Lorenz Pallhuber

The through-line

Pallhuber's consistent argument is that the constraint on procurement teams is not intelligence or strategy, it is hours. Teams have absorbed responsibilities without absorbing headcount, and the resulting administrative load crowds out the work humans are actually good at. His framing of the mission is to "leverage the incredible moment in technology to give people and teams finally the headcount that they deserve" [1]. The vehicle is agentic AI, which he describes plainly as "a really advanced AI system that can take actions on behalf of our users," producing "virtual co-workers that take on routine tasks" so people can focus on strategy and relationships [1]. That distinction between AI that observes and AI that executes and follows through runs through his discussions of how agents are already being applied in supply chain and procurement today [3][6].

The second half of the through-line is impatience with hesitation. He does not treat the technology as speculative or the rollout as heavy, and he argues repeatedly that the practical, scoped version of this is available now [1]. His warning to buyers is blunt: "people always say AI won't take your job. Um, you know, AI you you'll your job will be taken by somebody who's using AI. And I think that's very very accurate" [1].

On apathy as the real competitor

Asked how he differentiates against the crowd of agentic AI startups, Pallhuber rejects the premise. The market is early enough that competitors are not the problem: "if we look at the whole landscape of companies in this world that could benefit from Aentic AI. I would be surprised if any more than 0.01% of them have even started experimenting with them" [1]. What he actually competes with is inertia: "we really see apathy as our number one competitor. People doing the same thing over and over again and not realizing that the technology is both mature and ready for certain things" [1]. Other builders in the space he treats as allies rather than threats, on the grounds that the shared job is convincing the market that agents are ready for prime time [1].

He extends this into the biggest mistake he sees leaders make. Not deciding is itself a decision, and the larger competitors are already running experiments with dedicated teams [1]. The failure, in his account, is a lack of "the guts to try out something different" [1].

On what agents are actually ready for

Pallhuber is deliberate about not overclaiming. He concedes the limits in the same breath as the capability: "Yes, the agents aren't perfect and can do like anything that you give them. Sure. But it's enough if they can take over a lot of stuff that takes a lot of hours of people's time" [1]. The technology is "ready for certain things. It's not ready for other things" [1], and dozens of companies are running these agents live in production [1].

His model for where this lands is automotive autonomy levels. Level five does everything, level one is lane assist, and his target is "somewhere in level, you know, three to four where, you know, all of those tactical procurement things happen in the background enabled by an army of agents that do things for you" [1]. Concretely that means agents that get quotes from vendors, check in on orders, and reach out about invoice issues "in natural language in the channels people use on top of the existing software landscape" [1]. He also expects the ceiling to keep rising on its own, since the underlying models are improving daily and everything built on them improves with them, a point he illustrates with tools that can pass the LSAT and generate photorealistic video, capabilities he notes have nothing to do with his product but demonstrate the power of the layer underneath [1].

On human oversight and building for one swim lane

The autonomy Pallhuber wants is explicitly supervised. "It's not a hands-off process but actually a lot of the time you do need to ask somebody um for input" [1]. Hence the emphasis on keeping teams in the loop, routing escalations to them, and delivering "control and oversight and compliance" alongside the automation [1]. This is where he locates the value of focus. Generic tooling is available to everyone, so the work is in "focusing very very clearly on a swim lane for us that is helping procurement teams" and then building a world-class user experience tailored to that audience on top of the general-purpose models [1].

On expanding addressable spend rather than just cutting cost

The most interesting benefit in his telling is not doing the same work faster, it is doing work that was previously impossible to justify. He points to companies with hard thresholds: "we work with a couple of companies that say hey we can't source anything under 100k" [1]. With an AI sourcing agent that can take in specs, run an entire sourcing event, contact vendors in natural language and consolidate the results, competitive events become viable for a far larger share of spend [1]. His prompt to procurement leaders is to ask what sourcing events they would love to run but lack the capacity for, because "your addressable spend just got so much larger" [1].

He backs this with two deployments. A Florida-based, private equity owned manufacturer saw the sourcing agent save $2.3 million in the first two months while automating more than 600 touch points with vendors [1]. A New Jersey wholesale distributor has an agent managing essentially all POs and invoices end to end, from creation through three-way matching, at more than 95% straight-through processing, with PO and invoice entry and updates handled by the agent logging into the ERP and automating the clicks, cutting time spent on those processes by more than 90% [1].

On who buys this and why

The target is anyone carrying a high volume of tactical transactions: teams spending too much time on purchase orders and invoices, low value contracting and onboarding, or "clicking a million buttons in a source to pay tool or an ERP, kind of anywhere along the source to pay journey" [1]. In practice that skews toward manufacturers and distributors [1]. Private equity owned companies are a particular focus, partly because of margin pressure but also because sponsors have recognised the leverage across a portfolio, which has produced a pattern of one portfolio company leading to several [1]. He is careful not to frame the opportunity purely defensively, arguing there is untapped potential "to help companies um make procurement the value driver that it deserves to be, but in practice often doesn't have the tools to be," and agreeing that time is the other missing ingredient [1].

On starting small and managing the change

Pallhuber's implementation pitch is deliberately low-stakes: "We can try one workflow and we can have that up and running in two weeks and there won't be any IT build from your team that's required. We try it out. If it works, wonderful. Uh, we can, you know, layer in agent after agent, but you can start very small" [1]. The demo moment he values most is when someone watches a purchase order manage itself [1].

The harder half is people. He treats change management as a distinct hurdle rather than an afterthought, because "AI is scary and comes with a lot of connotations and people uh and worries and and and fears" [1]. His prescription is that leaders address the organisation directly and lead with the message that this is "a tool that will help us be better and focus more on our customers and do all the stuff that we're really really good at," delivered with credibility and put first rather than buried [1]. That connects back to the underlying promise of the technology: freeing professionals from manual work so they can move to higher-value, strategic decisions [3][6], and, in his own words, finally being freed "from you know clicking a million buttons in an ERP system" [1].

Takeaways

• The main obstacle to agentic AI adoption is inertia, not rival vendors: fewer than an estimated 0.01% of companies that could benefit have even begun experimenting [1].
• Agents should be judged on hours removed, not perfection: "it's enough if they can take over a lot of stuff that takes a lot of hours of people's time" [1].
• Target level three to four autonomy, with escalations, oversight and compliance built in, since "it's not a hands-off process" [1].
• The biggest prize is spend you never had capacity to source: companies with a 100k sourcing threshold can run competitive events far further down the tail [1].
• Proof points cited: $2.3 million saved and 600+ vendor touch points automated in two months at a Florida PE-owned manufacturer; 95%+ straight-through PO and invoice processing and a 90%+ time reduction at a New Jersey wholesale distributor [1].
• Start with a single workflow live in two weeks with no IT build from the customer, then layer in agents one at a time [1].
• Lead the internal rollout with a credible message about what the team gets to do instead, because fear of AI is a real deployment risk [1].
• Highest-fit buyers are high-transaction manufacturers, distributors and private equity owned companies anywhere along the source-to-pay journey [1].