AI didn’t negotiate my deal. It made me better at negotiating it.
“It’s not possible.”
Anyone who has spent time in sourcing has heard some version of those words. In this case, the product was simple: a 150 GSM, 60/40 CVC T-shirt. My target was $1.98. The supplier’s first quote from China was $2.65, almost 34% above target.
Read also: Inflation-Proofing Your Supply Chain: Finding Hidden Margins in Niche Logistics
A gap that size isn’t closed by asking a factory to “check once more.” But I also knew my target wasn’t imaginary. We had benchmarked the program across factories in China, India, Pakistan and Bangladesh, and each country came with its own cost structure, raw-material base and lead-time equation. I wasn’t forcing an arbitrary number onto a supplier. I needed to understand why this particular T-shirt cost $2.65, and this time I had new tools to help me find out.
When “Not Possible” Became a Cost Breakdown
Instead of asking for another discount, I asked for the open costing: fabric, yarn, consumption, wastage, trims, testing, packaging, overhead and margin. Then I started taking those numbers apart.
I used ChatGPT to analyze the costing and break individual assumptions into smaller parts. Claude helped on the technical side, with drawings and specification detail, so that loose ends in the product could be caught before they quietly turned into cost.
The more closely we examined the product, the less useful “Can you reduce the price?” became, and better questions replaced it. What consumption was the fabric cost built on, and how much wastage was included? Was every test being costed actually required for this program? Did every component need a nominated supplier, or were acceptable alternatives available? What would a higher MOQ do to the fabric price? Could we pack more efficiently and reduce CBM without compromising the product?
Those questions uncovered options that another discount request never would have. Fabric economics improved with volume. Some testing and nominated-supplier assumptions turned out not to be requirements. Tighter packaging improved cube utilization. Wastage and consumption could be challenged rather than accepted as fixed. None of those changes closed the gap alone, but together they brought the price from $2.65 to $2.10, without reducing the negotiation to a demand that the factory simply sacrifice its margin.
AI didn’t negotiate 55 cents off a T-shirt. It helped me stop treating 67 cents as one number.
The Negotiation Starts Before the Quote Arrives
We tend to think negotiation begins when a quotation lands in the inbox. In sourcing, it starts much earlier. A poorly defined product creates a poorly defined negotiation. If the construction is vague or the buyer doesn’t know which elements drive cost, there is no intelligent way to challenge a price.
This is where AI has changed my workflow most. I can now interrogate technical information, develop specifications and visualize construction options before the first sample round. The point isn’t to let AI make the sourcing decision. It’s to arrive at that decision knowing considerably more.
Four Countries, Four Supply Chains
One idea from Harvard Business School Online’s Negotiation Mastery course has stayed with me: BATNA, the best alternative to a negotiated agreement. In a simple negotiation, that alternative might be another supplier. Global sourcing is rarely that simple, and for this T-shirt, comparing four FOB quotes wouldn’t have told me which option was actually best.
Bangladesh could make the garment, but the polyester yarn and certain trims would still come from China. That added transactions, margins, transit time and logistics cost before production even began. Pakistan offered different raw-material advantages, including recycled cotton for this program, though some trims still came from China. China presented a different challenge: cotton traceability and compliance requirements for U.S.-bound goods affected how raw materials had to be sourced and documented. India gave us another manufacturing benchmark with its own economics.
So the BATNA question was never “Which factory gives me the next-lowest FOB?” It was “Which supply chain gives me the best commercial alternative?”
A $1.98 T-shirt made in China and a $1.98 T-shirt made somewhere else are not necessarily the same $1.98 T-shirt. One supply chain may need raw materials to cross another border before cutting starts. Another may add days in transit or carry different duty exposure, capacity limits or MOQ economics. A lower duty rate looks attractive until imported inputs, longer raw-material lead times and extra intermediaries eat into it. The cheapest factory quotation and the cheapest supply chain are not necessarily the same thing.
AI makes this kind of analysis far faster. Given the right sourcing data, it can help structure comparisons of material origin, duty exposure, transit time and alternative sourcing scenarios, and refresh those comparisons as conditions change. In today’s trade environment, a comparison can go stale quickly. But knowing your alternatives and choosing among them are different skills, and only the first one gets faster.
