ArticleTechnology & AI

AI That Works: Data First, Pilots Second

By Aaron McMillanJanuary 27, 2026undefined minsShareShareLeaders explain why data readiness and maturity trump AI hype (Credit: Getty Images)Leaders from SAP, LightSource and SpendHQ explain why data readiness and maturity trump AI hype – and where pilots fail

Last year saw plenty of AI hype, with promises it would revolutionise procurement.

By the end of the year, many procurement leaders were likely tired of hearing about AI transformation. The pressure to experiment is everywhere, but practical guidance on what actually works remains frustratingly vague.

Ironclad's 2025 State of AI in Procurement report, surveying more than 800 procurement professionals, provided a rare dose of clarity. It identified contracting as the domain where AI has demonstrably escaped the pilot trap, achieving 80% adoption and average impact ratings of 8.3 out of 10.

At the latest instalment of DPW Amsterdam, three industry leaders offered their perspectives: Baber Farooq, Senior Vice President of Product Marketing at SAP Ariba & SAP Fieldglass; Spencer Penn, CEO of LightSource; and Pierre Laprée, Chief Product Officer at SpendHQ.

Together, their insights and Ironclad’s data reveal a consistent pattern: AI succeeds where data is ready and implementation is deliberate – not where it is sprinkled on as a bolt-on feature.

Why contracting works

The 80% adoption rate matters less for the number itself than for what it represents about readiness and reliability. Gartner identified contract lifecycle management as the most mature and reliable AI application for procurement and the research confirms this pattern holds across industries.

The benefit scores are striking from Ironclad’s research: finance teams rate contract insights at 9.2/10, manufacturing teams score renewal tracking at 8.3/10, technology companies rate compliance monitoring at 8.8/10 and business services score risk assessment at 8.7/10.

Ironclad’s report explains contracting's lead with precision: contracts are already organised documents with defined sections, standardised language patterns and clear information hierarchies. They're the definitive source of truth for supplier relationships, pricing terms, obligations and commitments.

This structure enables verifiable AI outputs. Contract AI can cite specific clauses, page numbers and language that supports its extractions. By contrast, spend analysis faces wildly varying data quality across ERP systems, while supplier discovery confronts information scattered across the internet in inconsistent formats.

Contracting gives AI something concrete to work with, which explains why teams see value faster and more reliably here than in other applications.

AI succeeds where data is ready and implementation is deliberate – not where it is sprinkled on as a bolt-on feature (Credit: Getty Images)

Data readiness: The shared precondition

The three aforementioned leaders reinforce this data-first logic from different angles of the procurement landscape.

Baber positions data as "the fuel that drives artificial intelligence." His challenge at SAP has been structuring data correctly and accessing it from various sources to deliver powerful experiences. This extends beyond procurement silos: "If you're running an effective procurement strategy, it can't be siloed from the finance strategy, it can't be siloed from the HR strategy, frankly, because a significant portion of your talent strategy is actually external workforce."

The result is SAP's ambition to build "the first AI native platform for procurement powered by the SAP business technology platform," where AI operates across source-to-pay in a continuous "sense, reason and act" loop.

Spencer confronts a parallel issue in direct materials. Reflecting on his Tesla days, he recalls: "We were sourcing US$30bn of car parts on spreadsheets and emails, and when I went to the market, there really wasn't a solution to address it."

The gap persists. "On the left side of the equation upstream, we have the PLM system," he explains. "On the other side, the ERP. And in between you've had this whole chasm where there's a lot of work that actually has to happen – all in these manual spreadsheets, processes, meetings with no coherent digital ecosystem."

LightSource's response is Spec-to-Scale: "spec to source" (aligning on design through sourcing), "source to supply" (award through industrialisation) and "supply to scale" (initial production through ongoing lifecycle management). This creates a data backbone spanning engineering, procurement, manufacturing and finance – the foundation AI needs to function.

Pierre brings a spend intelligence perspective. SpendHQ has spent at least a decade "bringing clients a very solid data foundation that they can base their action upon." The shift now is volume: "We've been using data science and machine learning to generate insights for our clients for so long, but now we're just generating too much because there are so many signals out there."

