For decades, pricing a public tender was the most senior commercial task in a supplier organization. It required reading the technical specification, intuiting what the buyer actually wanted, mapping it to your own cost base, applying gut-feel competitor estimates, and landing on a number that was low enough to win but high enough to leave margin. It was the place where the most experience, the most context, and the most judgment converged. AI was supposed to come for the easy stuff first. Pricing was meant to be safe.
It isn't anymore. The reason has nothing to do with AI being smarter than the human pricer — and everything to do with the data structure now being accessible to a model that wasn't accessible to a person.
Three datasets, when combined, change the game. First: the supplier's own cost catalog — line-item material costs, labor rates, overhead multipliers, margin rules — expressed as a structured spreadsheet. Second: the tender's bill of quantities or technical specification, extracted from PDF into structured line items. Third: the historical award database for the same CPV codes in the same region — winning prices on similar contracts, going back several years, normalized for contract size.
Until 2024 those three datasets lived in three different worlds. The cost catalog was a spreadsheet on someone's laptop. The tender was a PDF nobody had time to read carefully. The historical awards were buried in TED's structured-data dumps that you had to write a parser for. A human had to bridge all three by memory, every time. Now an LLM bridges them in seconds.
What a modern autopricing engine does: it reads the tender PDF, extracts every line item, fuzzy-matches each line to your cost catalog, applies your standard margin rules, generates a fully priced bid in Excel and Word, and shows you the competitor band based on past awards in the same code and region. The output is a draft. A human still reviews, sanity-checks the outlier lines, adjusts the strategic premium up or down. But the 90% of the work that used to take a senior person two days now takes that person twenty minutes of review.
The competitive consequence is brutal for suppliers who don't adopt. Pricing has always been the bottleneck — most teams could only seriously price two or three bids per week. With autopricing, the same team can price ten. Their bid volume goes up by the factor their pricing capacity used to limit them by. Their win rate stays the same. Revenue triples without hiring.
The suppliers who treat autopricing as a productivity gain will gain market share. The suppliers who treat it as a threat will be the ones losing share to them. The transition window is closing — fast.