Yes, AI Can Negotiate
There is a complication I don’t want to dodge: AI no longer just helps people prepare for negotiations. It can conduct some of them. Autonomous procurement platforms are already designed to negotiate with suppliers within objectives and guardrails set by procurement teams, particularly where commercial variables and acceptable boundaries can be clearly defined.
For tail spend, repeat purchases and deals with structured parameters, that is a real capability, operating at a scale no human procurement team could replicate. Strategic sourcing raises a different question. It’s no longer whether AI can negotiate, but which negotiations we should hand it, and which decisions still need human commercial judgment.
Listening to What “No” Means
When a supplier says, “I can’t do it,” that can mean many things. The price may truly be out of reach. The MOQ may be too low, or the mill may not be giving the factory a good fabric price. The production window may be tight. A nominated component may be blocking alternatives. Or it may simply mean “not under the deal currently in front of me.” Telling these apart takes technical knowledge, but it also takes listening.
Price isn’t the only thing at stake, either. If I had pushed for the last 12 cents but lost production priority or cooperation the next time something went wrong, I wouldn’t necessarily have won. The lowest price and the best deal are not always the same thing.
Once I understood the costing, my goal wasn’t to prove the supplier wrong. It was to find what could move. “Can you do better on price?” became “What happens if we change this, or if I commit to that?” Buyer and supplier stopped fighting over one slice of margin and started engineering a deal that worked for both sides.
The Governance Question
Working from a real open costing also raised an issue every sourcing organization will have to face. Open costings can contain commercially sensitive information, including material prices, supplier margins, processes and negotiated terms, and my experience made me much more conscious of that risk.
As generative AI becomes part of everyday sourcing work, companies need clear policies around what information can be entered, which tools are approved and how supplier and customer information is protected. The productivity gain is enormous, but it cannot come at the expense of confidentiality.
The Advantage Moves to the Questions
Simply “using AI” won’t stay an advantage for long. Suppliers will use it to model costs, research alternatives and prepare their own positions. Soon, saying your procurement team uses AI will sound about as remarkable as saying it uses spreadsheets. The person across the table will have the same tools you do.
What will set people apart is everything around the technology. Who gives AI better inputs and challenges its outputs? Who understands the product well enough to spot an answer that doesn’t make sense? Who connects FOB to tariffs, raw materials, inventory and lead time? Who knows which five cents matter, and which aren’t worth chasing?
One buyer asks AI for a cheaper option. Another asks why the product costs what it does, what can move and what happens across the supply chain if it does. They’re using the same technology, but the negotiations they have can look completely different.
Where the Spreadsheet Ends
AI didn’t produce my deal. It helped me understand the product, tighten the specification, interrogate the costing and compare four countries as complete supply chains rather than four FOB quotes. But I still had to decide which questions mattered, what the supplier could realistically change, what I would concede and whether a deal was worth making at all.
That is why we stopped at $2.10 rather than forcing the conversation all the way to $1.98. We had closed 55 cents of a 67-cent gap, and the remaining 12 cents had to be weighed against everything else the supplier and the supply chain brought to the program. Once landed cost was factored in, none of the alternatives beat $2.10 from China.
At some point, negotiation stops being about whether another cent can technically be removed. It becomes a judgment about whether removing it creates more value than it destroys.
That brings me back to BATNA. Negotiating power doesn’t come from wanting the deal badly enough. It comes from understanding your alternatives well enough that you don’t have to accept the wrong one.
AI can raise our information quotient, but it can’t make the judgment call for us. That still comes from experience: knowing the product, knowing the supplier, understanding your alternatives and, sometimes, knowing when the last 12 cents aren’t worth chasing.
AI didn’t negotiate my deal. It made me much better prepared to negotiate it. And for me, that’s where its real value in sourcing begins.
Author bio
Shailja Agrawal is Vice President of Sourcing & Production and a global sourcing and manufacturing executive with nearly two decades of experience building and managing supply chains across Asia, Mexico, Central America and North America. Her work spans sourcing strategy, manufacturing economics, tariffs and trade, compliance, nearshoring and supplier negotiations, with a focus on the decisions behind what products really cost and where they should be made. She has completed Negotiation Mastery through Harvard Business School Online and writes from the operator’s side of the global supply chain.