SpendHQ's October 2025 investment in Sligo AI addresses this directly – adding "an agentic player on top of our insights" to "sift through the noise, prioritise, decide what's the most impactful, logical course of action”. 

In every case, AI is the superstructure; data readiness is the foundation.

Baber Farooq, Senior Vice President, Product Marketing SAP Ariba & SAP Fieldglass on stage at DPW Amsterdam

All procurement and supply chain leaders leaders should attend:

• Procurement & Supply Chain LIVE: The Net Zero Summit - QEII Centre, London, March 4-5
   

• Procurement & Supply Chain LIVE: The US Summit - Navy Pier, Chicago, April 21-22

Co-located with Sustainability LIVE, these events brings together CPOs, CSCO, CSOs, ESG leaders and senior decision-makers at a moment when sustainability, supply chains and commercial performance are increasingly interconnected.

Tickets can be booked online today for The Net Zero Summit and The US Summit. Group discounts available.

Always-on expertise over one-off tools

A second convergence emerges: the move from static analytics to proactive, configurable intelligence.

Baber highlights AI that "will sense it. It'll alert you for what has happened and recommend an action to take. And in many cases, if the action doesn't require a serious decision, it'll automatically take the action." 

This powers role-specific assistants: "There's a sourcing assistant, there's a category manager assistant, there's a buying assistant, there's an invoicing assistant. And underneath that, there are specific agents that do specific work."

Pierre frames agents differently: "I like to think of agents as an intern – you'll get an intern from the best schools, but until you've taught them everything, you've given them instruction, they're useless. But the minute you train them properly, they may become the next rockstar in your team."

The difference? This intern is "24/7, doesn't take holidays, doesn't get sick, and can do work in a way that is a lot more systematic than even your best team members could ever do."

Sligo's strength is customisability: "It'll give our client the ability to design their own agents and provide them with the tools that these agents need to actually perform the work."

Pierre Laprée, Chief Product Officer at SpendHQ

Confronting real risks

Ironclad addresses accuracy directly: "The concern – AI will extract incorrect information – is real and shouldn't be dismissed." Contracting's structure allows mitigation through citations to specific clauses and page numbers.

Pierre is more scathing about generic tools, particularly for data categorisation. "We've trained machine learning algorithms and neural networks over the years – we can do the same thing in 40 seconds on a MacBook Pro with 96% accuracy. But, the more important risk is consistency. If you run the same dataset on LLMs, you will get a different result.

"The question is, do you think your buyers should be empowered with proper data or should they try to find what changed since last time?"

Human oversight unites all perspectives. As Ironclad states: "AI handles extraction, monitoring – humans maintain decision authority."

Security receives particular focus from Pierre. SpendHQ's "hybrid approach" means clients can host the platform in the cloud for speed or "on their own infrastructure" for compliance and data sensitivity. The platform is "model agnostic, LLM agnostic" – companies can even hook agents to internally trained models for absolute control over where data lives and how it's processed.

Spencer Penn, CEO & Co-Founder, LightSource

The window for strategic advantage

Ironclad closes its report sharply: "The experimentation phase served its purpose. The question now is whether procurement organisations will learn from their success or spend another year piloting what's already been proven to work."

The 80% adoption rate for AI in contracting might seem like evidence that the opportunity has passed – that this is now table stakes rather than competitive advantage. But the variation in benefit scores and application sophistication suggests otherwise.

The data is clear; the vendors are ready. The question is whether procurement organisations will learn from the 80% who've already moved forward – or spend another year piloting what's already been proven to work.

Company portals

• Ironclad

• Lightsource

• SAP

• SpendHQ

Executives

• Baber Farooq

  Senior Vice President, Product Marketing SAP Ariba & SAP Fieldglass

• Pierre Laprée

  Chief Product Officer (CPO)

• Spencer Penn

  Co-Founder and CEO

Tags

Ai In ProcurementContract managementProcurement TechnologyDigital Transformation