, but will also enhance uniquely human skills – think customer-centricity, adaptability, systems thinking, and creativity. These are the skills that enable a sourcing leader to build strategic partnerships, navigate complex geopolitical landscapes, and drive true innovation with suppliers. And this is what the path the human-machine teams and partnership looks like as we consider the road ahead, with forward-looking companies already upgrading their sourcing job descriptions to prepare: 

Hype to Harmony

Companies are finding tangible benefits from AI in sourcing – across 12% of 900+ use cases in our AI Hub according to Zero100 analysis. But managing irrational exuberance means starting small with specific, well-defined use cases, thoughtfully re-shaping sourcing roles, and investing in developing the uniquely human skills that will continue to set your team apart. 

We may be in an AI bubble – but it doesn’t have to burst.