Ahead of his appearance on stage in Chicago at Expana’s Agrifood Intelligence Summit, Steve Tucker talks about starting in commodities at 18, why procurement still reports the past, and what makes a customer worth serving first when supply tightens.
Starting at Mars
I began my career at 18 at Mars, a world-class company that set the gold standard for operational discipline and commodity expertise. It was an incredible environment to start, where I managed a cocoa and vegetable oils portfolio and learned fast that even small price movements significantly impact a massive P&L. I later built on this foundation at United Biscuits and by trading commodities on the floor of the LIFFE exchange in London, where I gained a visceral understanding of market volatility. These experiences were transformative; they taught me that volatility is not the exception, but the job, and that margin management requires rigorous, proactive insight—the very problems I’m solving at Kodiact today.
The problem I saw at 18 and had no way to fix: we knew our markets and we knew our contracts, and the two never met. Price forecasts sat in one place. Formulas sat in another. Plant specifications sat in a third. My own head was the integration layer. When I left the room, the intelligence left with me.
Thirty years later most procurement teams still work the same way. The difference now is the tooling finally exists to hold the whole picture in one place.
Why is procurement still looking in the rear-view mirror?
Procurement has no data problem. Procurement has a resolution problem.
Ask a category manager what the business pays for glucose syrup across eleven plants in four countries. You will wait a week for the answer. The data exists. The data sits in different systems, under different material numbers, in different units, with different supplier names for the same legal entity. Someone reconciles all of this by hand before analysis even starts. By the time reconciliation finishes, the market moved.
The second reason is structural. Category strategy became a document. Teams build a deck once a year, present the deck, then store the deck. Markets do not respect an annual cycle. A strategy locked in January is wrong by March, and nobody has the hours to rebuild the analysis, so the team defends the old plan.
The third reason is rhythm. Variance gets reported after close, because reporting sits with finance and finance runs monthly. Procurement hears about a problem once the problem is already a number in a management account.
None of these are AI problems. All three are data structure and operating cadence problems. Fix those and forward-looking becomes the default.
Is this time different?
A sceptical CPO is right to be sceptical. ERP promised a single source of truth and delivered a system of record. Spend analytics looks backwards at what you spent, not what you should have spent. Big data promised insight and delivered dashboards nobody read.
Here is my test for any vendor. Ask them what their system knows about your materials before you train anybody. Every previous wave needed you to supply the meaning. You mapped the taxonomy. You cleaned the master data. You wrote the logic. The software gave you containers and you filled them.
What changed is machines reading unstructured content at scale. Specifications, contracts, price letters, index clauses, supplier certificates. Meaning gets extracted instead of typed in. Work needing six analysts and three months now takes hours. Nothing else in forty years moved that constraint.
Where humans still matter most: judgement under ambiguity, and relationships under stress. A model will tell you a supplier sits 8 percent above market with single-site risk. A model will not tell you whether the plant manager takes your call at 11pm during a recall, or whether squeezing price this quarter costs you allocation next year. Machines handle recall and reconciliation. People handle trust and consequence.
Build or buy?
Procurement teams should preserve their internal energy for high-value category strategy and relationship management, rather than trying to build complex, commodity-agnostic systems in-house. Outsourcing the heavy lifting of data reconciliation and material modeling to specialized platforms like Kodiact ensures you are not wasting resources on internal plumbing that specialized providers have already perfected.
I have built enterprise software for thirty years, including two companies acquired by larger platforms. Internal teams underestimate three things every time.
First, the ontology. A working commodity model spans thousands of materials, hundreds of specification attributes, unit conversions, yield factors, and the messy gap between what a supplier sells and what a plant consumes. Building the model once is a project. Maintaining the model as suppliers, specifications and regulations change is a permanent function with permanent headcount.
Second, the long tail. A demo works on clean data. Production breaks on the eleventh plant with a naming convention inherited from a 1998 acquisition. Ninety percent accuracy sounds strong and is unusable for a buyer signing a contract.
Third, the cost of working alone. A vendor serving thirty manufacturers sees thirty variations of the same problem. An internal team sees one. Pattern coverage compounds across customers and stays flat inside a single company.
My advice is this: focus your procurement team on talent development, supplier relationships, internal collaboration and best practice. Leave the complex data engineering to AI, which turns institutional knowledge into a scalable asset. Do not build a materials master data engine — build your team’s capability to use one.
Who gets served first?
Price wins in a buyer’s market. Allocation wins in a seller’s market. Most procurement organisations are designed entirely for the first condition and get caught out by the second.
Suppliers prioritise the customers who make their business more predictable and their week easier. In practice: accurate forecasts, stable specifications, payment on time, low complaint volume, a named contact who answers the phone, and a commercial history without opportunistic squeezing during the last shortage. Suppliers keep score. Suppliers have long memories. A reverse auction run in 2015 gets remembered in 2026.
So the answer is both. Preference gets earned and engineered.
Earned: reputation, consistency, behaviour during previous shortages.
Engineered: forecast accuracy, specification flexibility, volume commitment, smoother order patterns, joint planning cadence, one clear point of contact. All measurable. All improvable inside two quarters.
The step most teams skip is looking at themselves the way a supplier does. Score your own account. How attractive are you to serve, and how important is your spend to their business? A $2m spend inside a $4bn supplier buys you nothing in a shortage, whatever your leverage looks like on paper. Knowing your real position moves you from demanding to designing.
A message to your younger self
Learn the plant, not the spreadsheet.
At 18 I thought procurement was about price. I spent my energy on the negotiation, the index, the forecast. The people who taught me most were process engineers and plant schedulers, and I took years to work out why. Every commodity decision ends up as a physical constraint on a line. A cheaper oil changes the fry profile. A different sugar changes crystallisation. Savings on paper turn into scrap on the floor.
Second thing: write down what you know. I carried fifteen years of category knowledge in my head and lost most of the detail every time I changed roles. Our industry does the same at scale. Institutional memory walks out of the door with every retirement.
I would tell him both. He would ignore both and learn them the hard way anyway. Some lessons need scar tissue.
Steve Tucker is Co-Founder and CEO of Kodiact, an AI-native procurement intelligence platform built for direct materials in food, beverage and CPG. He began his career in commodity procurement at Mars and has spent thirty years in enterprise software.
To find out more about the upcoming Agrifood Intelligence Summit, click here.
Written by Said Sahardeed
