Bardia Shahali: eight-figure AI deals close in two weeks

Shahali's comments come from a conversation at Office Hours earlier this week, where he compared building sales teams at the application layer versus the infrastructure layer. At Fireworks, he said, customers are AI-native companies growing at extraordinary rates, so single deals can reach the tens of millions with a very fast sales cycle. The counterpoint he raised is competition: every inference provider chases the same accounts, making the stakes higher and the competition more intense.

He also described how annual planning becomes difficult in such a dynamic environment. Products evolve faster, model capabilities reach a new zenith every 45 days, and the competitive landscape shifts quickly, which is why his team moved to setting comp every quarter. On staffing, he warned that a small team risks not putting in the extra effort for outbound and expansion. At the infrastructure layer, forward-deployed engineers and sales engineers pair with account executives to deliver a highly complex product, while at the application layer the account executive and customer success manager deliver software similar to the previous era but with higher productivity and typically faster adoption and expansion.